diff --git "a/trainer_state.json" "b/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/trainer_state.json"
@@ -0,0 +1,206421 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.9244755244755245,
+  "eval_steps": 500,
+  "global_step": 34400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.863932898415657e-08,
+      "loss": 1.4191,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.727865796831314e-08,
+      "loss": 1.3378,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.5917986952469716e-08,
+      "loss": 1.2637,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.455731593662628e-08,
+      "loss": 1.2905,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.319664492078285e-08,
+      "loss": 1.2011,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1183597390493943e-07,
+      "loss": 1.3334,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3047530288909601e-07,
+      "loss": 1.3742,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4911463187325257e-07,
+      "loss": 1.156,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6775396085740915e-07,
+      "loss": 1.1699,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.863932898415657e-07,
+      "loss": 1.4714,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.050326188257223e-07,
+      "loss": 1.1523,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2367194780987887e-07,
+      "loss": 1.2572,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4231127679403545e-07,
+      "loss": 1.2592,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6095060577819203e-07,
+      "loss": 1.1911,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7958993476234855e-07,
+      "loss": 1.3532,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9822926374650514e-07,
+      "loss": 1.2609,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.168685927306617e-07,
+      "loss": 1.0892,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.355079217148183e-07,
+      "loss": 1.5378,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.541472506989749e-07,
+      "loss": 1.2088,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.727865796831314e-07,
+      "loss": 1.1546,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.91425908667288e-07,
+      "loss": 1.2144,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.100652376514446e-07,
+      "loss": 1.096,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.2870456663560115e-07,
+      "loss": 1.2098,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4734389561975773e-07,
+      "loss": 1.4649,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6598322460391426e-07,
+      "loss": 1.1227,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.846225535880709e-07,
+      "loss": 1.0912,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.032618825722274e-07,
+      "loss": 1.3319,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.219012115563841e-07,
+      "loss": 1.2784,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.405405405405406e-07,
+      "loss": 1.4131,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.591798695246971e-07,
+      "loss": 1.0508,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.778191985088537e-07,
+      "loss": 1.3707,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.964585274930103e-07,
+      "loss": 1.0212,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.150978564771669e-07,
+      "loss": 1.0727,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.337371854613234e-07,
+      "loss": 1.1946,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.5237651444548e-07,
+      "loss": 1.2556,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.710158434296366e-07,
+      "loss": 1.2909,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.896551724137931e-07,
+      "loss": 1.1846,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.082945013979498e-07,
+      "loss": 1.0411,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.269338303821064e-07,
+      "loss": 1.0576,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.455731593662628e-07,
+      "loss": 1.2003,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.642124883504194e-07,
+      "loss": 1.0828,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.82851817334576e-07,
+      "loss": 1.0911,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.014911463187326e-07,
+      "loss": 1.0298,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.201304753028892e-07,
+      "loss": 1.0777,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.387698042870457e-07,
+      "loss": 1.1043,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.574091332712023e-07,
+      "loss": 1.0993,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.760484622553589e-07,
+      "loss": 1.1877,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.946877912395155e-07,
+      "loss": 1.1761,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.13327120223672e-07,
+      "loss": 1.0614,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.319664492078285e-07,
+      "loss": 0.9935,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.506057781919852e-07,
+      "loss": 0.9439,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.692451071761418e-07,
+      "loss": 1.0239,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.878844361602984e-07,
+      "loss": 1.1991,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0065237651444548e-06,
+      "loss": 1.1677,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0251630941286115e-06,
+      "loss": 1.0436,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0438024231127681e-06,
+      "loss": 0.9414,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0624417520969245e-06,
+      "loss": 1.0669,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0810810810810812e-06,
+      "loss": 1.0032,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0997204100652378e-06,
+      "loss": 0.8486,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1183597390493942e-06,
+      "loss": 1.0217,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1369990680335509e-06,
+      "loss": 1.0882,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1556383970177075e-06,
+      "loss": 1.2013,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.174277726001864e-06,
+      "loss": 1.0268,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1929170549860205e-06,
+      "loss": 0.9295,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2115563839701772e-06,
+      "loss": 1.0865,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2301957129543338e-06,
+      "loss": 0.9602,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2488350419384904e-06,
+      "loss": 1.0743,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2674743709226469e-06,
+      "loss": 1.1283,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2861136999068035e-06,
+      "loss": 1.2138,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.30475302889096e-06,
+      "loss": 0.9242,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3233923578751166e-06,
+      "loss": 1.0194,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3420316868592732e-06,
+      "loss": 0.9206,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3606710158434296e-06,
+      "loss": 1.1392,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3793103448275862e-06,
+      "loss": 0.8832,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3979496738117429e-06,
+      "loss": 0.825,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4165890027958995e-06,
+      "loss": 0.9796,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4352283317800561e-06,
+      "loss": 0.8843,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4538676607642128e-06,
+      "loss": 0.8644,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4725069897483692e-06,
+      "loss": 0.9455,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4911463187325256e-06,
+      "loss": 0.8155,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5097856477166823e-06,
+      "loss": 0.7956,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5284249767008389e-06,
+      "loss": 0.9413,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5470643056849955e-06,
+      "loss": 0.9548,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.565703634669152e-06,
+      "loss": 0.8873,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5843429636533086e-06,
+      "loss": 0.8804,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6029822926374652e-06,
+      "loss": 0.8294,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6216216216216219e-06,
+      "loss": 0.8898,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6402609506057785e-06,
+      "loss": 0.8004,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6589002795899347e-06,
+      "loss": 0.9534,
+      "step": 89
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.6775396085740913e-06,
+      "loss": 0.9506,
+      "step": 90
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.696178937558248e-06,
+      "loss": 0.8872,
+      "step": 91
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7148182665424046e-06,
+      "loss": 0.8475,
+      "step": 92
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7334575955265612e-06,
+      "loss": 0.9569,
+      "step": 93
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7520969245107179e-06,
+      "loss": 1.1274,
+      "step": 94
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.7707362534948743e-06,
+      "loss": 0.8522,
+      "step": 95
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.789375582479031e-06,
+      "loss": 0.8561,
+      "step": 96
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8080149114631876e-06,
+      "loss": 0.8432,
+      "step": 97
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.826654240447344e-06,
+      "loss": 1.059,
+      "step": 98
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8452935694315006e-06,
+      "loss": 0.8608,
+      "step": 99
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.863932898415657e-06,
+      "loss": 0.9579,
+      "step": 100
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.8825722273998137e-06,
+      "loss": 0.8796,
+      "step": 101
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9012115563839703e-06,
+      "loss": 0.9657,
+      "step": 102
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.919850885368127e-06,
+      "loss": 0.9128,
+      "step": 103
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9384902143522836e-06,
+      "loss": 0.8852,
+      "step": 104
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.95712954333644e-06,
+      "loss": 0.961,
+      "step": 105
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.975768872320597e-06,
+      "loss": 0.9327,
+      "step": 106
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 1.9944082013047535e-06,
+      "loss": 1.008,
+      "step": 107
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0130475302889097e-06,
+      "loss": 0.8776,
+      "step": 108
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0316868592730663e-06,
+      "loss": 0.8697,
+      "step": 109
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.050326188257223e-06,
+      "loss": 0.7943,
+      "step": 110
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0689655172413796e-06,
+      "loss": 0.808,
+      "step": 111
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.0876048462255362e-06,
+      "loss": 0.852,
+      "step": 112
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.106244175209693e-06,
+      "loss": 1.0335,
+      "step": 113
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.124883504193849e-06,
+      "loss": 0.9101,
+      "step": 114
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1435228331780057e-06,
+      "loss": 0.8753,
+      "step": 115
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1621621621621623e-06,
+      "loss": 0.9791,
+      "step": 116
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.180801491146319e-06,
+      "loss": 0.8387,
+      "step": 117
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.1994408201304756e-06,
+      "loss": 0.8922,
+      "step": 118
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.218080149114632e-06,
+      "loss": 0.878,
+      "step": 119
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2367194780987884e-06,
+      "loss": 0.8194,
+      "step": 120
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.255358807082945e-06,
+      "loss": 0.8512,
+      "step": 121
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2739981360671017e-06,
+      "loss": 0.9007,
+      "step": 122
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.2926374650512583e-06,
+      "loss": 0.9367,
+      "step": 123
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.311276794035415e-06,
+      "loss": 0.8468,
+      "step": 124
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3299161230195716e-06,
+      "loss": 0.8093,
+      "step": 125
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.348555452003728e-06,
+      "loss": 0.8727,
+      "step": 126
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.3671947809878845e-06,
+      "loss": 0.918,
+      "step": 127
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.385834109972041e-06,
+      "loss": 0.8695,
+      "step": 128
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4044734389561977e-06,
+      "loss": 0.8262,
+      "step": 129
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4231127679403544e-06,
+      "loss": 0.8887,
+      "step": 130
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.441752096924511e-06,
+      "loss": 0.7677,
+      "step": 131
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4603914259086676e-06,
+      "loss": 0.8485,
+      "step": 132
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.4790307548928243e-06,
+      "loss": 0.9008,
+      "step": 133
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.497670083876981e-06,
+      "loss": 1.0456,
+      "step": 134
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.516309412861137e-06,
+      "loss": 0.9316,
+      "step": 135
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5349487418452937e-06,
+      "loss": 0.8401,
+      "step": 136
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5535880708294504e-06,
+      "loss": 0.8698,
+      "step": 137
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.572227399813607e-06,
+      "loss": 0.9327,
+      "step": 138
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.5908667287977636e-06,
+      "loss": 0.8208,
+      "step": 139
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.60950605778192e-06,
+      "loss": 0.871,
+      "step": 140
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6281453867660765e-06,
+      "loss": 0.8568,
+      "step": 141
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.646784715750233e-06,
+      "loss": 0.9191,
+      "step": 142
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6654240447343897e-06,
+      "loss": 0.9627,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.6840633737185464e-06,
+      "loss": 0.8656,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.702702702702703e-06,
+      "loss": 0.926,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7213420316868592e-06,
+      "loss": 0.8513,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7399813606710163e-06,
+      "loss": 0.8967,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7586206896551725e-06,
+      "loss": 0.8308,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7772600186393296e-06,
+      "loss": 0.8926,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.7958993476234858e-06,
+      "loss": 0.8489,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.814538676607642e-06,
+      "loss": 0.8243,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.833178005591799e-06,
+      "loss": 0.7025,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8518173345759552e-06,
+      "loss": 0.8729,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8704566635601123e-06,
+      "loss": 0.8187,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.8890959925442685e-06,
+      "loss": 0.8563,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9077353215284256e-06,
+      "loss": 0.8238,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9263746505125818e-06,
+      "loss": 0.8002,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9450139794967384e-06,
+      "loss": 0.8414,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.963653308480895e-06,
+      "loss": 0.7769,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 2.9822926374650513e-06,
+      "loss": 0.663,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0009319664492083e-06,
+      "loss": 0.7825,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0195712954333645e-06,
+      "loss": 0.7812,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.038210624417521e-06,
+      "loss": 0.7454,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0568499534016778e-06,
+      "loss": 0.8548,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.0754892823858344e-06,
+      "loss": 0.9134,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.094128611369991e-06,
+      "loss": 0.8746,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1127679403541477e-06,
+      "loss": 0.9472,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.131407269338304e-06,
+      "loss": 0.7378,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.1500465983224605e-06,
+      "loss": 0.8512,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.168685927306617e-06,
+      "loss": 0.8695,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.187325256290774e-06,
+      "loss": 0.8614,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2059645852749304e-06,
+      "loss": 0.8662,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2246039142590866e-06,
+      "loss": 0.7448,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2432432432432437e-06,
+      "loss": 0.7202,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.2618825722274e-06,
+      "loss": 0.7912,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.280521901211557e-06,
+      "loss": 0.818,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.299161230195713e-06,
+      "loss": 0.781,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3178005591798694e-06,
+      "loss": 0.8019,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3364398881640264e-06,
+      "loss": 0.8601,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3550792171481827e-06,
+      "loss": 0.8248,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3737185461323397e-06,
+      "loss": 0.7604,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.392357875116496e-06,
+      "loss": 0.6962,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.410997204100653e-06,
+      "loss": 0.8235,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.429636533084809e-06,
+      "loss": 0.852,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.448275862068966e-06,
+      "loss": 0.9576,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4669151910531225e-06,
+      "loss": 0.7913,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4855545200372787e-06,
+      "loss": 0.9454,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5041938490214357e-06,
+      "loss": 0.7428,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.522833178005592e-06,
+      "loss": 0.9378,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5414725069897486e-06,
+      "loss": 0.8162,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.560111835973905e-06,
+      "loss": 0.7334,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.578751164958062e-06,
+      "loss": 0.8569,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5973904939422185e-06,
+      "loss": 0.6515,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.616029822926375e-06,
+      "loss": 0.7271,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6346691519105313e-06,
+      "loss": 0.8409,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.653308480894688e-06,
+      "loss": 0.7559,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6719478098788446e-06,
+      "loss": 0.9145,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6905871388630012e-06,
+      "loss": 0.9357,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.709226467847158e-06,
+      "loss": 0.8085,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.727865796831314e-06,
+      "loss": 0.7099,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.746505125815471e-06,
+      "loss": 0.8536,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7651444547996273e-06,
+      "loss": 0.7679,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7837837837837844e-06,
+      "loss": 0.814,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.8024231127679406e-06,
+      "loss": 0.8453,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.821062441752098e-06,
+      "loss": 0.7753,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.839701770736254e-06,
+      "loss": 0.7149,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.85834109972041e-06,
+      "loss": 0.7316,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.876980428704567e-06,
+      "loss": 0.8062,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.895619757688723e-06,
+      "loss": 0.8416,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.91425908667288e-06,
+      "loss": 0.7999,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.932898415657037e-06,
+      "loss": 0.8119,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.951537744641194e-06,
+      "loss": 0.7708,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.97017707362535e-06,
+      "loss": 0.7575,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.988816402609507e-06,
+      "loss": 0.8059,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.007455731593663e-06,
+      "loss": 0.7958,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.026095060577819e-06,
+      "loss": 0.7649,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.044734389561976e-06,
+      "loss": 0.7681,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.063373718546133e-06,
+      "loss": 0.7557,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.08201304753029e-06,
+      "loss": 0.9817,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.100652376514446e-06,
+      "loss": 0.7287,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.119291705498603e-06,
+      "loss": 0.9004,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.137931034482759e-06,
+      "loss": 0.8906,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.156570363466915e-06,
+      "loss": 0.7532,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.1752096924510724e-06,
+      "loss": 0.8205,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.193849021435229e-06,
+      "loss": 0.8864,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.212488350419386e-06,
+      "loss": 0.7549,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.231127679403542e-06,
+      "loss": 0.6755,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.249767008387698e-06,
+      "loss": 0.7446,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.268406337371855e-06,
+      "loss": 0.799,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.287045666356011e-06,
+      "loss": 0.6873,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3056849953401684e-06,
+      "loss": 0.778,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.324324324324325e-06,
+      "loss": 0.7302,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.342963653308481e-06,
+      "loss": 0.903,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.361602982292638e-06,
+      "loss": 0.7925,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.380242311276794e-06,
+      "loss": 0.8926,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.398881640260951e-06,
+      "loss": 0.8272,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.417520969245107e-06,
+      "loss": 0.8699,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.436160298229264e-06,
+      "loss": 0.9494,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.454799627213421e-06,
+      "loss": 0.855,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.473438956197577e-06,
+      "loss": 0.7886,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.492078285181734e-06,
+      "loss": 0.7844,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.51071761416589e-06,
+      "loss": 0.685,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.529356943150046e-06,
+      "loss": 0.8508,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.547996272134203e-06,
+      "loss": 0.7477,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.56663560111836e-06,
+      "loss": 0.7852,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.585274930102517e-06,
+      "loss": 0.7457,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.603914259086673e-06,
+      "loss": 0.7812,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.62255358807083e-06,
+      "loss": 0.8331,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.641192917054986e-06,
+      "loss": 0.8753,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.659832246039143e-06,
+      "loss": 0.7482,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.6784715750232994e-06,
+      "loss": 0.6512,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.697110904007456e-06,
+      "loss": 0.7527,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.715750232991613e-06,
+      "loss": 0.9291,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.734389561975769e-06,
+      "loss": 0.7907,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.753028890959926e-06,
+      "loss": 0.6841,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.771668219944082e-06,
+      "loss": 0.7445,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.790307548928239e-06,
+      "loss": 0.8466,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8089468779123954e-06,
+      "loss": 0.8309,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.8275862068965525e-06,
+      "loss": 0.85,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.846225535880709e-06,
+      "loss": 0.7796,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.864864864864866e-06,
+      "loss": 0.8119,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.883504193849022e-06,
+      "loss": 0.8055,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.902143522833178e-06,
+      "loss": 0.6347,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.920782851817335e-06,
+      "loss": 0.7901,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9394221808014915e-06,
+      "loss": 0.8021,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.9580615097856485e-06,
+      "loss": 0.9081,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.976700838769805e-06,
+      "loss": 0.7647,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.995340167753962e-06,
+      "loss": 0.6963,
+      "step": 268
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.013979496738118e-06,
+      "loss": 0.7759,
+      "step": 269
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.032618825722274e-06,
+      "loss": 0.9086,
+      "step": 270
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.051258154706431e-06,
+      "loss": 0.8515,
+      "step": 271
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.0698974836905875e-06,
+      "loss": 0.813,
+      "step": 272
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.088536812674744e-06,
+      "loss": 0.7829,
+      "step": 273
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.107176141658901e-06,
+      "loss": 0.9163,
+      "step": 274
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.125815470643058e-06,
+      "loss": 0.7178,
+      "step": 275
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.144454799627214e-06,
+      "loss": 0.791,
+      "step": 276
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.16309412861137e-06,
+      "loss": 0.8053,
+      "step": 277
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.181733457595527e-06,
+      "loss": 0.8357,
+      "step": 278
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.2003727865796835e-06,
+      "loss": 0.8337,
+      "step": 279
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.21901211556384e-06,
+      "loss": 0.8048,
+      "step": 280
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.237651444547997e-06,
+      "loss": 0.798,
+      "step": 281
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.256290773532153e-06,
+      "loss": 0.9345,
+      "step": 282
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.27493010251631e-06,
+      "loss": 0.8052,
+      "step": 283
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.293569431500466e-06,
+      "loss": 0.8627,
+      "step": 284
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.312208760484623e-06,
+      "loss": 0.7052,
+      "step": 285
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3308480894687795e-06,
+      "loss": 0.7398,
+      "step": 286
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.3494874184529366e-06,
+      "loss": 0.6799,
+      "step": 287
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.368126747437093e-06,
+      "loss": 0.8094,
+      "step": 288
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.386766076421249e-06,
+      "loss": 0.7523,
+      "step": 289
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.405405405405406e-06,
+      "loss": 0.7807,
+      "step": 290
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.424044734389562e-06,
+      "loss": 0.8897,
+      "step": 291
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4426840633737185e-06,
+      "loss": 0.7865,
+      "step": 292
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4613233923578755e-06,
+      "loss": 0.8911,
+      "step": 293
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.4799627213420326e-06,
+      "loss": 0.6529,
+      "step": 294
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.498602050326188e-06,
+      "loss": 0.8404,
+      "step": 295
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.517241379310345e-06,
+      "loss": 0.7355,
+      "step": 296
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.535880708294502e-06,
+      "loss": 0.7831,
+      "step": 297
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.554520037278659e-06,
+      "loss": 0.8167,
+      "step": 298
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5731593662628145e-06,
+      "loss": 0.7253,
+      "step": 299
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.5917986952469715e-06,
+      "loss": 0.8106,
+      "step": 300
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.610438024231129e-06,
+      "loss": 0.7802,
+      "step": 301
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.629077353215284e-06,
+      "loss": 0.7586,
+      "step": 302
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.647716682199441e-06,
+      "loss": 0.6697,
+      "step": 303
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.666356011183598e-06,
+      "loss": 0.7119,
+      "step": 304
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.684995340167755e-06,
+      "loss": 0.8118,
+      "step": 305
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7036346691519105e-06,
+      "loss": 0.6273,
+      "step": 306
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.7222739981360675e-06,
+      "loss": 0.609,
+      "step": 307
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.740913327120225e-06,
+      "loss": 0.8131,
+      "step": 308
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.75955265610438e-06,
+      "loss": 0.6911,
+      "step": 309
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.778191985088537e-06,
+      "loss": 0.7181,
+      "step": 310
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.796831314072694e-06,
+      "loss": 0.6848,
+      "step": 311
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.815470643056851e-06,
+      "loss": 0.7599,
+      "step": 312
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8341099720410065e-06,
+      "loss": 0.8206,
+      "step": 313
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.8527493010251635e-06,
+      "loss": 0.7095,
+      "step": 314
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.871388630009321e-06,
+      "loss": 0.748,
+      "step": 315
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.890027958993477e-06,
+      "loss": 0.809,
+      "step": 316
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.908667287977633e-06,
+      "loss": 0.7856,
+      "step": 317
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.92730661696179e-06,
+      "loss": 0.76,
+      "step": 318
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.945945945945947e-06,
+      "loss": 0.8019,
+      "step": 319
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9645852749301025e-06,
+      "loss": 0.8059,
+      "step": 320
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 5.9832246039142596e-06,
+      "loss": 0.7875,
+      "step": 321
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.001863932898417e-06,
+      "loss": 0.751,
+      "step": 322
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.020503261882573e-06,
+      "loss": 0.6414,
+      "step": 323
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.039142590866729e-06,
+      "loss": 0.893,
+      "step": 324
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.057781919850886e-06,
+      "loss": 0.7376,
+      "step": 325
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.076421248835042e-06,
+      "loss": 0.8213,
+      "step": 326
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.0950605778191985e-06,
+      "loss": 0.7372,
+      "step": 327
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.1136999068033556e-06,
+      "loss": 0.8257,
+      "step": 328
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.132339235787513e-06,
+      "loss": 0.8143,
+      "step": 329
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.150978564771669e-06,
+      "loss": 0.6855,
+      "step": 330
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.169617893755825e-06,
+      "loss": 0.7565,
+      "step": 331
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.188257222739982e-06,
+      "loss": 0.8408,
+      "step": 332
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.206896551724138e-06,
+      "loss": 0.7946,
+      "step": 333
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.225535880708295e-06,
+      "loss": 0.7982,
+      "step": 334
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.244175209692452e-06,
+      "loss": 0.816,
+      "step": 335
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.262814538676608e-06,
+      "loss": 0.8118,
+      "step": 336
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.281453867660765e-06,
+      "loss": 0.769,
+      "step": 337
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.300093196644921e-06,
+      "loss": 0.7793,
+      "step": 338
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.318732525629077e-06,
+      "loss": 0.6314,
+      "step": 339
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.337371854613234e-06,
+      "loss": 0.7936,
+      "step": 340
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.356011183597391e-06,
+      "loss": 0.782,
+      "step": 341
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.374650512581548e-06,
+      "loss": 0.8044,
+      "step": 342
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.393289841565704e-06,
+      "loss": 0.7995,
+      "step": 343
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.411929170549861e-06,
+      "loss": 0.9036,
+      "step": 344
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.430568499534018e-06,
+      "loss": 0.8418,
+      "step": 345
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.449207828518173e-06,
+      "loss": 0.9111,
+      "step": 346
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.46784715750233e-06,
+      "loss": 0.7033,
+      "step": 347
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.486486486486487e-06,
+      "loss": 0.812,
+      "step": 348
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.505125815470643e-06,
+      "loss": 0.7802,
+      "step": 349
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.5237651444548e-06,
+      "loss": 0.6843,
+      "step": 350
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.542404473438957e-06,
+      "loss": 0.6984,
+      "step": 351
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.561043802423114e-06,
+      "loss": 0.7454,
+      "step": 352
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.579683131407269e-06,
+      "loss": 0.7745,
+      "step": 353
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.598322460391426e-06,
+      "loss": 0.6918,
+      "step": 354
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.616961789375583e-06,
+      "loss": 0.7124,
+      "step": 355
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.635601118359739e-06,
+      "loss": 0.6892,
+      "step": 356
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.654240447343896e-06,
+      "loss": 0.717,
+      "step": 357
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.672879776328053e-06,
+      "loss": 0.8014,
+      "step": 358
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.69151910531221e-06,
+      "loss": 0.7679,
+      "step": 359
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.710158434296365e-06,
+      "loss": 0.9503,
+      "step": 360
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.728797763280522e-06,
+      "loss": 0.7293,
+      "step": 361
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.7474370922646794e-06,
+      "loss": 0.7196,
+      "step": 362
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.7660764212488365e-06,
+      "loss": 0.7813,
+      "step": 363
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.784715750232992e-06,
+      "loss": 0.777,
+      "step": 364
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.803355079217149e-06,
+      "loss": 0.7556,
+      "step": 365
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.821994408201306e-06,
+      "loss": 0.7738,
+      "step": 366
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.840633737185461e-06,
+      "loss": 0.7996,
+      "step": 367
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.859273066169618e-06,
+      "loss": 0.7954,
+      "step": 368
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.8779123951537754e-06,
+      "loss": 0.8881,
+      "step": 369
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.896551724137932e-06,
+      "loss": 0.7735,
+      "step": 370
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.915191053122088e-06,
+      "loss": 0.8142,
+      "step": 371
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.933830382106245e-06,
+      "loss": 0.8237,
+      "step": 372
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.952469711090402e-06,
+      "loss": 0.7349,
+      "step": 373
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.971109040074557e-06,
+      "loss": 0.726,
+      "step": 374
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 6.989748369058714e-06,
+      "loss": 0.7524,
+      "step": 375
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.0083876980428715e-06,
+      "loss": 0.6812,
+      "step": 376
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.027027027027028e-06,
+      "loss": 0.6784,
+      "step": 377
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.045666356011184e-06,
+      "loss": 0.902,
+      "step": 378
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.064305684995341e-06,
+      "loss": 0.7209,
+      "step": 379
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.082945013979497e-06,
+      "loss": 0.7343,
+      "step": 380
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.101584342963654e-06,
+      "loss": 0.802,
+      "step": 381
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.12022367194781e-06,
+      "loss": 0.7049,
+      "step": 382
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.138863000931967e-06,
+      "loss": 0.7683,
+      "step": 383
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.157502329916124e-06,
+      "loss": 0.7318,
+      "step": 384
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.17614165890028e-06,
+      "loss": 0.7129,
+      "step": 385
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.194780987884437e-06,
+      "loss": 0.7412,
+      "step": 386
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.213420316868593e-06,
+      "loss": 0.6679,
+      "step": 387
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.23205964585275e-06,
+      "loss": 0.784,
+      "step": 388
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.2506989748369064e-06,
+      "loss": 0.7799,
+      "step": 389
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.269338303821063e-06,
+      "loss": 0.8005,
+      "step": 390
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.28797763280522e-06,
+      "loss": 0.8158,
+      "step": 391
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.306616961789376e-06,
+      "loss": 0.7068,
+      "step": 392
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.325256290773532e-06,
+      "loss": 0.7453,
+      "step": 393
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.343895619757689e-06,
+      "loss": 0.7888,
+      "step": 394
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.362534948741846e-06,
+      "loss": 0.6547,
+      "step": 395
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.3811742777260024e-06,
+      "loss": 0.8301,
+      "step": 396
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.399813606710159e-06,
+      "loss": 0.7533,
+      "step": 397
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.418452935694316e-06,
+      "loss": 0.6233,
+      "step": 398
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.437092264678473e-06,
+      "loss": 0.739,
+      "step": 399
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.455731593662628e-06,
+      "loss": 0.7281,
+      "step": 400
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.474370922646785e-06,
+      "loss": 0.6604,
+      "step": 401
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.493010251630942e-06,
+      "loss": 0.7656,
+      "step": 402
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.511649580615098e-06,
+      "loss": 0.8542,
+      "step": 403
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.530288909599255e-06,
+      "loss": 0.7358,
+      "step": 404
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.548928238583412e-06,
+      "loss": 0.7033,
+      "step": 405
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.567567567567569e-06,
+      "loss": 0.801,
+      "step": 406
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.586206896551724e-06,
+      "loss": 0.725,
+      "step": 407
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.604846225535881e-06,
+      "loss": 0.7159,
+      "step": 408
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.623485554520038e-06,
+      "loss": 0.8045,
+      "step": 409
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.642124883504195e-06,
+      "loss": 0.8217,
+      "step": 410
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.660764212488352e-06,
+      "loss": 0.7715,
+      "step": 411
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.679403541472508e-06,
+      "loss": 0.8666,
+      "step": 412
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.698042870456664e-06,
+      "loss": 0.6892,
+      "step": 413
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.71668219944082e-06,
+      "loss": 0.7831,
+      "step": 414
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.735321528424976e-06,
+      "loss": 0.7165,
+      "step": 415
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.753960857409134e-06,
+      "loss": 0.7295,
+      "step": 416
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.77260018639329e-06,
+      "loss": 0.773,
+      "step": 417
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.791239515377447e-06,
+      "loss": 0.8522,
+      "step": 418
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.809878844361603e-06,
+      "loss": 0.7229,
+      "step": 419
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.82851817334576e-06,
+      "loss": 0.7794,
+      "step": 420
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.847157502329917e-06,
+      "loss": 0.6911,
+      "step": 421
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.865796831314073e-06,
+      "loss": 0.7461,
+      "step": 422
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.88443616029823e-06,
+      "loss": 0.7577,
+      "step": 423
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.903075489282387e-06,
+      "loss": 0.7998,
+      "step": 424
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.921714818266542e-06,
+      "loss": 0.6518,
+      "step": 425
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.9403541472507e-06,
+      "loss": 0.6668,
+      "step": 426
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.958993476234856e-06,
+      "loss": 0.8423,
+      "step": 427
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.977632805219014e-06,
+      "loss": 0.7082,
+      "step": 428
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.996272134203168e-06,
+      "loss": 0.7574,
+      "step": 429
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.014911463187326e-06,
+      "loss": 0.658,
+      "step": 430
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.033550792171483e-06,
+      "loss": 0.6858,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.052190121155639e-06,
+      "loss": 0.7685,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.070829450139795e-06,
+      "loss": 0.7633,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.089468779123953e-06,
+      "loss": 0.8287,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.108108108108109e-06,
+      "loss": 0.5849,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.126747437092265e-06,
+      "loss": 0.8237,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.145386766076421e-06,
+      "loss": 0.9568,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.16402609506058e-06,
+      "loss": 0.6927,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.182665424044734e-06,
+      "loss": 0.731,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.201304753028892e-06,
+      "loss": 0.6948,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.219944082013048e-06,
+      "loss": 0.6568,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.238583410997206e-06,
+      "loss": 0.8561,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.25722273998136e-06,
+      "loss": 0.6849,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.275862068965518e-06,
+      "loss": 0.6254,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.294501397949675e-06,
+      "loss": 0.7351,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 8.31314072693383e-06,
+      "loss": 0.7572,
+      "step": 446
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.331780055917987e-06,
+      "loss": 0.7967,
+      "step": 447
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.350419384902145e-06,
+      "loss": 0.7999,
+      "step": 448
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.369058713886301e-06,
+      "loss": 0.7507,
+      "step": 449
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.387698042870457e-06,
+      "loss": 0.7159,
+      "step": 450
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.406337371854613e-06,
+      "loss": 0.7232,
+      "step": 451
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.424976700838771e-06,
+      "loss": 0.7711,
+      "step": 452
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.443616029822928e-06,
+      "loss": 0.7663,
+      "step": 453
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.462255358807084e-06,
+      "loss": 0.7395,
+      "step": 454
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.48089468779124e-06,
+      "loss": 0.7609,
+      "step": 455
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.499534016775396e-06,
+      "loss": 0.7245,
+      "step": 456
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.518173345759554e-06,
+      "loss": 0.8078,
+      "step": 457
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.53681267474371e-06,
+      "loss": 0.6932,
+      "step": 458
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.555452003727867e-06,
+      "loss": 0.7522,
+      "step": 459
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.574091332712023e-06,
+      "loss": 0.5679,
+      "step": 460
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.592730661696179e-06,
+      "loss": 0.6607,
+      "step": 461
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.611369990680337e-06,
+      "loss": 0.7986,
+      "step": 462
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.630009319664493e-06,
+      "loss": 0.7719,
+      "step": 463
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.64864864864865e-06,
+      "loss": 0.8551,
+      "step": 464
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.667287977632806e-06,
+      "loss": 0.8262,
+      "step": 465
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.685927306616962e-06,
+      "loss": 0.733,
+      "step": 466
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.70456663560112e-06,
+      "loss": 0.7649,
+      "step": 467
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.723205964585276e-06,
+      "loss": 0.7448,
+      "step": 468
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.741845293569432e-06,
+      "loss": 0.7958,
+      "step": 469
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.760484622553588e-06,
+      "loss": 0.755,
+      "step": 470
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.779123951537746e-06,
+      "loss": 0.7619,
+      "step": 471
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.797763280521902e-06,
+      "loss": 0.8423,
+      "step": 472
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.816402609506059e-06,
+      "loss": 0.8596,
+      "step": 473
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.835041938490215e-06,
+      "loss": 0.6888,
+      "step": 474
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.853681267474373e-06,
+      "loss": 0.695,
+      "step": 475
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.872320596458527e-06,
+      "loss": 0.753,
+      "step": 476
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.890959925442685e-06,
+      "loss": 0.949,
+      "step": 477
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.909599254426841e-06,
+      "loss": 0.8755,
+      "step": 478
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.928238583410998e-06,
+      "loss": 0.638,
+      "step": 479
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.946877912395154e-06,
+      "loss": 0.9076,
+      "step": 480
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.965517241379312e-06,
+      "loss": 0.7502,
+      "step": 481
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.984156570363468e-06,
+      "loss": 0.7039,
+      "step": 482
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.002795899347624e-06,
+      "loss": 0.7837,
+      "step": 483
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.02143522833178e-06,
+      "loss": 0.7975,
+      "step": 484
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.040074557315938e-06,
+      "loss": 0.7473,
+      "step": 485
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.058713886300093e-06,
+      "loss": 0.6561,
+      "step": 486
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.07735321528425e-06,
+      "loss": 0.8024,
+      "step": 487
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.095992544268407e-06,
+      "loss": 0.6172,
+      "step": 488
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.114631873252565e-06,
+      "loss": 0.8731,
+      "step": 489
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.13327120223672e-06,
+      "loss": 0.8192,
+      "step": 490
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.151910531220877e-06,
+      "loss": 0.6983,
+      "step": 491
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.170549860205033e-06,
+      "loss": 0.7597,
+      "step": 492
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.189189189189191e-06,
+      "loss": 0.7195,
+      "step": 493
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.207828518173346e-06,
+      "loss": 0.8036,
+      "step": 494
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.226467847157504e-06,
+      "loss": 0.6971,
+      "step": 495
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.24510717614166e-06,
+      "loss": 0.786,
+      "step": 496
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.263746505125816e-06,
+      "loss": 0.7045,
+      "step": 497
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.282385834109972e-06,
+      "loss": 0.7284,
+      "step": 498
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.30102516309413e-06,
+      "loss": 0.7388,
+      "step": 499
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.319664492078286e-06,
+      "loss": 0.8181,
+      "step": 500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.338303821062443e-06,
+      "loss": 0.667,
+      "step": 501
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.356943150046599e-06,
+      "loss": 0.8384,
+      "step": 502
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.375582479030755e-06,
+      "loss": 0.736,
+      "step": 503
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.394221808014911e-06,
+      "loss": 0.6473,
+      "step": 504
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.41286113699907e-06,
+      "loss": 0.8179,
+      "step": 505
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.431500465983225e-06,
+      "loss": 0.7297,
+      "step": 506
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.450139794967382e-06,
+      "loss": 0.7839,
+      "step": 507
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.468779123951538e-06,
+      "loss": 0.7914,
+      "step": 508
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.487418452935696e-06,
+      "loss": 0.8081,
+      "step": 509
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.506057781919852e-06,
+      "loss": 0.6848,
+      "step": 510
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.524697110904008e-06,
+      "loss": 0.7321,
+      "step": 511
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.543336439888164e-06,
+      "loss": 0.6524,
+      "step": 512
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.56197576887232e-06,
+      "loss": 0.664,
+      "step": 513
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.580615097856478e-06,
+      "loss": 0.838,
+      "step": 514
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.599254426840635e-06,
+      "loss": 0.6432,
+      "step": 515
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.617893755824791e-06,
+      "loss": 0.7546,
+      "step": 516
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.636533084808947e-06,
+      "loss": 0.7382,
+      "step": 517
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.655172413793105e-06,
+      "loss": 0.7104,
+      "step": 518
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.673811742777261e-06,
+      "loss": 0.6697,
+      "step": 519
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.692451071761417e-06,
+      "loss": 0.7403,
+      "step": 520
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.711090400745574e-06,
+      "loss": 0.7437,
+      "step": 521
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.729729729729732e-06,
+      "loss": 0.7733,
+      "step": 522
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.748369058713886e-06,
+      "loss": 0.8731,
+      "step": 523
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.767008387698044e-06,
+      "loss": 0.7311,
+      "step": 524
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.7856477166822e-06,
+      "loss": 0.7097,
+      "step": 525
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.804287045666356e-06,
+      "loss": 0.6669,
+      "step": 526
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.822926374650513e-06,
+      "loss": 0.7047,
+      "step": 527
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.84156570363467e-06,
+      "loss": 0.8518,
+      "step": 528
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.860205032618827e-06,
+      "loss": 0.7214,
+      "step": 529
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.878844361602983e-06,
+      "loss": 0.7858,
+      "step": 530
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.897483690587139e-06,
+      "loss": 0.7233,
+      "step": 531
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.916123019571297e-06,
+      "loss": 0.6595,
+      "step": 532
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.934762348555452e-06,
+      "loss": 0.8281,
+      "step": 533
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.95340167753961e-06,
+      "loss": 0.7712,
+      "step": 534
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.972041006523766e-06,
+      "loss": 0.7745,
+      "step": 535
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 9.990680335507924e-06,
+      "loss": 0.7571,
+      "step": 536
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0009319664492078e-05,
+      "loss": 0.7475,
+      "step": 537
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0027958993476236e-05,
+      "loss": 0.7572,
+      "step": 538
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0046598322460392e-05,
+      "loss": 0.8059,
+      "step": 539
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0065237651444548e-05,
+      "loss": 0.6736,
+      "step": 540
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0083876980428706e-05,
+      "loss": 0.7101,
+      "step": 541
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0102516309412863e-05,
+      "loss": 0.787,
+      "step": 542
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0121155638397017e-05,
+      "loss": 0.7057,
+      "step": 543
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0139794967381175e-05,
+      "loss": 0.7164,
+      "step": 544
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0158434296365331e-05,
+      "loss": 0.7937,
+      "step": 545
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0177073625349487e-05,
+      "loss": 0.848,
+      "step": 546
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0195712954333645e-05,
+      "loss": 0.5987,
+      "step": 547
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0214352283317801e-05,
+      "loss": 0.7489,
+      "step": 548
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.023299161230196e-05,
+      "loss": 0.679,
+      "step": 549
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0251630941286116e-05,
+      "loss": 0.702,
+      "step": 550
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.027027027027027e-05,
+      "loss": 0.8054,
+      "step": 551
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0288909599254428e-05,
+      "loss": 0.7615,
+      "step": 552
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0307548928238584e-05,
+      "loss": 0.7422,
+      "step": 553
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.032618825722274e-05,
+      "loss": 0.7505,
+      "step": 554
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0344827586206898e-05,
+      "loss": 0.7823,
+      "step": 555
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0363466915191055e-05,
+      "loss": 0.7341,
+      "step": 556
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0382106244175209e-05,
+      "loss": 0.7124,
+      "step": 557
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0400745573159367e-05,
+      "loss": 0.9396,
+      "step": 558
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0419384902143523e-05,
+      "loss": 0.7408,
+      "step": 559
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.043802423112768e-05,
+      "loss": 0.7684,
+      "step": 560
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0456663560111837e-05,
+      "loss": 0.8541,
+      "step": 561
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0475302889095993e-05,
+      "loss": 0.7485,
+      "step": 562
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0493942218080151e-05,
+      "loss": 0.6754,
+      "step": 563
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0512581547064306e-05,
+      "loss": 0.662,
+      "step": 564
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0531220876048462e-05,
+      "loss": 0.7882,
+      "step": 565
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.054986020503262e-05,
+      "loss": 0.7434,
+      "step": 566
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0568499534016776e-05,
+      "loss": 0.7521,
+      "step": 567
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0587138863000932e-05,
+      "loss": 0.7501,
+      "step": 568
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.060577819198509e-05,
+      "loss": 0.7504,
+      "step": 569
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0624417520969247e-05,
+      "loss": 0.7686,
+      "step": 570
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0643056849953401e-05,
+      "loss": 0.841,
+      "step": 571
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0661696178937559e-05,
+      "loss": 0.7773,
+      "step": 572
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0680335507921715e-05,
+      "loss": 0.7531,
+      "step": 573
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0698974836905873e-05,
+      "loss": 0.7592,
+      "step": 574
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.071761416589003e-05,
+      "loss": 0.6617,
+      "step": 575
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0736253494874186e-05,
+      "loss": 0.6771,
+      "step": 576
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0754892823858343e-05,
+      "loss": 0.6839,
+      "step": 577
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0773532152842498e-05,
+      "loss": 0.6327,
+      "step": 578
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0792171481826654e-05,
+      "loss": 0.6668,
+      "step": 579
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0810810810810812e-05,
+      "loss": 0.869,
+      "step": 580
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0829450139794968e-05,
+      "loss": 0.7253,
+      "step": 581
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0848089468779124e-05,
+      "loss": 0.837,
+      "step": 582
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0866728797763282e-05,
+      "loss": 0.7097,
+      "step": 583
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0885368126747437e-05,
+      "loss": 0.6715,
+      "step": 584
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0904007455731595e-05,
+      "loss": 0.7709,
+      "step": 585
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0922646784715751e-05,
+      "loss": 0.7331,
+      "step": 586
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0941286113699907e-05,
+      "loss": 0.6801,
+      "step": 587
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0959925442684065e-05,
+      "loss": 0.6838,
+      "step": 588
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0978564771668221e-05,
+      "loss": 0.6532,
+      "step": 589
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.0997204100652376e-05,
+      "loss": 0.769,
+      "step": 590
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1015843429636535e-05,
+      "loss": 0.6603,
+      "step": 591
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.103448275862069e-05,
+      "loss": 0.7228,
+      "step": 592
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1053122087604846e-05,
+      "loss": 0.6683,
+      "step": 593
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1071761416589004e-05,
+      "loss": 0.6593,
+      "step": 594
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.109040074557316e-05,
+      "loss": 0.766,
+      "step": 595
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1109040074557318e-05,
+      "loss": 0.7455,
+      "step": 596
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1127679403541474e-05,
+      "loss": 0.5968,
+      "step": 597
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1146318732525629e-05,
+      "loss": 0.7945,
+      "step": 598
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1164958061509787e-05,
+      "loss": 0.7187,
+      "step": 599
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1183597390493943e-05,
+      "loss": 0.9322,
+      "step": 600
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.12022367194781e-05,
+      "loss": 0.8307,
+      "step": 601
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1220876048462257e-05,
+      "loss": 0.7555,
+      "step": 602
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1239515377446413e-05,
+      "loss": 0.7989,
+      "step": 603
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1258154706430568e-05,
+      "loss": 0.79,
+      "step": 604
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1276794035414726e-05,
+      "loss": 0.7389,
+      "step": 605
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1295433364398882e-05,
+      "loss": 0.6392,
+      "step": 606
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1314072693383038e-05,
+      "loss": 0.6435,
+      "step": 607
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1332712022367196e-05,
+      "loss": 0.6929,
+      "step": 608
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1351351351351352e-05,
+      "loss": 0.7092,
+      "step": 609
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.136999068033551e-05,
+      "loss": 0.6982,
+      "step": 610
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1388630009319665e-05,
+      "loss": 0.7115,
+      "step": 611
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1407269338303821e-05,
+      "loss": 0.7064,
+      "step": 612
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1425908667287979e-05,
+      "loss": 0.806,
+      "step": 613
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1444547996272135e-05,
+      "loss": 0.6837,
+      "step": 614
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1463187325256291e-05,
+      "loss": 0.7173,
+      "step": 615
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.148182665424045e-05,
+      "loss": 0.6373,
+      "step": 616
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1500465983224605e-05,
+      "loss": 0.7096,
+      "step": 617
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.151910531220876e-05,
+      "loss": 0.6947,
+      "step": 618
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1537744641192918e-05,
+      "loss": 0.8329,
+      "step": 619
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1556383970177074e-05,
+      "loss": 0.8248,
+      "step": 620
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1575023299161232e-05,
+      "loss": 0.8796,
+      "step": 621
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1593662628145388e-05,
+      "loss": 0.6713,
+      "step": 622
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1612301957129544e-05,
+      "loss": 0.7247,
+      "step": 623
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1630941286113702e-05,
+      "loss": 0.623,
+      "step": 624
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.1649580615097857e-05,
+      "loss": 0.681,
+      "step": 625
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1668219944082013e-05,
+      "loss": 0.7923,
+      "step": 626
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1686859273066171e-05,
+      "loss": 0.727,
+      "step": 627
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1705498602050327e-05,
+      "loss": 0.7581,
+      "step": 628
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1724137931034483e-05,
+      "loss": 0.6791,
+      "step": 629
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1742777260018641e-05,
+      "loss": 0.7318,
+      "step": 630
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1761416589002796e-05,
+      "loss": 0.8006,
+      "step": 631
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1780055917986954e-05,
+      "loss": 0.7257,
+      "step": 632
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.179869524697111e-05,
+      "loss": 0.6534,
+      "step": 633
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1817334575955266e-05,
+      "loss": 0.8001,
+      "step": 634
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1835973904939424e-05,
+      "loss": 0.69,
+      "step": 635
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.185461323392358e-05,
+      "loss": 0.8069,
+      "step": 636
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1873252562907736e-05,
+      "loss": 0.7087,
+      "step": 637
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1891891891891894e-05,
+      "loss": 0.8389,
+      "step": 638
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1910531220876049e-05,
+      "loss": 0.639,
+      "step": 639
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1929170549860205e-05,
+      "loss": 0.7748,
+      "step": 640
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1947809878844363e-05,
+      "loss": 0.6891,
+      "step": 641
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1966449207828519e-05,
+      "loss": 0.6774,
+      "step": 642
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.1985088536812677e-05,
+      "loss": 0.7081,
+      "step": 643
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2003727865796833e-05,
+      "loss": 0.8058,
+      "step": 644
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2022367194780988e-05,
+      "loss": 0.8398,
+      "step": 645
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2041006523765146e-05,
+      "loss": 0.7024,
+      "step": 646
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2059645852749302e-05,
+      "loss": 0.7992,
+      "step": 647
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2078285181733458e-05,
+      "loss": 0.7928,
+      "step": 648
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2096924510717616e-05,
+      "loss": 0.6557,
+      "step": 649
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2115563839701772e-05,
+      "loss": 0.8257,
+      "step": 650
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2134203168685927e-05,
+      "loss": 0.5999,
+      "step": 651
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2152842497670085e-05,
+      "loss": 0.7269,
+      "step": 652
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.217148182665424e-05,
+      "loss": 0.6874,
+      "step": 653
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2190121155638397e-05,
+      "loss": 0.8719,
+      "step": 654
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2208760484622555e-05,
+      "loss": 0.6776,
+      "step": 655
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2227399813606711e-05,
+      "loss": 0.7664,
+      "step": 656
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2246039142590869e-05,
+      "loss": 0.8045,
+      "step": 657
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2264678471575025e-05,
+      "loss": 0.6343,
+      "step": 658
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.228331780055918e-05,
+      "loss": 0.7973,
+      "step": 659
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2301957129543338e-05,
+      "loss": 0.7543,
+      "step": 660
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2320596458527494e-05,
+      "loss": 0.7007,
+      "step": 661
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.233923578751165e-05,
+      "loss": 0.7722,
+      "step": 662
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2357875116495808e-05,
+      "loss": 0.6591,
+      "step": 663
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2376514445479964e-05,
+      "loss": 0.7634,
+      "step": 664
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2395153774464119e-05,
+      "loss": 0.8275,
+      "step": 665
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2413793103448277e-05,
+      "loss": 0.6749,
+      "step": 666
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2432432432432433e-05,
+      "loss": 0.6658,
+      "step": 667
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.245107176141659e-05,
+      "loss": 0.7284,
+      "step": 668
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2469711090400747e-05,
+      "loss": 0.7763,
+      "step": 669
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2488350419384903e-05,
+      "loss": 0.8182,
+      "step": 670
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2506989748369061e-05,
+      "loss": 0.6411,
+      "step": 671
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2525629077353216e-05,
+      "loss": 0.7703,
+      "step": 672
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2544268406337372e-05,
+      "loss": 0.6872,
+      "step": 673
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.256290773532153e-05,
+      "loss": 0.9028,
+      "step": 674
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2581547064305686e-05,
+      "loss": 0.7449,
+      "step": 675
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2600186393289842e-05,
+      "loss": 0.6181,
+      "step": 676
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2618825722274e-05,
+      "loss": 0.7895,
+      "step": 677
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2637465051258155e-05,
+      "loss": 0.7387,
+      "step": 678
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2656104380242314e-05,
+      "loss": 0.7093,
+      "step": 679
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2674743709226469e-05,
+      "loss": 0.7319,
+      "step": 680
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2693383038210625e-05,
+      "loss": 0.7776,
+      "step": 681
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2712022367194783e-05,
+      "loss": 0.7884,
+      "step": 682
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2730661696178939e-05,
+      "loss": 0.7639,
+      "step": 683
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2749301025163095e-05,
+      "loss": 0.6592,
+      "step": 684
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2767940354147253e-05,
+      "loss": 0.7657,
+      "step": 685
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2786579683131408e-05,
+      "loss": 0.5674,
+      "step": 686
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2805219012115564e-05,
+      "loss": 0.8272,
+      "step": 687
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2823858341099722e-05,
+      "loss": 0.7611,
+      "step": 688
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2842497670083878e-05,
+      "loss": 0.6659,
+      "step": 689
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2861136999068036e-05,
+      "loss": 0.74,
+      "step": 690
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2879776328052192e-05,
+      "loss": 0.7323,
+      "step": 691
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2898415657036347e-05,
+      "loss": 0.7228,
+      "step": 692
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2917054986020504e-05,
+      "loss": 0.6993,
+      "step": 693
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.293569431500466e-05,
+      "loss": 0.7001,
+      "step": 694
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2954333643988817e-05,
+      "loss": 0.7474,
+      "step": 695
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2972972972972975e-05,
+      "loss": 0.7654,
+      "step": 696
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.2991612301957131e-05,
+      "loss": 0.6538,
+      "step": 697
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3010251630941286e-05,
+      "loss": 0.6666,
+      "step": 698
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3028890959925443e-05,
+      "loss": 0.7664,
+      "step": 699
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.30475302889096e-05,
+      "loss": 0.5914,
+      "step": 700
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3066169617893756e-05,
+      "loss": 0.7367,
+      "step": 701
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3084808946877914e-05,
+      "loss": 0.6723,
+      "step": 702
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.310344827586207e-05,
+      "loss": 0.8435,
+      "step": 703
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3122087604846228e-05,
+      "loss": 0.6672,
+      "step": 704
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3140726933830384e-05,
+      "loss": 0.705,
+      "step": 705
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3159366262814539e-05,
+      "loss": 0.7805,
+      "step": 706
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3178005591798697e-05,
+      "loss": 0.7055,
+      "step": 707
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3196644920782853e-05,
+      "loss": 0.6534,
+      "step": 708
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3215284249767009e-05,
+      "loss": 0.6517,
+      "step": 709
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3233923578751167e-05,
+      "loss": 0.5932,
+      "step": 710
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3252562907735323e-05,
+      "loss": 0.7089,
+      "step": 711
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3271202236719478e-05,
+      "loss": 0.797,
+      "step": 712
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3289841565703635e-05,
+      "loss": 0.697,
+      "step": 713
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3308480894687792e-05,
+      "loss": 0.7283,
+      "step": 714
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.332712022367195e-05,
+      "loss": 0.7664,
+      "step": 715
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3345759552656106e-05,
+      "loss": 0.7619,
+      "step": 716
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3364398881640262e-05,
+      "loss": 0.7924,
+      "step": 717
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.338303821062442e-05,
+      "loss": 0.7234,
+      "step": 718
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3401677539608574e-05,
+      "loss": 0.6202,
+      "step": 719
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.342031686859273e-05,
+      "loss": 0.7537,
+      "step": 720
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3438956197576889e-05,
+      "loss": 0.7203,
+      "step": 721
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3457595526561045e-05,
+      "loss": 0.8278,
+      "step": 722
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3476234855545201e-05,
+      "loss": 0.7444,
+      "step": 723
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3494874184529359e-05,
+      "loss": 0.7056,
+      "step": 724
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3513513513513515e-05,
+      "loss": 0.669,
+      "step": 725
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3532152842497673e-05,
+      "loss": 0.7712,
+      "step": 726
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3550792171481827e-05,
+      "loss": 0.6709,
+      "step": 727
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3569431500465984e-05,
+      "loss": 0.6588,
+      "step": 728
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3588070829450142e-05,
+      "loss": 0.714,
+      "step": 729
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3606710158434298e-05,
+      "loss": 0.7392,
+      "step": 730
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3625349487418454e-05,
+      "loss": 0.7463,
+      "step": 731
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3643988816402612e-05,
+      "loss": 0.7122,
+      "step": 732
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3662628145386766e-05,
+      "loss": 0.7221,
+      "step": 733
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3681267474370923e-05,
+      "loss": 0.6014,
+      "step": 734
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.369990680335508e-05,
+      "loss": 0.6829,
+      "step": 735
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3718546132339237e-05,
+      "loss": 0.7602,
+      "step": 736
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3737185461323395e-05,
+      "loss": 0.7467,
+      "step": 737
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3755824790307551e-05,
+      "loss": 0.7204,
+      "step": 738
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3774464119291705e-05,
+      "loss": 0.7034,
+      "step": 739
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3793103448275863e-05,
+      "loss": 0.816,
+      "step": 740
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.381174277726002e-05,
+      "loss": 0.6222,
+      "step": 741
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3830382106244176e-05,
+      "loss": 0.7014,
+      "step": 742
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3849021435228334e-05,
+      "loss": 0.7157,
+      "step": 743
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.386766076421249e-05,
+      "loss": 0.7394,
+      "step": 744
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3886300093196644e-05,
+      "loss": 0.73,
+      "step": 745
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3904939422180804e-05,
+      "loss": 0.7796,
+      "step": 746
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3923578751164958e-05,
+      "loss": 0.6505,
+      "step": 747
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3942218080149115e-05,
+      "loss": 0.7209,
+      "step": 748
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3960857409133273e-05,
+      "loss": 0.7849,
+      "step": 749
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3979496738117429e-05,
+      "loss": 0.7263,
+      "step": 750
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.3998136067101587e-05,
+      "loss": 0.7648,
+      "step": 751
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4016775396085743e-05,
+      "loss": 0.731,
+      "step": 752
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4035414725069897e-05,
+      "loss": 0.8489,
+      "step": 753
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4054054054054055e-05,
+      "loss": 0.86,
+      "step": 754
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4072693383038212e-05,
+      "loss": 0.7155,
+      "step": 755
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4091332712022368e-05,
+      "loss": 0.7874,
+      "step": 756
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4109972041006526e-05,
+      "loss": 0.7469,
+      "step": 757
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4128611369990682e-05,
+      "loss": 0.7708,
+      "step": 758
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4147250698974836e-05,
+      "loss": 0.7011,
+      "step": 759
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4165890027958994e-05,
+      "loss": 0.748,
+      "step": 760
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.418452935694315e-05,
+      "loss": 0.6343,
+      "step": 761
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4203168685927308e-05,
+      "loss": 0.8323,
+      "step": 762
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4221808014911465e-05,
+      "loss": 0.8329,
+      "step": 763
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.424044734389562e-05,
+      "loss": 0.8305,
+      "step": 764
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4259086672879779e-05,
+      "loss": 0.7327,
+      "step": 765
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4277726001863933e-05,
+      "loss": 0.8023,
+      "step": 766
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.429636533084809e-05,
+      "loss": 0.756,
+      "step": 767
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4315004659832247e-05,
+      "loss": 0.6926,
+      "step": 768
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4333643988816404e-05,
+      "loss": 0.7165,
+      "step": 769
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.435228331780056e-05,
+      "loss": 0.738,
+      "step": 770
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4370922646784718e-05,
+      "loss": 0.674,
+      "step": 771
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4389561975768874e-05,
+      "loss": 0.72,
+      "step": 772
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4408201304753032e-05,
+      "loss": 0.746,
+      "step": 773
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4426840633737186e-05,
+      "loss": 0.7296,
+      "step": 774
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4445479962721343e-05,
+      "loss": 0.8332,
+      "step": 775
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.44641192917055e-05,
+      "loss": 0.837,
+      "step": 776
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4482758620689657e-05,
+      "loss": 0.7843,
+      "step": 777
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4501397949673813e-05,
+      "loss": 0.7554,
+      "step": 778
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.452003727865797e-05,
+      "loss": 0.7047,
+      "step": 779
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4538676607642125e-05,
+      "loss": 0.7083,
+      "step": 780
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4557315936626281e-05,
+      "loss": 0.752,
+      "step": 781
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.457595526561044e-05,
+      "loss": 0.6681,
+      "step": 782
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4594594594594596e-05,
+      "loss": 0.6542,
+      "step": 783
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4613233923578752e-05,
+      "loss": 0.7447,
+      "step": 784
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.463187325256291e-05,
+      "loss": 0.7156,
+      "step": 785
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4650512581547064e-05,
+      "loss": 0.6585,
+      "step": 786
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4669151910531224e-05,
+      "loss": 0.7232,
+      "step": 787
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4687791239515378e-05,
+      "loss": 0.8201,
+      "step": 788
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4706430568499535e-05,
+      "loss": 0.6464,
+      "step": 789
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4725069897483692e-05,
+      "loss": 0.7274,
+      "step": 790
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4743709226467849e-05,
+      "loss": 0.6109,
+      "step": 791
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4762348555452005e-05,
+      "loss": 0.6498,
+      "step": 792
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4780987884436163e-05,
+      "loss": 0.7047,
+      "step": 793
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4799627213420317e-05,
+      "loss": 0.7256,
+      "step": 794
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4818266542404474e-05,
+      "loss": 0.7457,
+      "step": 795
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4836905871388631e-05,
+      "loss": 0.6728,
+      "step": 796
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4855545200372788e-05,
+      "loss": 0.8166,
+      "step": 797
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4874184529356946e-05,
+      "loss": 0.7468,
+      "step": 798
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4892823858341102e-05,
+      "loss": 0.7879,
+      "step": 799
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4911463187325256e-05,
+      "loss": 0.5997,
+      "step": 800
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4930102516309414e-05,
+      "loss": 0.756,
+      "step": 801
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.494874184529357e-05,
+      "loss": 0.7718,
+      "step": 802
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4967381174277727e-05,
+      "loss": 0.645,
+      "step": 803
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.4986020503261884e-05,
+      "loss": 0.7142,
+      "step": 804
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.500465983224604e-05,
+      "loss": 0.7106,
+      "step": 805
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5023299161230195e-05,
+      "loss": 0.8103,
+      "step": 806
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5041938490214353e-05,
+      "loss": 0.6867,
+      "step": 807
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.506057781919851e-05,
+      "loss": 0.8351,
+      "step": 808
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5079217148182667e-05,
+      "loss": 0.741,
+      "step": 809
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5097856477166823e-05,
+      "loss": 0.6792,
+      "step": 810
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.511649580615098e-05,
+      "loss": 0.7839,
+      "step": 811
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5135135135135138e-05,
+      "loss": 0.7822,
+      "step": 812
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5153774464119294e-05,
+      "loss": 0.7517,
+      "step": 813
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5172413793103448e-05,
+      "loss": 0.6193,
+      "step": 814
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5191053122087606e-05,
+      "loss": 0.6878,
+      "step": 815
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5209692451071762e-05,
+      "loss": 0.7325,
+      "step": 816
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5228331780055919e-05,
+      "loss": 0.7413,
+      "step": 817
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5246971109040077e-05,
+      "loss": 0.6725,
+      "step": 818
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5265610438024233e-05,
+      "loss": 0.6104,
+      "step": 819
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.528424976700839e-05,
+      "loss": 0.7067,
+      "step": 820
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5302889095992545e-05,
+      "loss": 0.7911,
+      "step": 821
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5321528424976703e-05,
+      "loss": 0.7306,
+      "step": 822
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.534016775396086e-05,
+      "loss": 0.601,
+      "step": 823
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5358807082945015e-05,
+      "loss": 0.8067,
+      "step": 824
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.537744641192917e-05,
+      "loss": 0.7167,
+      "step": 825
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5396085740913328e-05,
+      "loss": 0.686,
+      "step": 826
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5414725069897486e-05,
+      "loss": 0.8925,
+      "step": 827
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.543336439888164e-05,
+      "loss": 0.6575,
+      "step": 828
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5452003727865798e-05,
+      "loss": 0.7418,
+      "step": 829
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5470643056849953e-05,
+      "loss": 0.7715,
+      "step": 830
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.548928238583411e-05,
+      "loss": 0.8341,
+      "step": 831
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.550792171481827e-05,
+      "loss": 0.7426,
+      "step": 832
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5526561043802423e-05,
+      "loss": 0.7971,
+      "step": 833
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.554520037278658e-05,
+      "loss": 0.6141,
+      "step": 834
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.556383970177074e-05,
+      "loss": 0.5737,
+      "step": 835
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5582479030754893e-05,
+      "loss": 0.7158,
+      "step": 836
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.560111835973905e-05,
+      "loss": 0.6045,
+      "step": 837
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5619757688723206e-05,
+      "loss": 0.6909,
+      "step": 838
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5638397017707364e-05,
+      "loss": 0.7653,
+      "step": 839
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.565703634669152e-05,
+      "loss": 0.7921,
+      "step": 840
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5675675675675676e-05,
+      "loss": 0.7224,
+      "step": 841
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5694315004659834e-05,
+      "loss": 0.6489,
+      "step": 842
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5712954333643992e-05,
+      "loss": 0.8299,
+      "step": 843
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5731593662628146e-05,
+      "loss": 0.8376,
+      "step": 844
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5750232991612304e-05,
+      "loss": 0.7145,
+      "step": 845
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.576887232059646e-05,
+      "loss": 0.8202,
+      "step": 846
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5787511649580617e-05,
+      "loss": 0.7874,
+      "step": 847
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5806150978564775e-05,
+      "loss": 0.5912,
+      "step": 848
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.582479030754893e-05,
+      "loss": 0.6692,
+      "step": 849
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5843429636533084e-05,
+      "loss": 0.6017,
+      "step": 850
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.586206896551724e-05,
+      "loss": 0.6957,
+      "step": 851
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.58807082945014e-05,
+      "loss": 0.7147,
+      "step": 852
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5899347623485554e-05,
+      "loss": 0.772,
+      "step": 853
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5917986952469712e-05,
+      "loss": 0.8673,
+      "step": 854
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.593662628145387e-05,
+      "loss": 0.7159,
+      "step": 855
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5955265610438028e-05,
+      "loss": 0.7921,
+      "step": 856
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5973904939422182e-05,
+      "loss": 0.7888,
+      "step": 857
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.5992544268406337e-05,
+      "loss": 0.7091,
+      "step": 858
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6011183597390495e-05,
+      "loss": 0.7172,
+      "step": 859
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6029822926374653e-05,
+      "loss": 0.593,
+      "step": 860
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6048462255358807e-05,
+      "loss": 0.7316,
+      "step": 861
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6067101584342965e-05,
+      "loss": 0.6533,
+      "step": 862
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6085740913327123e-05,
+      "loss": 0.8092,
+      "step": 863
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6104380242311277e-05,
+      "loss": 0.827,
+      "step": 864
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6123019571295435e-05,
+      "loss": 0.8015,
+      "step": 865
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.614165890027959e-05,
+      "loss": 0.6808,
+      "step": 866
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6160298229263748e-05,
+      "loss": 0.7988,
+      "step": 867
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6178937558247906e-05,
+      "loss": 0.6016,
+      "step": 868
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.619757688723206e-05,
+      "loss": 0.6697,
+      "step": 869
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6216216216216218e-05,
+      "loss": 0.7582,
+      "step": 870
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6234855545200373e-05,
+      "loss": 0.7323,
+      "step": 871
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.625349487418453e-05,
+      "loss": 0.7093,
+      "step": 872
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.627213420316869e-05,
+      "loss": 0.6269,
+      "step": 873
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6290773532152843e-05,
+      "loss": 0.6928,
+      "step": 874
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6309412861137e-05,
+      "loss": 0.7194,
+      "step": 875
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.632805219012116e-05,
+      "loss": 0.7174,
+      "step": 876
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6346691519105313e-05,
+      "loss": 0.6879,
+      "step": 877
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6365330848089468e-05,
+      "loss": 0.7432,
+      "step": 878
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6383970177073626e-05,
+      "loss": 0.772,
+      "step": 879
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6402609506057784e-05,
+      "loss": 0.8377,
+      "step": 880
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.642124883504194e-05,
+      "loss": 0.6993,
+      "step": 881
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6439888164026096e-05,
+      "loss": 0.5896,
+      "step": 882
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6458527493010254e-05,
+      "loss": 0.7686,
+      "step": 883
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6477166821994412e-05,
+      "loss": 0.7155,
+      "step": 884
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6495806150978566e-05,
+      "loss": 0.7032,
+      "step": 885
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.651444547996272e-05,
+      "loss": 0.6481,
+      "step": 886
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.653308480894688e-05,
+      "loss": 0.6229,
+      "step": 887
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6551724137931037e-05,
+      "loss": 0.6541,
+      "step": 888
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.657036346691519e-05,
+      "loss": 0.6091,
+      "step": 889
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.658900279589935e-05,
+      "loss": 0.7937,
+      "step": 890
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6607642124883504e-05,
+      "loss": 0.787,
+      "step": 891
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.662628145386766e-05,
+      "loss": 0.6433,
+      "step": 892
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.664492078285182e-05,
+      "loss": 0.6638,
+      "step": 893
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6663560111835974e-05,
+      "loss": 0.7367,
+      "step": 894
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6682199440820132e-05,
+      "loss": 0.7155,
+      "step": 895
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.670083876980429e-05,
+      "loss": 0.8036,
+      "step": 896
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6719478098788444e-05,
+      "loss": 0.8555,
+      "step": 897
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6738117427772602e-05,
+      "loss": 0.8291,
+      "step": 898
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6756756756756757e-05,
+      "loss": 0.8272,
+      "step": 899
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6775396085740915e-05,
+      "loss": 0.6139,
+      "step": 900
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6794035414725072e-05,
+      "loss": 0.5862,
+      "step": 901
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6812674743709227e-05,
+      "loss": 0.7105,
+      "step": 902
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6831314072693385e-05,
+      "loss": 0.7546,
+      "step": 903
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6849953401677543e-05,
+      "loss": 0.617,
+      "step": 904
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6868592730661697e-05,
+      "loss": 0.7638,
+      "step": 905
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6887232059645855e-05,
+      "loss": 0.724,
+      "step": 906
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.690587138863001e-05,
+      "loss": 0.7312,
+      "step": 907
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6924510717614168e-05,
+      "loss": 0.7946,
+      "step": 908
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6943150046598326e-05,
+      "loss": 0.6794,
+      "step": 909
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.696178937558248e-05,
+      "loss": 0.6955,
+      "step": 910
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6980428704566635e-05,
+      "loss": 0.853,
+      "step": 911
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.6999068033550792e-05,
+      "loss": 0.86,
+      "step": 912
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.701770736253495e-05,
+      "loss": 0.7696,
+      "step": 913
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7036346691519108e-05,
+      "loss": 0.7889,
+      "step": 914
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7054986020503263e-05,
+      "loss": 0.6921,
+      "step": 915
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.707362534948742e-05,
+      "loss": 0.6422,
+      "step": 916
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.709226467847158e-05,
+      "loss": 0.7719,
+      "step": 917
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7110904007455733e-05,
+      "loss": 0.745,
+      "step": 918
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7129543336439888e-05,
+      "loss": 0.7576,
+      "step": 919
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7148182665424046e-05,
+      "loss": 0.6944,
+      "step": 920
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7166821994408203e-05,
+      "loss": 0.7219,
+      "step": 921
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7185461323392358e-05,
+      "loss": 0.7154,
+      "step": 922
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7204100652376516e-05,
+      "loss": 0.7516,
+      "step": 923
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7222739981360674e-05,
+      "loss": 0.7459,
+      "step": 924
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7241379310344828e-05,
+      "loss": 0.8113,
+      "step": 925
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7260018639328986e-05,
+      "loss": 0.6794,
+      "step": 926
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.727865796831314e-05,
+      "loss": 0.8527,
+      "step": 927
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.72972972972973e-05,
+      "loss": 0.7785,
+      "step": 928
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7315936626281457e-05,
+      "loss": 0.6928,
+      "step": 929
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.733457595526561e-05,
+      "loss": 0.8661,
+      "step": 930
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.735321528424977e-05,
+      "loss": 0.7122,
+      "step": 931
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7371854613233923e-05,
+      "loss": 0.8509,
+      "step": 932
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.739049394221808e-05,
+      "loss": 0.6495,
+      "step": 933
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.740913327120224e-05,
+      "loss": 0.7552,
+      "step": 934
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7427772600186394e-05,
+      "loss": 0.5724,
+      "step": 935
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.744641192917055e-05,
+      "loss": 0.6394,
+      "step": 936
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.746505125815471e-05,
+      "loss": 0.66,
+      "step": 937
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7483690587138864e-05,
+      "loss": 0.6928,
+      "step": 938
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7502329916123022e-05,
+      "loss": 0.7123,
+      "step": 939
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7520969245107177e-05,
+      "loss": 0.6727,
+      "step": 940
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7539608574091334e-05,
+      "loss": 0.7848,
+      "step": 941
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7558247903075492e-05,
+      "loss": 0.6826,
+      "step": 942
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7576887232059647e-05,
+      "loss": 0.8711,
+      "step": 943
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7595526561043805e-05,
+      "loss": 0.7341,
+      "step": 944
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7614165890027963e-05,
+      "loss": 0.854,
+      "step": 945
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7632805219012117e-05,
+      "loss": 0.725,
+      "step": 946
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7651444547996272e-05,
+      "loss": 0.8663,
+      "step": 947
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.767008387698043e-05,
+      "loss": 0.6842,
+      "step": 948
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7688723205964588e-05,
+      "loss": 0.8315,
+      "step": 949
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7707362534948745e-05,
+      "loss": 0.7656,
+      "step": 950
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.77260018639329e-05,
+      "loss": 0.7006,
+      "step": 951
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7744641192917054e-05,
+      "loss": 0.6532,
+      "step": 952
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7763280521901212e-05,
+      "loss": 0.6713,
+      "step": 953
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.778191985088537e-05,
+      "loss": 0.8081,
+      "step": 954
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7800559179869525e-05,
+      "loss": 0.6864,
+      "step": 955
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7819198508853683e-05,
+      "loss": 0.7376,
+      "step": 956
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.783783783783784e-05,
+      "loss": 0.6706,
+      "step": 957
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7856477166821995e-05,
+      "loss": 0.7468,
+      "step": 958
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7875116495806153e-05,
+      "loss": 0.6694,
+      "step": 959
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7893755824790308e-05,
+      "loss": 0.6739,
+      "step": 960
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7912395153774465e-05,
+      "loss": 0.8162,
+      "step": 961
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7931034482758623e-05,
+      "loss": 0.7361,
+      "step": 962
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7949673811742778e-05,
+      "loss": 0.7025,
+      "step": 963
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7968313140726936e-05,
+      "loss": 0.6973,
+      "step": 964
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.7986952469711094e-05,
+      "loss": 0.6293,
+      "step": 965
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8005591798695248e-05,
+      "loss": 0.7365,
+      "step": 966
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8024231127679406e-05,
+      "loss": 0.7588,
+      "step": 967
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.804287045666356e-05,
+      "loss": 0.719,
+      "step": 968
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.806150978564772e-05,
+      "loss": 0.7327,
+      "step": 969
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8080149114631876e-05,
+      "loss": 0.8299,
+      "step": 970
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.809878844361603e-05,
+      "loss": 0.6976,
+      "step": 971
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8117427772600185e-05,
+      "loss": 0.7007,
+      "step": 972
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8136067101584343e-05,
+      "loss": 0.7028,
+      "step": 973
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.81547064305685e-05,
+      "loss": 0.6684,
+      "step": 974
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.817334575955266e-05,
+      "loss": 0.7636,
+      "step": 975
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8191985088536814e-05,
+      "loss": 0.7156,
+      "step": 976
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.821062441752097e-05,
+      "loss": 0.6398,
+      "step": 977
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.822926374650513e-05,
+      "loss": 0.6526,
+      "step": 978
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8247903075489284e-05,
+      "loss": 0.5897,
+      "step": 979
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.826654240447344e-05,
+      "loss": 0.7994,
+      "step": 980
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8285181733457596e-05,
+      "loss": 0.7733,
+      "step": 981
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.8303821062441754e-05,
+      "loss": 0.7627,
+      "step": 982
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.832246039142591e-05,
+      "loss": 0.7185,
+      "step": 983
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8341099720410067e-05,
+      "loss": 0.7702,
+      "step": 984
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.835973904939422e-05,
+      "loss": 0.7722,
+      "step": 985
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8378378378378383e-05,
+      "loss": 0.7399,
+      "step": 986
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8397017707362537e-05,
+      "loss": 0.7149,
+      "step": 987
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.841565703634669e-05,
+      "loss": 0.6549,
+      "step": 988
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.843429636533085e-05,
+      "loss": 0.8503,
+      "step": 989
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8452935694315007e-05,
+      "loss": 0.6803,
+      "step": 990
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8471575023299162e-05,
+      "loss": 0.7417,
+      "step": 991
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.849021435228332e-05,
+      "loss": 0.7607,
+      "step": 992
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8508853681267474e-05,
+      "loss": 0.7669,
+      "step": 993
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8527493010251632e-05,
+      "loss": 0.7417,
+      "step": 994
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.854613233923579e-05,
+      "loss": 0.6973,
+      "step": 995
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8564771668219945e-05,
+      "loss": 0.7333,
+      "step": 996
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8583410997204103e-05,
+      "loss": 0.7243,
+      "step": 997
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.860205032618826e-05,
+      "loss": 0.6553,
+      "step": 998
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8620689655172415e-05,
+      "loss": 0.8592,
+      "step": 999
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8639328984156573e-05,
+      "loss": 0.7206,
+      "step": 1000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8657968313140727e-05,
+      "loss": 0.715,
+      "step": 1001
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8676607642124885e-05,
+      "loss": 0.6275,
+      "step": 1002
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8695246971109043e-05,
+      "loss": 0.669,
+      "step": 1003
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8713886300093198e-05,
+      "loss": 0.707,
+      "step": 1004
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8732525629077352e-05,
+      "loss": 0.7078,
+      "step": 1005
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.875116495806151e-05,
+      "loss": 0.7222,
+      "step": 1006
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8769804287045668e-05,
+      "loss": 0.7108,
+      "step": 1007
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8788443616029823e-05,
+      "loss": 0.8807,
+      "step": 1008
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.880708294501398e-05,
+      "loss": 0.6577,
+      "step": 1009
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.882572227399814e-05,
+      "loss": 0.6956,
+      "step": 1010
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8844361602982296e-05,
+      "loss": 0.5947,
+      "step": 1011
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.886300093196645e-05,
+      "loss": 0.6861,
+      "step": 1012
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8881640260950605e-05,
+      "loss": 0.6636,
+      "step": 1013
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8900279589934763e-05,
+      "loss": 0.6694,
+      "step": 1014
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.891891891891892e-05,
+      "loss": 0.6757,
+      "step": 1015
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8937558247903076e-05,
+      "loss": 0.6478,
+      "step": 1016
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8956197576887234e-05,
+      "loss": 0.7578,
+      "step": 1017
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.897483690587139e-05,
+      "loss": 0.7606,
+      "step": 1018
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.8993476234855546e-05,
+      "loss": 0.6918,
+      "step": 1019
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9012115563839704e-05,
+      "loss": 0.6314,
+      "step": 1020
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.903075489282386e-05,
+      "loss": 0.6567,
+      "step": 1021
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9049394221808016e-05,
+      "loss": 0.6414,
+      "step": 1022
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9068033550792174e-05,
+      "loss": 0.6139,
+      "step": 1023
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.908667287977633e-05,
+      "loss": 0.722,
+      "step": 1024
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9105312208760487e-05,
+      "loss": 0.7393,
+      "step": 1025
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.912395153774464e-05,
+      "loss": 0.71,
+      "step": 1026
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.91425908667288e-05,
+      "loss": 0.703,
+      "step": 1027
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9161230195712957e-05,
+      "loss": 0.6065,
+      "step": 1028
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.917986952469711e-05,
+      "loss": 0.8045,
+      "step": 1029
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.919850885368127e-05,
+      "loss": 0.6948,
+      "step": 1030
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9217148182665427e-05,
+      "loss": 0.8387,
+      "step": 1031
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9235787511649582e-05,
+      "loss": 0.7025,
+      "step": 1032
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.925442684063374e-05,
+      "loss": 0.7845,
+      "step": 1033
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9273066169617894e-05,
+      "loss": 0.7539,
+      "step": 1034
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9291705498602052e-05,
+      "loss": 0.7295,
+      "step": 1035
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.931034482758621e-05,
+      "loss": 0.7209,
+      "step": 1036
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9328984156570365e-05,
+      "loss": 0.7613,
+      "step": 1037
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9347623485554522e-05,
+      "loss": 0.7143,
+      "step": 1038
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.936626281453868e-05,
+      "loss": 0.728,
+      "step": 1039
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9384902143522835e-05,
+      "loss": 0.6937,
+      "step": 1040
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.940354147250699e-05,
+      "loss": 0.6416,
+      "step": 1041
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9422180801491147e-05,
+      "loss": 0.7642,
+      "step": 1042
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9440820130475305e-05,
+      "loss": 0.7385,
+      "step": 1043
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9459459459459463e-05,
+      "loss": 0.6495,
+      "step": 1044
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9478098788443618e-05,
+      "loss": 0.7398,
+      "step": 1045
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9496738117427772e-05,
+      "loss": 0.7188,
+      "step": 1046
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.951537744641193e-05,
+      "loss": 0.6628,
+      "step": 1047
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9534016775396088e-05,
+      "loss": 0.6619,
+      "step": 1048
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9552656104380242e-05,
+      "loss": 0.7722,
+      "step": 1049
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.95712954333644e-05,
+      "loss": 0.7106,
+      "step": 1050
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9589934762348558e-05,
+      "loss": 0.7899,
+      "step": 1051
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9608574091332713e-05,
+      "loss": 0.7797,
+      "step": 1052
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.962721342031687e-05,
+      "loss": 0.6944,
+      "step": 1053
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9645852749301025e-05,
+      "loss": 0.7148,
+      "step": 1054
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9664492078285183e-05,
+      "loss": 0.631,
+      "step": 1055
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.968313140726934e-05,
+      "loss": 0.6902,
+      "step": 1056
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9701770736253495e-05,
+      "loss": 0.6831,
+      "step": 1057
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9720410065237653e-05,
+      "loss": 0.713,
+      "step": 1058
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.973904939422181e-05,
+      "loss": 0.7099,
+      "step": 1059
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9757688723205966e-05,
+      "loss": 0.7708,
+      "step": 1060
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9776328052190124e-05,
+      "loss": 0.6308,
+      "step": 1061
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9794967381174278e-05,
+      "loss": 0.7833,
+      "step": 1062
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9813606710158436e-05,
+      "loss": 0.8013,
+      "step": 1063
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9832246039142594e-05,
+      "loss": 0.7891,
+      "step": 1064
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.985088536812675e-05,
+      "loss": 0.6189,
+      "step": 1065
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9869524697110903e-05,
+      "loss": 0.7409,
+      "step": 1066
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.988816402609506e-05,
+      "loss": 0.6329,
+      "step": 1067
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.990680335507922e-05,
+      "loss": 0.6424,
+      "step": 1068
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9925442684063377e-05,
+      "loss": 0.7307,
+      "step": 1069
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994408201304753e-05,
+      "loss": 0.6674,
+      "step": 1070
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996272134203169e-05,
+      "loss": 0.6701,
+      "step": 1071
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981360671015847e-05,
+      "loss": 0.6573,
+      "step": 1072
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2e-05,
+      "loss": 0.6268,
+      "step": 1073
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999942324883929e-05,
+      "loss": 0.7076,
+      "step": 1074
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999884649767858e-05,
+      "loss": 0.7152,
+      "step": 1075
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9998269746517868e-05,
+      "loss": 0.6845,
+      "step": 1076
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9997692995357156e-05,
+      "loss": 0.527,
+      "step": 1077
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9997116244196443e-05,
+      "loss": 0.7878,
+      "step": 1078
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999653949303573e-05,
+      "loss": 0.7048,
+      "step": 1079
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999596274187502e-05,
+      "loss": 0.7454,
+      "step": 1080
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999538599071431e-05,
+      "loss": 0.7639,
+      "step": 1081
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9994809239553598e-05,
+      "loss": 0.5792,
+      "step": 1082
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9994232488392885e-05,
+      "loss": 0.617,
+      "step": 1083
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9993655737232173e-05,
+      "loss": 0.7309,
+      "step": 1084
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999307898607146e-05,
+      "loss": 0.761,
+      "step": 1085
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999250223491075e-05,
+      "loss": 0.7856,
+      "step": 1086
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.999192548375004e-05,
+      "loss": 0.8349,
+      "step": 1087
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9991348732589327e-05,
+      "loss": 0.7264,
+      "step": 1088
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9990771981428615e-05,
+      "loss": 0.7475,
+      "step": 1089
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9990195230267902e-05,
+      "loss": 0.677,
+      "step": 1090
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998961847910719e-05,
+      "loss": 0.7578,
+      "step": 1091
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9989041727946478e-05,
+      "loss": 0.791,
+      "step": 1092
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9988464976785766e-05,
+      "loss": 0.8617,
+      "step": 1093
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9987888225625057e-05,
+      "loss": 0.9244,
+      "step": 1094
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9987311474464344e-05,
+      "loss": 0.6054,
+      "step": 1095
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9986734723303632e-05,
+      "loss": 0.6922,
+      "step": 1096
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998615797214292e-05,
+      "loss": 0.766,
+      "step": 1097
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9985581220982207e-05,
+      "loss": 0.6802,
+      "step": 1098
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9985004469821495e-05,
+      "loss": 0.7171,
+      "step": 1099
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9984427718660786e-05,
+      "loss": 0.6856,
+      "step": 1100
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9983850967500074e-05,
+      "loss": 0.7277,
+      "step": 1101
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998327421633936e-05,
+      "loss": 0.8068,
+      "step": 1102
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9982697465178653e-05,
+      "loss": 0.7012,
+      "step": 1103
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.998212071401794e-05,
+      "loss": 0.8022,
+      "step": 1104
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9981543962857225e-05,
+      "loss": 0.7924,
+      "step": 1105
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980967211696516e-05,
+      "loss": 0.6234,
+      "step": 1106
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9980390460535803e-05,
+      "loss": 0.8735,
+      "step": 1107
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997981370937509e-05,
+      "loss": 0.6662,
+      "step": 1108
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9979236958214382e-05,
+      "loss": 0.7098,
+      "step": 1109
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997866020705367e-05,
+      "loss": 0.734,
+      "step": 1110
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9978083455892957e-05,
+      "loss": 0.7076,
+      "step": 1111
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9977506704732245e-05,
+      "loss": 0.7144,
+      "step": 1112
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9976929953571533e-05,
+      "loss": 0.7807,
+      "step": 1113
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997635320241082e-05,
+      "loss": 0.7371,
+      "step": 1114
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997577645125011e-05,
+      "loss": 0.8385,
+      "step": 1115
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99751997000894e-05,
+      "loss": 0.6525,
+      "step": 1116
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9974622948928687e-05,
+      "loss": 0.705,
+      "step": 1117
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9974046197767975e-05,
+      "loss": 0.8017,
+      "step": 1118
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9973469446607262e-05,
+      "loss": 0.7328,
+      "step": 1119
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997289269544655e-05,
+      "loss": 0.7222,
+      "step": 1120
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997231594428584e-05,
+      "loss": 0.7645,
+      "step": 1121
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.997173919312513e-05,
+      "loss": 0.7771,
+      "step": 1122
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9971162441964416e-05,
+      "loss": 0.7526,
+      "step": 1123
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9970585690803704e-05,
+      "loss": 0.7454,
+      "step": 1124
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9970008939642992e-05,
+      "loss": 0.5991,
+      "step": 1125
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996943218848228e-05,
+      "loss": 0.8479,
+      "step": 1126
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9968855437321567e-05,
+      "loss": 0.7393,
+      "step": 1127
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9968278686160858e-05,
+      "loss": 0.7047,
+      "step": 1128
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967701935000146e-05,
+      "loss": 0.6772,
+      "step": 1129
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9967125183839434e-05,
+      "loss": 0.7701,
+      "step": 1130
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996654843267872e-05,
+      "loss": 0.6803,
+      "step": 1131
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996597168151801e-05,
+      "loss": 0.7651,
+      "step": 1132
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965394930357297e-05,
+      "loss": 0.6423,
+      "step": 1133
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964818179196588e-05,
+      "loss": 0.6046,
+      "step": 1134
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964241428035875e-05,
+      "loss": 0.7658,
+      "step": 1135
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963664676875163e-05,
+      "loss": 0.7601,
+      "step": 1136
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963087925714454e-05,
+      "loss": 0.7696,
+      "step": 1137
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962511174553742e-05,
+      "loss": 0.6731,
+      "step": 1138
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996193442339303e-05,
+      "loss": 0.636,
+      "step": 1139
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961357672232317e-05,
+      "loss": 0.7468,
+      "step": 1140
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960780921071605e-05,
+      "loss": 0.5983,
+      "step": 1141
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960204169910893e-05,
+      "loss": 0.7326,
+      "step": 1142
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959627418750184e-05,
+      "loss": 0.7807,
+      "step": 1143
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995905066758947e-05,
+      "loss": 0.6315,
+      "step": 1144
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995847391642876e-05,
+      "loss": 0.7727,
+      "step": 1145
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957897165268047e-05,
+      "loss": 0.7244,
+      "step": 1146
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957320414107334e-05,
+      "loss": 0.691,
+      "step": 1147
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956743662946622e-05,
+      "loss": 0.8063,
+      "step": 1148
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956166911785913e-05,
+      "loss": 0.7842,
+      "step": 1149
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99555901606252e-05,
+      "loss": 0.7475,
+      "step": 1150
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995501340946449e-05,
+      "loss": 0.7221,
+      "step": 1151
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954436658303776e-05,
+      "loss": 0.8199,
+      "step": 1152
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953859907143064e-05,
+      "loss": 0.6944,
+      "step": 1153
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995328315598235e-05,
+      "loss": 0.6115,
+      "step": 1154
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952706404821643e-05,
+      "loss": 0.7869,
+      "step": 1155
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995212965366093e-05,
+      "loss": 0.7479,
+      "step": 1156
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951552902500218e-05,
+      "loss": 0.7123,
+      "step": 1157
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950976151339506e-05,
+      "loss": 0.706,
+      "step": 1158
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950399400178793e-05,
+      "loss": 0.6669,
+      "step": 1159
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994982264901808e-05,
+      "loss": 0.6485,
+      "step": 1160
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994924589785737e-05,
+      "loss": 0.7244,
+      "step": 1161
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.994866914669666e-05,
+      "loss": 0.6523,
+      "step": 1162
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9948092395535948e-05,
+      "loss": 0.7044,
+      "step": 1163
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9947515644375235e-05,
+      "loss": 0.6973,
+      "step": 1164
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9946938893214526e-05,
+      "loss": 0.7192,
+      "step": 1165
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9946362142053814e-05,
+      "loss": 0.5842,
+      "step": 1166
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9945785390893098e-05,
+      "loss": 0.8572,
+      "step": 1167
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.994520863973239e-05,
+      "loss": 0.8414,
+      "step": 1168
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9944631888571677e-05,
+      "loss": 0.6434,
+      "step": 1169
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9944055137410965e-05,
+      "loss": 0.7106,
+      "step": 1170
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9943478386250256e-05,
+      "loss": 0.7806,
+      "step": 1171
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9942901635089543e-05,
+      "loss": 0.7436,
+      "step": 1172
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.994232488392883e-05,
+      "loss": 0.6091,
+      "step": 1173
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.994174813276812e-05,
+      "loss": 0.6286,
+      "step": 1174
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9941171381607407e-05,
+      "loss": 0.5873,
+      "step": 1175
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9940594630446694e-05,
+      "loss": 0.7179,
+      "step": 1176
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9940017879285985e-05,
+      "loss": 0.6614,
+      "step": 1177
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9939441128125273e-05,
+      "loss": 0.7187,
+      "step": 1178
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993886437696456e-05,
+      "loss": 0.729,
+      "step": 1179
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993828762580385e-05,
+      "loss": 0.6831,
+      "step": 1180
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9937710874643136e-05,
+      "loss": 0.5788,
+      "step": 1181
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9937134123482424e-05,
+      "loss": 0.6698,
+      "step": 1182
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9936557372321715e-05,
+      "loss": 0.695,
+      "step": 1183
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935980621161002e-05,
+      "loss": 0.786,
+      "step": 1184
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993540387000029e-05,
+      "loss": 0.7313,
+      "step": 1185
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934827118839578e-05,
+      "loss": 0.8178,
+      "step": 1186
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934250367678866e-05,
+      "loss": 0.7859,
+      "step": 1187
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933673616518153e-05,
+      "loss": 0.643,
+      "step": 1188
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933096865357444e-05,
+      "loss": 0.6689,
+      "step": 1189
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932520114196732e-05,
+      "loss": 0.736,
+      "step": 1190
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993194336303602e-05,
+      "loss": 0.7847,
+      "step": 1191
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993136661187531e-05,
+      "loss": 0.7822,
+      "step": 1192
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930789860714595e-05,
+      "loss": 0.6878,
+      "step": 1193
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930213109553883e-05,
+      "loss": 0.6773,
+      "step": 1194
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992963635839317e-05,
+      "loss": 0.7261,
+      "step": 1195
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992905960723246e-05,
+      "loss": 0.7404,
+      "step": 1196
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992848285607175e-05,
+      "loss": 0.5993,
+      "step": 1197
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927906104911037e-05,
+      "loss": 0.6835,
+      "step": 1198
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927329353750328e-05,
+      "loss": 0.7758,
+      "step": 1199
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926752602589616e-05,
+      "loss": 0.6877,
+      "step": 1200
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926175851428903e-05,
+      "loss": 0.6461,
+      "step": 1201
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992559910026819e-05,
+      "loss": 0.7324,
+      "step": 1202
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992502234910748e-05,
+      "loss": 0.7294,
+      "step": 1203
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924445597946766e-05,
+      "loss": 0.7424,
+      "step": 1204
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923868846786057e-05,
+      "loss": 0.8032,
+      "step": 1205
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923292095625345e-05,
+      "loss": 0.7068,
+      "step": 1206
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922715344464633e-05,
+      "loss": 0.7615,
+      "step": 1207
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992213859330392e-05,
+      "loss": 0.6991,
+      "step": 1208
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921561842143208e-05,
+      "loss": 0.6813,
+      "step": 1209
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920985090982496e-05,
+      "loss": 0.7197,
+      "step": 1210
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920408339821787e-05,
+      "loss": 0.7796,
+      "step": 1211
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919831588661075e-05,
+      "loss": 0.7299,
+      "step": 1212
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919254837500362e-05,
+      "loss": 0.7625,
+      "step": 1213
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991867808633965e-05,
+      "loss": 0.7454,
+      "step": 1214
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918101335178938e-05,
+      "loss": 0.6915,
+      "step": 1215
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917524584018225e-05,
+      "loss": 0.6391,
+      "step": 1216
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916947832857516e-05,
+      "loss": 0.7188,
+      "step": 1217
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916371081696804e-05,
+      "loss": 0.6796,
+      "step": 1218
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915794330536092e-05,
+      "loss": 0.5763,
+      "step": 1219
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991521757937538e-05,
+      "loss": 0.7271,
+      "step": 1220
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914640828214667e-05,
+      "loss": 0.7179,
+      "step": 1221
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9914064077053955e-05,
+      "loss": 0.7335,
+      "step": 1222
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913487325893246e-05,
+      "loss": 0.7094,
+      "step": 1223
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912910574732534e-05,
+      "loss": 0.6891,
+      "step": 1224
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991233382357182e-05,
+      "loss": 0.7996,
+      "step": 1225
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911757072411112e-05,
+      "loss": 0.6439,
+      "step": 1226
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99111803212504e-05,
+      "loss": 0.6978,
+      "step": 1227
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910603570089688e-05,
+      "loss": 0.6256,
+      "step": 1228
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910026818928975e-05,
+      "loss": 0.7496,
+      "step": 1229
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909450067768263e-05,
+      "loss": 0.6285,
+      "step": 1230
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990887331660755e-05,
+      "loss": 0.8006,
+      "step": 1231
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990829656544684e-05,
+      "loss": 0.7074,
+      "step": 1232
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990771981428613e-05,
+      "loss": 0.7085,
+      "step": 1233
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907143063125417e-05,
+      "loss": 0.7778,
+      "step": 1234
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906566311964705e-05,
+      "loss": 0.7299,
+      "step": 1235
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905989560803993e-05,
+      "loss": 0.7398,
+      "step": 1236
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990541280964328e-05,
+      "loss": 0.7887,
+      "step": 1237
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904836058482568e-05,
+      "loss": 0.6353,
+      "step": 1238
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990425930732186e-05,
+      "loss": 0.7737,
+      "step": 1239
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903682556161147e-05,
+      "loss": 0.6528,
+      "step": 1240
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903105805000434e-05,
+      "loss": 0.8114,
+      "step": 1241
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902529053839722e-05,
+      "loss": 0.6774,
+      "step": 1242
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990195230267901e-05,
+      "loss": 0.7704,
+      "step": 1243
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901375551518297e-05,
+      "loss": 0.7355,
+      "step": 1244
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990079880035759e-05,
+      "loss": 0.7345,
+      "step": 1245
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900222049196876e-05,
+      "loss": 0.6595,
+      "step": 1246
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899645298036164e-05,
+      "loss": 0.5212,
+      "step": 1247
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989906854687545e-05,
+      "loss": 0.7708,
+      "step": 1248
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989849179571474e-05,
+      "loss": 0.7135,
+      "step": 1249
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897915044554027e-05,
+      "loss": 0.5348,
+      "step": 1250
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897338293393318e-05,
+      "loss": 0.6632,
+      "step": 1251
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896761542232606e-05,
+      "loss": 0.7317,
+      "step": 1252
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896184791071893e-05,
+      "loss": 0.8141,
+      "step": 1253
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895608039911184e-05,
+      "loss": 0.7681,
+      "step": 1254
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989503128875047e-05,
+      "loss": 0.7762,
+      "step": 1255
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894454537589756e-05,
+      "loss": 0.6265,
+      "step": 1256
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9893877786429048e-05,
+      "loss": 0.7941,
+      "step": 1257
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9893301035268335e-05,
+      "loss": 0.7743,
+      "step": 1258
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9892724284107623e-05,
+      "loss": 0.7144,
+      "step": 1259
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9892147532946914e-05,
+      "loss": 0.6873,
+      "step": 1260
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98915707817862e-05,
+      "loss": 0.6016,
+      "step": 1261
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989099403062549e-05,
+      "loss": 0.8806,
+      "step": 1262
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9890417279464777e-05,
+      "loss": 0.7413,
+      "step": 1263
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9889840528304065e-05,
+      "loss": 0.6999,
+      "step": 1264
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9889263777143352e-05,
+      "loss": 0.6559,
+      "step": 1265
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988868702598264e-05,
+      "loss": 0.566,
+      "step": 1266
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988811027482193e-05,
+      "loss": 0.6653,
+      "step": 1267
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988753352366122e-05,
+      "loss": 0.7203,
+      "step": 1268
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9886956772500507e-05,
+      "loss": 0.7779,
+      "step": 1269
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9886380021339794e-05,
+      "loss": 0.7257,
+      "step": 1270
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9885803270179082e-05,
+      "loss": 0.733,
+      "step": 1271
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988522651901837e-05,
+      "loss": 0.806,
+      "step": 1272
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988464976785766e-05,
+      "loss": 0.6605,
+      "step": 1273
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988407301669695e-05,
+      "loss": 0.8159,
+      "step": 1274
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9883496265536236e-05,
+      "loss": 0.8351,
+      "step": 1275
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9882919514375524e-05,
+      "loss": 0.6969,
+      "step": 1276
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988234276321481e-05,
+      "loss": 0.7097,
+      "step": 1277
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98817660120541e-05,
+      "loss": 0.768,
+      "step": 1278
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.988118926089339e-05,
+      "loss": 0.5455,
+      "step": 1279
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9880612509732678e-05,
+      "loss": 0.8167,
+      "step": 1280
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9880035758571965e-05,
+      "loss": 0.6889,
+      "step": 1281
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9879459007411253e-05,
+      "loss": 0.7087,
+      "step": 1282
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987888225625054e-05,
+      "loss": 0.6468,
+      "step": 1283
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987830550508983e-05,
+      "loss": 0.7927,
+      "step": 1284
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987772875392912e-05,
+      "loss": 0.8436,
+      "step": 1285
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9877152002768407e-05,
+      "loss": 0.7398,
+      "step": 1286
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9876575251607695e-05,
+      "loss": 0.7008,
+      "step": 1287
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9875998500446986e-05,
+      "loss": 0.7245,
+      "step": 1288
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9875421749286274e-05,
+      "loss": 0.8312,
+      "step": 1289
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987484499812556e-05,
+      "loss": 0.8024,
+      "step": 1290
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987426824696485e-05,
+      "loss": 0.6771,
+      "step": 1291
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9873691495804137e-05,
+      "loss": 0.746,
+      "step": 1292
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9873114744643424e-05,
+      "loss": 0.7675,
+      "step": 1293
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9872537993482716e-05,
+      "loss": 0.6997,
+      "step": 1294
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9871961242322003e-05,
+      "loss": 0.6066,
+      "step": 1295
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987138449116129e-05,
+      "loss": 0.6162,
+      "step": 1296
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.987080774000058e-05,
+      "loss": 0.7056,
+      "step": 1297
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9870230988839866e-05,
+      "loss": 0.6995,
+      "step": 1298
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9869654237679154e-05,
+      "loss": 0.6861,
+      "step": 1299
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.986907748651844e-05,
+      "loss": 0.6582,
+      "step": 1300
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9868500735357733e-05,
+      "loss": 0.617,
+      "step": 1301
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.986792398419702e-05,
+      "loss": 0.6587,
+      "step": 1302
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9867347233036308e-05,
+      "loss": 0.7266,
+      "step": 1303
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9866770481875596e-05,
+      "loss": 0.7782,
+      "step": 1304
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9866193730714883e-05,
+      "loss": 0.7844,
+      "step": 1305
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.986561697955417e-05,
+      "loss": 0.6165,
+      "step": 1306
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9865040228393462e-05,
+      "loss": 0.7145,
+      "step": 1307
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.986446347723275e-05,
+      "loss": 0.7774,
+      "step": 1308
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9863886726072038e-05,
+      "loss": 0.5757,
+      "step": 1309
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9863309974911325e-05,
+      "loss": 0.724,
+      "step": 1310
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9862733223750613e-05,
+      "loss": 0.6154,
+      "step": 1311
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98621564725899e-05,
+      "loss": 0.6935,
+      "step": 1312
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9861579721429192e-05,
+      "loss": 0.6979,
+      "step": 1313
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.986100297026848e-05,
+      "loss": 0.7599,
+      "step": 1314
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9860426219107767e-05,
+      "loss": 0.6532,
+      "step": 1315
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9859849467947058e-05,
+      "loss": 0.6943,
+      "step": 1316
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9859272716786342e-05,
+      "loss": 0.6316,
+      "step": 1317
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985869596562563e-05,
+      "loss": 0.603,
+      "step": 1318
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985811921446492e-05,
+      "loss": 0.7158,
+      "step": 1319
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985754246330421e-05,
+      "loss": 0.6667,
+      "step": 1320
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9856965712143497e-05,
+      "loss": 0.7557,
+      "step": 1321
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9856388960982788e-05,
+      "loss": 0.6605,
+      "step": 1322
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9855812209822075e-05,
+      "loss": 0.7249,
+      "step": 1323
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9855235458661363e-05,
+      "loss": 0.5527,
+      "step": 1324
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985465870750065e-05,
+      "loss": 0.6432,
+      "step": 1325
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985408195633994e-05,
+      "loss": 0.7204,
+      "step": 1326
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9853505205179226e-05,
+      "loss": 0.8108,
+      "step": 1327
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9852928454018517e-05,
+      "loss": 0.9199,
+      "step": 1328
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9852351702857805e-05,
+      "loss": 0.7717,
+      "step": 1329
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9851774951697093e-05,
+      "loss": 0.6921,
+      "step": 1330
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.985119820053638e-05,
+      "loss": 0.8706,
+      "step": 1331
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9850621449375668e-05,
+      "loss": 0.7728,
+      "step": 1332
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9850044698214956e-05,
+      "loss": 0.6837,
+      "step": 1333
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9849467947054247e-05,
+      "loss": 0.6643,
+      "step": 1334
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9848891195893534e-05,
+      "loss": 0.7231,
+      "step": 1335
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9848314444732822e-05,
+      "loss": 0.6423,
+      "step": 1336
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.984773769357211e-05,
+      "loss": 0.6455,
+      "step": 1337
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9847160942411397e-05,
+      "loss": 0.7124,
+      "step": 1338
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9846584191250685e-05,
+      "loss": 0.6851,
+      "step": 1339
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9846007440089973e-05,
+      "loss": 0.7305,
+      "step": 1340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845430688929264e-05,
+      "loss": 0.666,
+      "step": 1341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984485393776855e-05,
+      "loss": 0.6174,
+      "step": 1342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984427718660784e-05,
+      "loss": 0.7351,
+      "step": 1343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843700435447127e-05,
+      "loss": 0.6847,
+      "step": 1344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843123684286415e-05,
+      "loss": 0.7312,
+      "step": 1345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842546933125702e-05,
+      "loss": 0.6872,
+      "step": 1346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841970181964993e-05,
+      "loss": 0.7304,
+      "step": 1347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984139343080428e-05,
+      "loss": 0.6597,
+      "step": 1348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984081667964357e-05,
+      "loss": 0.7251,
+      "step": 1349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984023992848286e-05,
+      "loss": 0.7151,
+      "step": 1350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9839663177322148e-05,
+      "loss": 0.8172,
+      "step": 1351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9839086426161435e-05,
+      "loss": 0.652,
+      "step": 1352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9838509675000723e-05,
+      "loss": 0.7019,
+      "step": 1353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.983793292384001e-05,
+      "loss": 0.6845,
+      "step": 1354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9837356172679298e-05,
+      "loss": 0.6517,
+      "step": 1355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.983677942151859e-05,
+      "loss": 0.7218,
+      "step": 1356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9836202670357877e-05,
+      "loss": 0.7638,
+      "step": 1357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9835625919197165e-05,
+      "loss": 0.6437,
+      "step": 1358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9835049168036452e-05,
+      "loss": 0.8001,
+      "step": 1359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.983447241687574e-05,
+      "loss": 0.6961,
+      "step": 1360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9833895665715028e-05,
+      "loss": 0.6428,
+      "step": 1361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.983331891455432e-05,
+      "loss": 0.5929,
+      "step": 1362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9832742163393606e-05,
+      "loss": 0.6971,
+      "step": 1363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9832165412232894e-05,
+      "loss": 0.756,
+      "step": 1364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9831588661072182e-05,
+      "loss": 0.7377,
+      "step": 1365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.983101190991147e-05,
+      "loss": 0.742,
+      "step": 1366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9830435158750757e-05,
+      "loss": 0.718,
+      "step": 1367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982985840759005e-05,
+      "loss": 0.6835,
+      "step": 1368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9829281656429336e-05,
+      "loss": 0.6989,
+      "step": 1369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9828704905268624e-05,
+      "loss": 0.7375,
+      "step": 1370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982812815410791e-05,
+      "loss": 0.7127,
+      "step": 1371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98275514029472e-05,
+      "loss": 0.7396,
+      "step": 1372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9826974651786487e-05,
+      "loss": 0.5578,
+      "step": 1373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9826397900625774e-05,
+      "loss": 0.7014,
+      "step": 1374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9825821149465065e-05,
+      "loss": 0.8089,
+      "step": 1375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9825244398304353e-05,
+      "loss": 0.7451,
+      "step": 1376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982466764714364e-05,
+      "loss": 0.7091,
+      "step": 1377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9824090895982932e-05,
+      "loss": 0.6178,
+      "step": 1378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9823514144822216e-05,
+      "loss": 0.6035,
+      "step": 1379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9822937393661504e-05,
+      "loss": 0.6413,
+      "step": 1380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9822360642500795e-05,
+      "loss": 0.7677,
+      "step": 1381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9821783891340083e-05,
+      "loss": 0.6867,
+      "step": 1382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982120714017937e-05,
+      "loss": 0.5972,
+      "step": 1383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982063038901866e-05,
+      "loss": 0.703,
+      "step": 1384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.982005363785795e-05,
+      "loss": 0.5814,
+      "step": 1385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9819476886697237e-05,
+      "loss": 0.581,
+      "step": 1386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9818900135536524e-05,
+      "loss": 0.7754,
+      "step": 1387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9818323384375812e-05,
+      "loss": 0.5769,
+      "step": 1388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98177466332151e-05,
+      "loss": 0.7464,
+      "step": 1389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981716988205439e-05,
+      "loss": 0.7961,
+      "step": 1390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981659313089368e-05,
+      "loss": 0.5897,
+      "step": 1391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9816016379732966e-05,
+      "loss": 0.76,
+      "step": 1392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9815439628572254e-05,
+      "loss": 0.7252,
+      "step": 1393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981486287741154e-05,
+      "loss": 0.7186,
+      "step": 1394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981428612625083e-05,
+      "loss": 0.7912,
+      "step": 1395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981370937509012e-05,
+      "loss": 0.7465,
+      "step": 1396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9813132623929408e-05,
+      "loss": 0.7088,
+      "step": 1397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9812555872768696e-05,
+      "loss": 0.7059,
+      "step": 1398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9811979121607983e-05,
+      "loss": 0.74,
+      "step": 1399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981140237044727e-05,
+      "loss": 0.7659,
+      "step": 1400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981082561928656e-05,
+      "loss": 0.6722,
+      "step": 1401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.981024886812585e-05,
+      "loss": 0.6929,
+      "step": 1402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9809672116965138e-05,
+      "loss": 0.6643,
+      "step": 1403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9809095365804425e-05,
+      "loss": 0.679,
+      "step": 1404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9808518614643713e-05,
+      "loss": 0.7247,
+      "step": 1405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9807941863483e-05,
+      "loss": 0.6764,
+      "step": 1406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.980736511232229e-05,
+      "loss": 0.7211,
+      "step": 1407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9806788361161576e-05,
+      "loss": 0.6304,
+      "step": 1408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9806211610000867e-05,
+      "loss": 0.74,
+      "step": 1409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9805634858840155e-05,
+      "loss": 0.6769,
+      "step": 1410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9805058107679442e-05,
+      "loss": 0.7029,
+      "step": 1411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9804481356518734e-05,
+      "loss": 0.7719,
+      "step": 1412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.980390460535802e-05,
+      "loss": 0.6264,
+      "step": 1413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.980332785419731e-05,
+      "loss": 0.6701,
+      "step": 1414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9802751103036597e-05,
+      "loss": 0.6121,
+      "step": 1415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9802174351875884e-05,
+      "loss": 0.7663,
+      "step": 1416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9801597600715172e-05,
+      "loss": 0.7285,
+      "step": 1417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9801020849554463e-05,
+      "loss": 0.7297,
+      "step": 1418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.980044409839375e-05,
+      "loss": 0.6781,
+      "step": 1419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979986734723304e-05,
+      "loss": 0.7801,
+      "step": 1420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9799290596072326e-05,
+      "loss": 0.581,
+      "step": 1421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9798713844911614e-05,
+      "loss": 0.73,
+      "step": 1422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.97981370937509e-05,
+      "loss": 0.643,
+      "step": 1423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9797560342590193e-05,
+      "loss": 0.6896,
+      "step": 1424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979698359142948e-05,
+      "loss": 0.8243,
+      "step": 1425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9796406840268768e-05,
+      "loss": 0.7943,
+      "step": 1426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9795830089108056e-05,
+      "loss": 0.7808,
+      "step": 1427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9795253337947343e-05,
+      "loss": 0.7157,
+      "step": 1428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979467658678663e-05,
+      "loss": 0.7034,
+      "step": 1429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9794099835625922e-05,
+      "loss": 0.7224,
+      "step": 1430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979352308446521e-05,
+      "loss": 0.7243,
+      "step": 1431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9792946333304497e-05,
+      "loss": 0.5849,
+      "step": 1432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9792369582143785e-05,
+      "loss": 0.517,
+      "step": 1433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9791792830983073e-05,
+      "loss": 0.7028,
+      "step": 1434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979121607982236e-05,
+      "loss": 0.5507,
+      "step": 1435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979063932866165e-05,
+      "loss": 0.5883,
+      "step": 1436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.979006257750094e-05,
+      "loss": 0.8045,
+      "step": 1437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9789485826340227e-05,
+      "loss": 0.7292,
+      "step": 1438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9788909075179515e-05,
+      "loss": 0.8015,
+      "step": 1439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9788332324018806e-05,
+      "loss": 0.8132,
+      "step": 1440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9787755572858093e-05,
+      "loss": 0.8021,
+      "step": 1441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9787178821697378e-05,
+      "loss": 0.7184,
+      "step": 1442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.978660207053667e-05,
+      "loss": 0.7767,
+      "step": 1443
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9786025319375956e-05,
+      "loss": 0.7406,
+      "step": 1444
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9785448568215244e-05,
+      "loss": 0.7348,
+      "step": 1445
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9784871817054535e-05,
+      "loss": 0.7359,
+      "step": 1446
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9784295065893823e-05,
+      "loss": 0.5295,
+      "step": 1447
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.978371831473311e-05,
+      "loss": 0.7213,
+      "step": 1448
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9783141563572398e-05,
+      "loss": 0.7542,
+      "step": 1449
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9782564812411686e-05,
+      "loss": 0.7467,
+      "step": 1450
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9781988061250974e-05,
+      "loss": 0.7648,
+      "step": 1451
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9781411310090265e-05,
+      "loss": 0.6779,
+      "step": 1452
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9780834558929552e-05,
+      "loss": 0.6635,
+      "step": 1453
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.978025780776884e-05,
+      "loss": 0.6462,
+      "step": 1454
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9779681056608128e-05,
+      "loss": 0.733,
+      "step": 1455
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9779104305447415e-05,
+      "loss": 0.671,
+      "step": 1456
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9778527554286703e-05,
+      "loss": 0.6455,
+      "step": 1457
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9777950803125994e-05,
+      "loss": 0.8354,
+      "step": 1458
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9777374051965282e-05,
+      "loss": 0.652,
+      "step": 1459
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.977679730080457e-05,
+      "loss": 0.67,
+      "step": 1460
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9776220549643857e-05,
+      "loss": 0.5972,
+      "step": 1461
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9775643798483145e-05,
+      "loss": 0.7843,
+      "step": 1462
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9775067047322433e-05,
+      "loss": 0.6514,
+      "step": 1463
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9774490296161724e-05,
+      "loss": 0.6568,
+      "step": 1464
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.977391354500101e-05,
+      "loss": 0.5995,
+      "step": 1465
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.97733367938403e-05,
+      "loss": 0.67,
+      "step": 1466
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.977276004267959e-05,
+      "loss": 0.7053,
+      "step": 1467
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9772183291518874e-05,
+      "loss": 0.5111,
+      "step": 1468
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9771606540358162e-05,
+      "loss": 0.6726,
+      "step": 1469
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9771029789197453e-05,
+      "loss": 0.6362,
+      "step": 1470
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.977045303803674e-05,
+      "loss": 0.77,
+      "step": 1471
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.976987628687603e-05,
+      "loss": 0.6829,
+      "step": 1472
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.976929953571532e-05,
+      "loss": 0.7258,
+      "step": 1473
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9768722784554607e-05,
+      "loss": 0.6309,
+      "step": 1474
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9768146033393895e-05,
+      "loss": 0.7406,
+      "step": 1475
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9767569282233183e-05,
+      "loss": 0.6845,
+      "step": 1476
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.976699253107247e-05,
+      "loss": 0.7208,
+      "step": 1477
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9766415779911758e-05,
+      "loss": 0.7119,
+      "step": 1478
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9765839028751046e-05,
+      "loss": 0.5718,
+      "step": 1479
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9765262277590337e-05,
+      "loss": 0.652,
+      "step": 1480
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9764685526429624e-05,
+      "loss": 0.6778,
+      "step": 1481
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9764108775268912e-05,
+      "loss": 0.7318,
+      "step": 1482
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.97635320241082e-05,
+      "loss": 0.6098,
+      "step": 1483
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9762955272947488e-05,
+      "loss": 0.7695,
+      "step": 1484
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9762378521786775e-05,
+      "loss": 0.6192,
+      "step": 1485
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9761801770626066e-05,
+      "loss": 0.6251,
+      "step": 1486
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9761225019465354e-05,
+      "loss": 0.637,
+      "step": 1487
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.976064826830464e-05,
+      "loss": 0.6722,
+      "step": 1488
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.976007151714393e-05,
+      "loss": 0.7971,
+      "step": 1489
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9759494765983217e-05,
+      "loss": 0.7213,
+      "step": 1490
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9758918014822505e-05,
+      "loss": 0.4555,
+      "step": 1491
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9758341263661796e-05,
+      "loss": 0.6449,
+      "step": 1492
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9757764512501083e-05,
+      "loss": 0.6703,
+      "step": 1493
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.975718776134037e-05,
+      "loss": 0.7263,
+      "step": 1494
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.975661101017966e-05,
+      "loss": 0.6402,
+      "step": 1495
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9756034259018947e-05,
+      "loss": 0.8087,
+      "step": 1496
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9755457507858234e-05,
+      "loss": 0.6887,
+      "step": 1497
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9754880756697525e-05,
+      "loss": 0.6608,
+      "step": 1498
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9754304005536813e-05,
+      "loss": 0.6984,
+      "step": 1499
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.97537272543761e-05,
+      "loss": 0.7666,
+      "step": 1500
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9753150503215392e-05,
+      "loss": 0.7616,
+      "step": 1501
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.975257375205468e-05,
+      "loss": 0.6839,
+      "step": 1502
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9751997000893967e-05,
+      "loss": 0.6354,
+      "step": 1503
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9751420249733255e-05,
+      "loss": 0.5678,
+      "step": 1504
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9750843498572542e-05,
+      "loss": 0.8416,
+      "step": 1505
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.975026674741183e-05,
+      "loss": 0.6276,
+      "step": 1506
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.974968999625112e-05,
+      "loss": 0.7334,
+      "step": 1507
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.974911324509041e-05,
+      "loss": 0.7435,
+      "step": 1508
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9748536493929697e-05,
+      "loss": 0.6475,
+      "step": 1509
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9747959742768984e-05,
+      "loss": 0.7507,
+      "step": 1510
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9747382991608272e-05,
+      "loss": 0.7533,
+      "step": 1511
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.974680624044756e-05,
+      "loss": 0.5408,
+      "step": 1512
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9746229489286847e-05,
+      "loss": 0.8046,
+      "step": 1513
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.974565273812614e-05,
+      "loss": 0.6714,
+      "step": 1514
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9745075986965426e-05,
+      "loss": 0.6823,
+      "step": 1515
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9744499235804714e-05,
+      "loss": 0.7084,
+      "step": 1516
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9743922484644e-05,
+      "loss": 0.743,
+      "step": 1517
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.974334573348329e-05,
+      "loss": 0.803,
+      "step": 1518
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9742768982322577e-05,
+      "loss": 0.6455,
+      "step": 1519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9742192231161868e-05,
+      "loss": 0.6331,
+      "step": 1520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9741615480001156e-05,
+      "loss": 0.6126,
+      "step": 1521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9741038728840443e-05,
+      "loss": 0.8349,
+      "step": 1522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.974046197767973e-05,
+      "loss": 0.7836,
+      "step": 1523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.973988522651902e-05,
+      "loss": 0.7323,
+      "step": 1524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9739308475358306e-05,
+      "loss": 0.679,
+      "step": 1525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9738731724197597e-05,
+      "loss": 0.7767,
+      "step": 1526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9738154973036885e-05,
+      "loss": 0.6444,
+      "step": 1527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9737578221876173e-05,
+      "loss": 0.6136,
+      "step": 1528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9737001470715464e-05,
+      "loss": 0.5735,
+      "step": 1529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9736424719554748e-05,
+      "loss": 0.6767,
+      "step": 1530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9735847968394036e-05,
+      "loss": 0.6786,
+      "step": 1531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9735271217233327e-05,
+      "loss": 0.6708,
+      "step": 1532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9734694466072615e-05,
+      "loss": 0.8027,
+      "step": 1533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9734117714911902e-05,
+      "loss": 0.7382,
+      "step": 1534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9733540963751193e-05,
+      "loss": 0.7382,
+      "step": 1535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.973296421259048e-05,
+      "loss": 0.5995,
+      "step": 1536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.973238746142977e-05,
+      "loss": 0.692,
+      "step": 1537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9731810710269056e-05,
+      "loss": 0.7054,
+      "step": 1538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9731233959108344e-05,
+      "loss": 0.5581,
+      "step": 1539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9730657207947632e-05,
+      "loss": 0.7833,
+      "step": 1540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9730080456786923e-05,
+      "loss": 0.6195,
+      "step": 1541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972950370562621e-05,
+      "loss": 0.7966,
+      "step": 1542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9728926954465498e-05,
+      "loss": 0.7256,
+      "step": 1543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9728350203304786e-05,
+      "loss": 0.7322,
+      "step": 1544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9727773452144074e-05,
+      "loss": 0.5707,
+      "step": 1545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972719670098336e-05,
+      "loss": 0.6657,
+      "step": 1546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972661994982265e-05,
+      "loss": 0.6451,
+      "step": 1547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972604319866194e-05,
+      "loss": 0.6055,
+      "step": 1548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9725466447501228e-05,
+      "loss": 0.7684,
+      "step": 1549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9724889696340515e-05,
+      "loss": 0.6988,
+      "step": 1550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9724312945179803e-05,
+      "loss": 0.6792,
+      "step": 1551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972373619401909e-05,
+      "loss": 0.6694,
+      "step": 1552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972315944285838e-05,
+      "loss": 0.7079,
+      "step": 1553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972258269169767e-05,
+      "loss": 0.592,
+      "step": 1554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9722005940536957e-05,
+      "loss": 0.8342,
+      "step": 1555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9721429189376245e-05,
+      "loss": 0.6669,
+      "step": 1556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9720852438215533e-05,
+      "loss": 0.6066,
+      "step": 1557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.972027568705482e-05,
+      "loss": 0.6641,
+      "step": 1558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9719698935894108e-05,
+      "loss": 0.7425,
+      "step": 1559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97191221847334e-05,
+      "loss": 0.7356,
+      "step": 1560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9718545433572687e-05,
+      "loss": 0.6857,
+      "step": 1561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9717968682411974e-05,
+      "loss": 0.7451,
+      "step": 1562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9717391931251265e-05,
+      "loss": 0.7909,
+      "step": 1563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9716815180090553e-05,
+      "loss": 0.7087,
+      "step": 1564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.971623842892984e-05,
+      "loss": 0.6781,
+      "step": 1565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.971566167776913e-05,
+      "loss": 0.6889,
+      "step": 1566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9715084926608416e-05,
+      "loss": 0.7302,
+      "step": 1567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9714508175447704e-05,
+      "loss": 0.6657,
+      "step": 1568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9713931424286995e-05,
+      "loss": 0.6403,
+      "step": 1569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9713354673126283e-05,
+      "loss": 0.8393,
+      "step": 1570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.971277792196557e-05,
+      "loss": 0.6133,
+      "step": 1571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9712201170804858e-05,
+      "loss": 0.7263,
+      "step": 1572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9711624419644146e-05,
+      "loss": 0.7504,
+      "step": 1573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9711047668483433e-05,
+      "loss": 0.6468,
+      "step": 1574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9710470917322724e-05,
+      "loss": 0.6864,
+      "step": 1575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9709894166162012e-05,
+      "loss": 0.6417,
+      "step": 1576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97093174150013e-05,
+      "loss": 0.6188,
+      "step": 1577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9708740663840588e-05,
+      "loss": 0.6139,
+      "step": 1578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9708163912679875e-05,
+      "loss": 0.6829,
+      "step": 1579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9707587161519163e-05,
+      "loss": 0.7243,
+      "step": 1580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970701041035845e-05,
+      "loss": 0.6215,
+      "step": 1581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970643365919774e-05,
+      "loss": 0.7519,
+      "step": 1582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970585690803703e-05,
+      "loss": 0.5801,
+      "step": 1583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9705280156876317e-05,
+      "loss": 0.7221,
+      "step": 1584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9704703405715605e-05,
+      "loss": 0.7278,
+      "step": 1585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9704126654554892e-05,
+      "loss": 0.6803,
+      "step": 1586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970354990339418e-05,
+      "loss": 0.7525,
+      "step": 1587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970297315223347e-05,
+      "loss": 0.7409,
+      "step": 1588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970239640107276e-05,
+      "loss": 0.7402,
+      "step": 1589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9701819649912047e-05,
+      "loss": 0.7665,
+      "step": 1590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9701242898751338e-05,
+      "loss": 0.7174,
+      "step": 1591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9700666147590622e-05,
+      "loss": 0.7071,
+      "step": 1592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.970008939642991e-05,
+      "loss": 0.6805,
+      "step": 1593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96995126452692e-05,
+      "loss": 0.6906,
+      "step": 1594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.969893589410849e-05,
+      "loss": 0.6395,
+      "step": 1595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9698359142947776e-05,
+      "loss": 0.7508,
+      "step": 1596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9697782391787067e-05,
+      "loss": 0.6572,
+      "step": 1597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9697205640626355e-05,
+      "loss": 0.7944,
+      "step": 1598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9696628889465642e-05,
+      "loss": 0.7093,
+      "step": 1599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.969605213830493e-05,
+      "loss": 0.7355,
+      "step": 1600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9695475387144218e-05,
+      "loss": 0.6928,
+      "step": 1601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9694898635983506e-05,
+      "loss": 0.7876,
+      "step": 1602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9694321884822797e-05,
+      "loss": 0.7067,
+      "step": 1603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9693745133662084e-05,
+      "loss": 0.7456,
+      "step": 1604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9693168382501372e-05,
+      "loss": 0.8226,
+      "step": 1605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.969259163134066e-05,
+      "loss": 0.635,
+      "step": 1606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9692014880179947e-05,
+      "loss": 0.7144,
+      "step": 1607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9691438129019235e-05,
+      "loss": 0.5722,
+      "step": 1608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9690861377858526e-05,
+      "loss": 0.7071,
+      "step": 1609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9690284626697814e-05,
+      "loss": 0.6676,
+      "step": 1610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96897078755371e-05,
+      "loss": 0.6907,
+      "step": 1611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.968913112437639e-05,
+      "loss": 0.7206,
+      "step": 1612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9688554373215677e-05,
+      "loss": 0.6193,
+      "step": 1613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9687977622054965e-05,
+      "loss": 0.7494,
+      "step": 1614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9687400870894252e-05,
+      "loss": 0.6529,
+      "step": 1615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9686824119733543e-05,
+      "loss": 0.7491,
+      "step": 1616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.968624736857283e-05,
+      "loss": 0.7769,
+      "step": 1617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.968567061741212e-05,
+      "loss": 0.6298,
+      "step": 1618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9685093866251406e-05,
+      "loss": 0.7018,
+      "step": 1619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9684517115090694e-05,
+      "loss": 0.7392,
+      "step": 1620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9683940363929982e-05,
+      "loss": 0.662,
+      "step": 1621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9683363612769273e-05,
+      "loss": 0.6772,
+      "step": 1622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.968278686160856e-05,
+      "loss": 0.7521,
+      "step": 1623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9682210110447848e-05,
+      "loss": 0.6066,
+      "step": 1624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.968163335928714e-05,
+      "loss": 0.7552,
+      "step": 1625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9681056608126427e-05,
+      "loss": 0.7931,
+      "step": 1626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9680479856965715e-05,
+      "loss": 0.6724,
+      "step": 1627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9679903105805002e-05,
+      "loss": 0.5628,
+      "step": 1628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.967932635464429e-05,
+      "loss": 0.7122,
+      "step": 1629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9678749603483578e-05,
+      "loss": 0.7245,
+      "step": 1630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.967817285232287e-05,
+      "loss": 0.7608,
+      "step": 1631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9677596101162156e-05,
+      "loss": 0.7038,
+      "step": 1632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9677019350001444e-05,
+      "loss": 0.5632,
+      "step": 1633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9676442598840732e-05,
+      "loss": 0.7288,
+      "step": 1634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.967586584768002e-05,
+      "loss": 0.6816,
+      "step": 1635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9675289096519307e-05,
+      "loss": 0.7771,
+      "step": 1636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9674712345358598e-05,
+      "loss": 0.6995,
+      "step": 1637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9674135594197886e-05,
+      "loss": 0.665,
+      "step": 1638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9673558843037174e-05,
+      "loss": 0.6796,
+      "step": 1639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.967298209187646e-05,
+      "loss": 0.8144,
+      "step": 1640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.967240534071575e-05,
+      "loss": 0.705,
+      "step": 1641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9671828589555037e-05,
+      "loss": 0.7018,
+      "step": 1642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9671251838394328e-05,
+      "loss": 0.6723,
+      "step": 1643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9670675087233615e-05,
+      "loss": 0.9327,
+      "step": 1644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9670098336072903e-05,
+      "loss": 0.772,
+      "step": 1645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966952158491219e-05,
+      "loss": 0.692,
+      "step": 1646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966894483375148e-05,
+      "loss": 0.6569,
+      "step": 1647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9668368082590766e-05,
+      "loss": 0.693,
+      "step": 1648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9667791331430054e-05,
+      "loss": 0.6182,
+      "step": 1649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9667214580269345e-05,
+      "loss": 0.8067,
+      "step": 1650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9666637829108633e-05,
+      "loss": 0.6974,
+      "step": 1651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966606107794792e-05,
+      "loss": 0.5766,
+      "step": 1652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966548432678721e-05,
+      "loss": 0.6951,
+      "step": 1653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9664907575626496e-05,
+      "loss": 0.6892,
+      "step": 1654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9664330824465783e-05,
+      "loss": 0.6587,
+      "step": 1655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9663754073305074e-05,
+      "loss": 0.7394,
+      "step": 1656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9663177322144362e-05,
+      "loss": 0.5983,
+      "step": 1657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966260057098365e-05,
+      "loss": 0.6234,
+      "step": 1658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966202381982294e-05,
+      "loss": 0.8111,
+      "step": 1659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.966144706866223e-05,
+      "loss": 0.7028,
+      "step": 1660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9660870317501516e-05,
+      "loss": 0.7055,
+      "step": 1661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9660293566340804e-05,
+      "loss": 0.715,
+      "step": 1662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965971681518009e-05,
+      "loss": 0.6741,
+      "step": 1663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965914006401938e-05,
+      "loss": 0.6764,
+      "step": 1664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965856331285867e-05,
+      "loss": 0.6538,
+      "step": 1665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9657986561697958e-05,
+      "loss": 0.7187,
+      "step": 1666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9657409810537246e-05,
+      "loss": 0.6379,
+      "step": 1667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9656833059376533e-05,
+      "loss": 0.5624,
+      "step": 1668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965625630821582e-05,
+      "loss": 0.611,
+      "step": 1669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965567955705511e-05,
+      "loss": 0.6461,
+      "step": 1670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96551028058944e-05,
+      "loss": 0.6381,
+      "step": 1671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9654526054733688e-05,
+      "loss": 0.6623,
+      "step": 1672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9653949303572975e-05,
+      "loss": 0.731,
+      "step": 1673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9653372552412263e-05,
+      "loss": 0.7022,
+      "step": 1674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965279580125155e-05,
+      "loss": 0.6798,
+      "step": 1675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9652219050090838e-05,
+      "loss": 0.6091,
+      "step": 1676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.965164229893013e-05,
+      "loss": 0.6806,
+      "step": 1677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9651065547769417e-05,
+      "loss": 0.6715,
+      "step": 1678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9650488796608705e-05,
+      "loss": 0.7782,
+      "step": 1679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9649912045447996e-05,
+      "loss": 0.7624,
+      "step": 1680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964933529428728e-05,
+      "loss": 0.7939,
+      "step": 1681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9648758543126568e-05,
+      "loss": 0.8521,
+      "step": 1682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9648181791965855e-05,
+      "loss": 0.6941,
+      "step": 1683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9647605040805147e-05,
+      "loss": 0.8209,
+      "step": 1684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9647028289644434e-05,
+      "loss": 0.7219,
+      "step": 1685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9646451538483722e-05,
+      "loss": 0.5341,
+      "step": 1686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9645874787323013e-05,
+      "loss": 0.705,
+      "step": 1687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.96452980361623e-05,
+      "loss": 0.7874,
+      "step": 1688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964472128500159e-05,
+      "loss": 0.7424,
+      "step": 1689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9644144533840876e-05,
+      "loss": 0.6825,
+      "step": 1690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9643567782680164e-05,
+      "loss": 0.6529,
+      "step": 1691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964299103151945e-05,
+      "loss": 0.7234,
+      "step": 1692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9642414280358742e-05,
+      "loss": 0.6334,
+      "step": 1693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.964183752919803e-05,
+      "loss": 0.6778,
+      "step": 1694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9641260778037318e-05,
+      "loss": 0.6986,
+      "step": 1695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9640684026876606e-05,
+      "loss": 0.6278,
+      "step": 1696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9640107275715893e-05,
+      "loss": 0.8004,
+      "step": 1697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.963953052455518e-05,
+      "loss": 0.7073,
+      "step": 1698
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9638953773394472e-05,
+      "loss": 0.6639,
+      "step": 1699
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963837702223376e-05,
+      "loss": 0.8043,
+      "step": 1700
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9637800271073047e-05,
+      "loss": 0.7824,
+      "step": 1701
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9637223519912335e-05,
+      "loss": 0.6561,
+      "step": 1702
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9636646768751623e-05,
+      "loss": 0.7006,
+      "step": 1703
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963607001759091e-05,
+      "loss": 0.6876,
+      "step": 1704
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.96354932664302e-05,
+      "loss": 0.8392,
+      "step": 1705
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963491651526949e-05,
+      "loss": 0.7919,
+      "step": 1706
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9634339764108777e-05,
+      "loss": 0.7301,
+      "step": 1707
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9633763012948065e-05,
+      "loss": 0.696,
+      "step": 1708
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9633186261787352e-05,
+      "loss": 0.6609,
+      "step": 1709
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963260951062664e-05,
+      "loss": 0.7384,
+      "step": 1710
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963203275946593e-05,
+      "loss": 0.6292,
+      "step": 1711
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.963145600830522e-05,
+      "loss": 0.708,
+      "step": 1712
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9630879257144506e-05,
+      "loss": 0.6461,
+      "step": 1713
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9630302505983797e-05,
+      "loss": 0.6854,
+      "step": 1714
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9629725754823085e-05,
+      "loss": 0.534,
+      "step": 1715
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.962914900366237e-05,
+      "loss": 0.6842,
+      "step": 1716
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.962857225250166e-05,
+      "loss": 0.712,
+      "step": 1717
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9627995501340948e-05,
+      "loss": 0.7133,
+      "step": 1718
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9627418750180236e-05,
+      "loss": 0.7249,
+      "step": 1719
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9626841999019524e-05,
+      "loss": 0.8198,
+      "step": 1720
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9626265247858815e-05,
+      "loss": 0.7044,
+      "step": 1721
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9625688496698102e-05,
+      "loss": 0.8188,
+      "step": 1722
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.962511174553739e-05,
+      "loss": 0.6824,
+      "step": 1723
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9624534994376678e-05,
+      "loss": 0.5906,
+      "step": 1724
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9623958243215965e-05,
+      "loss": 0.7173,
+      "step": 1725
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9623381492055253e-05,
+      "loss": 0.797,
+      "step": 1726
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9622804740894544e-05,
+      "loss": 0.6093,
+      "step": 1727
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9622227989733832e-05,
+      "loss": 0.7463,
+      "step": 1728
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.962165123857312e-05,
+      "loss": 0.6777,
+      "step": 1729
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9621074487412407e-05,
+      "loss": 0.6752,
+      "step": 1730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9620497736251695e-05,
+      "loss": 0.7042,
+      "step": 1731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9619920985090983e-05,
+      "loss": 0.6352,
+      "step": 1732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9619344233930274e-05,
+      "loss": 0.6713,
+      "step": 1733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961876748276956e-05,
+      "loss": 0.6508,
+      "step": 1734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961819073160885e-05,
+      "loss": 0.7307,
+      "step": 1735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9617613980448137e-05,
+      "loss": 0.8204,
+      "step": 1736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9617037229287424e-05,
+      "loss": 0.7438,
+      "step": 1737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9616460478126712e-05,
+      "loss": 0.6955,
+      "step": 1738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9615883726966003e-05,
+      "loss": 0.6176,
+      "step": 1739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961530697580529e-05,
+      "loss": 0.7814,
+      "step": 1740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961473022464458e-05,
+      "loss": 0.7079,
+      "step": 1741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961415347348387e-05,
+      "loss": 0.6258,
+      "step": 1742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9613576722323154e-05,
+      "loss": 0.5775,
+      "step": 1743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961299997116244e-05,
+      "loss": 0.7203,
+      "step": 1744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9612423220001733e-05,
+      "loss": 0.7613,
+      "step": 1745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.961184646884102e-05,
+      "loss": 0.7481,
+      "step": 1746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9611269717680308e-05,
+      "loss": 0.8233,
+      "step": 1747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.96106929665196e-05,
+      "loss": 0.7466,
+      "step": 1748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9610116215358887e-05,
+      "loss": 0.7283,
+      "step": 1749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9609539464198174e-05,
+      "loss": 0.5595,
+      "step": 1750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9608962713037462e-05,
+      "loss": 0.6601,
+      "step": 1751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.960838596187675e-05,
+      "loss": 0.6729,
+      "step": 1752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9607809210716037e-05,
+      "loss": 0.6556,
+      "step": 1753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9607232459555325e-05,
+      "loss": 0.6718,
+      "step": 1754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9606655708394616e-05,
+      "loss": 0.8055,
+      "step": 1755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9606078957233904e-05,
+      "loss": 0.6376,
+      "step": 1756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.960550220607319e-05,
+      "loss": 0.6333,
+      "step": 1757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.960492545491248e-05,
+      "loss": 0.9173,
+      "step": 1758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9604348703751767e-05,
+      "loss": 0.7541,
+      "step": 1759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9603771952591055e-05,
+      "loss": 0.5551,
+      "step": 1760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9603195201430346e-05,
+      "loss": 0.6956,
+      "step": 1761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9602618450269633e-05,
+      "loss": 0.6584,
+      "step": 1762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.960204169910892e-05,
+      "loss": 0.7367,
+      "step": 1763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.960146494794821e-05,
+      "loss": 0.7272,
+      "step": 1764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9600888196787496e-05,
+      "loss": 0.6987,
+      "step": 1765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9600311445626784e-05,
+      "loss": 0.7587,
+      "step": 1766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9599734694466075e-05,
+      "loss": 0.6533,
+      "step": 1767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9599157943305363e-05,
+      "loss": 0.7375,
+      "step": 1768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959858119214465e-05,
+      "loss": 0.7056,
+      "step": 1769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9598004440983938e-05,
+      "loss": 0.7048,
+      "step": 1770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9597427689823226e-05,
+      "loss": 0.6893,
+      "step": 1771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9596850938662514e-05,
+      "loss": 0.7187,
+      "step": 1772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9596274187501805e-05,
+      "loss": 0.7282,
+      "step": 1773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9595697436341092e-05,
+      "loss": 0.6946,
+      "step": 1774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959512068518038e-05,
+      "loss": 0.7147,
+      "step": 1775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959454393401967e-05,
+      "loss": 0.721,
+      "step": 1776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959396718285896e-05,
+      "loss": 0.6683,
+      "step": 1777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9593390431698243e-05,
+      "loss": 0.8156,
+      "step": 1778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9592813680537534e-05,
+      "loss": 0.692,
+      "step": 1779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9592236929376822e-05,
+      "loss": 0.6071,
+      "step": 1780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959166017821611e-05,
+      "loss": 0.791,
+      "step": 1781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95910834270554e-05,
+      "loss": 0.7139,
+      "step": 1782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.959050667589469e-05,
+      "loss": 0.7156,
+      "step": 1783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9589929924733976e-05,
+      "loss": 0.7339,
+      "step": 1784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9589353173573264e-05,
+      "loss": 0.6558,
+      "step": 1785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.958877642241255e-05,
+      "loss": 0.68,
+      "step": 1786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.958819967125184e-05,
+      "loss": 0.764,
+      "step": 1787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9587622920091127e-05,
+      "loss": 0.7405,
+      "step": 1788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9587046168930418e-05,
+      "loss": 0.5456,
+      "step": 1789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9586469417769706e-05,
+      "loss": 0.6829,
+      "step": 1790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9585892666608993e-05,
+      "loss": 0.6779,
+      "step": 1791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.958531591544828e-05,
+      "loss": 0.6906,
+      "step": 1792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.958473916428757e-05,
+      "loss": 0.6808,
+      "step": 1793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9584162413126856e-05,
+      "loss": 0.7799,
+      "step": 1794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9583585661966147e-05,
+      "loss": 0.6477,
+      "step": 1795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9583008910805435e-05,
+      "loss": 0.7287,
+      "step": 1796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9582432159644723e-05,
+      "loss": 0.6906,
+      "step": 1797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.958185540848401e-05,
+      "loss": 0.6698,
+      "step": 1798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9581278657323298e-05,
+      "loss": 0.7696,
+      "step": 1799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9580701906162586e-05,
+      "loss": 0.7634,
+      "step": 1800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9580125155001877e-05,
+      "loss": 0.7127,
+      "step": 1801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9579548403841165e-05,
+      "loss": 0.7277,
+      "step": 1802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9578971652680452e-05,
+      "loss": 0.7002,
+      "step": 1803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9578394901519743e-05,
+      "loss": 0.7174,
+      "step": 1804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9577818150359028e-05,
+      "loss": 0.6338,
+      "step": 1805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9577241399198315e-05,
+      "loss": 0.6513,
+      "step": 1806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9576664648037606e-05,
+      "loss": 0.8134,
+      "step": 1807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9576087896876894e-05,
+      "loss": 0.6449,
+      "step": 1808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9575511145716182e-05,
+      "loss": 0.7198,
+      "step": 1809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9574934394555473e-05,
+      "loss": 0.6604,
+      "step": 1810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.957435764339476e-05,
+      "loss": 0.7738,
+      "step": 1811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9573780892234048e-05,
+      "loss": 0.6276,
+      "step": 1812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9573204141073336e-05,
+      "loss": 0.7433,
+      "step": 1813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9572627389912624e-05,
+      "loss": 0.5911,
+      "step": 1814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.957205063875191e-05,
+      "loss": 0.6018,
+      "step": 1815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9571473887591202e-05,
+      "loss": 0.5882,
+      "step": 1816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.957089713643049e-05,
+      "loss": 0.8416,
+      "step": 1817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9570320385269778e-05,
+      "loss": 0.7037,
+      "step": 1818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9569743634109065e-05,
+      "loss": 0.729,
+      "step": 1819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9569166882948353e-05,
+      "loss": 0.6569,
+      "step": 1820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956859013178764e-05,
+      "loss": 0.6394,
+      "step": 1821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956801338062693e-05,
+      "loss": 0.6254,
+      "step": 1822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956743662946622e-05,
+      "loss": 0.6947,
+      "step": 1823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9566859878305507e-05,
+      "loss": 0.6572,
+      "step": 1824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9566283127144795e-05,
+      "loss": 0.734,
+      "step": 1825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9565706375984083e-05,
+      "loss": 0.6403,
+      "step": 1826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956512962482337e-05,
+      "loss": 0.674,
+      "step": 1827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9564552873662658e-05,
+      "loss": 0.6423,
+      "step": 1828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956397612250195e-05,
+      "loss": 0.6354,
+      "step": 1829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9563399371341237e-05,
+      "loss": 0.6535,
+      "step": 1830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9562822620180524e-05,
+      "loss": 0.6706,
+      "step": 1831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9562245869019812e-05,
+      "loss": 0.6981,
+      "step": 1832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95616691178591e-05,
+      "loss": 0.7064,
+      "step": 1833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9561092366698387e-05,
+      "loss": 0.767,
+      "step": 1834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.956051561553768e-05,
+      "loss": 0.7752,
+      "step": 1835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9559938864376966e-05,
+      "loss": 0.647,
+      "step": 1836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9559362113216254e-05,
+      "loss": 0.6399,
+      "step": 1837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9558785362055545e-05,
+      "loss": 0.7454,
+      "step": 1838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9558208610894833e-05,
+      "loss": 0.6899,
+      "step": 1839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955763185973412e-05,
+      "loss": 0.7292,
+      "step": 1840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9557055108573408e-05,
+      "loss": 0.7268,
+      "step": 1841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9556478357412696e-05,
+      "loss": 0.7268,
+      "step": 1842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9555901606251983e-05,
+      "loss": 0.8391,
+      "step": 1843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9555324855091274e-05,
+      "loss": 0.6478,
+      "step": 1844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9554748103930562e-05,
+      "loss": 0.6248,
+      "step": 1845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955417135276985e-05,
+      "loss": 0.6715,
+      "step": 1846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9553594601609137e-05,
+      "loss": 0.6122,
+      "step": 1847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9553017850448425e-05,
+      "loss": 0.5871,
+      "step": 1848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9552441099287713e-05,
+      "loss": 0.7448,
+      "step": 1849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9551864348127004e-05,
+      "loss": 0.6899,
+      "step": 1850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955128759696629e-05,
+      "loss": 0.7233,
+      "step": 1851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.955071084580558e-05,
+      "loss": 0.6622,
+      "step": 1852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9550134094644867e-05,
+      "loss": 0.6956,
+      "step": 1853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9549557343484155e-05,
+      "loss": 0.6776,
+      "step": 1854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9548980592323442e-05,
+      "loss": 0.6782,
+      "step": 1855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9548403841162733e-05,
+      "loss": 0.7451,
+      "step": 1856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954782709000202e-05,
+      "loss": 0.7114,
+      "step": 1857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954725033884131e-05,
+      "loss": 0.6615,
+      "step": 1858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9546673587680596e-05,
+      "loss": 0.74,
+      "step": 1859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9546096836519884e-05,
+      "loss": 0.7406,
+      "step": 1860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9545520085359172e-05,
+      "loss": 0.709,
+      "step": 1861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954494333419846e-05,
+      "loss": 0.6347,
+      "step": 1862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954436658303775e-05,
+      "loss": 0.7096,
+      "step": 1863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9543789831877038e-05,
+      "loss": 0.6439,
+      "step": 1864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9543213080716326e-05,
+      "loss": 0.755,
+      "step": 1865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9542636329555617e-05,
+      "loss": 0.5706,
+      "step": 1866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.95420595783949e-05,
+      "loss": 0.8155,
+      "step": 1867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954148282723419e-05,
+      "loss": 0.7294,
+      "step": 1868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.954090607607348e-05,
+      "loss": 0.6766,
+      "step": 1869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9540329324912768e-05,
+      "loss": 0.7581,
+      "step": 1870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9539752573752055e-05,
+      "loss": 0.6068,
+      "step": 1871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9539175822591347e-05,
+      "loss": 0.7171,
+      "step": 1872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9538599071430634e-05,
+      "loss": 0.6309,
+      "step": 1873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9538022320269922e-05,
+      "loss": 0.7178,
+      "step": 1874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.953744556910921e-05,
+      "loss": 0.7408,
+      "step": 1875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9536868817948497e-05,
+      "loss": 0.7582,
+      "step": 1876
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9536292066787785e-05,
+      "loss": 0.7419,
+      "step": 1877
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9535715315627076e-05,
+      "loss": 0.7412,
+      "step": 1878
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9535138564466364e-05,
+      "loss": 0.7582,
+      "step": 1879
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.953456181330565e-05,
+      "loss": 0.6171,
+      "step": 1880
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.953398506214494e-05,
+      "loss": 0.7582,
+      "step": 1881
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9533408310984227e-05,
+      "loss": 0.6538,
+      "step": 1882
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9532831559823514e-05,
+      "loss": 0.6679,
+      "step": 1883
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9532254808662806e-05,
+      "loss": 0.7622,
+      "step": 1884
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9531678057502093e-05,
+      "loss": 0.6271,
+      "step": 1885
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.953110130634138e-05,
+      "loss": 0.7969,
+      "step": 1886
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.953052455518067e-05,
+      "loss": 0.6919,
+      "step": 1887
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9529947804019956e-05,
+      "loss": 0.6021,
+      "step": 1888
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9529371052859244e-05,
+      "loss": 0.7462,
+      "step": 1889
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9528794301698535e-05,
+      "loss": 0.7407,
+      "step": 1890
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9528217550537823e-05,
+      "loss": 0.73,
+      "step": 1891
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.952764079937711e-05,
+      "loss": 0.7174,
+      "step": 1892
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9527064048216398e-05,
+      "loss": 0.7095,
+      "step": 1893
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9526487297055686e-05,
+      "loss": 0.7382,
+      "step": 1894
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9525910545894973e-05,
+      "loss": 0.6394,
+      "step": 1895
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.952533379473426e-05,
+      "loss": 0.6821,
+      "step": 1896
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9524757043573552e-05,
+      "loss": 0.6422,
+      "step": 1897
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.952418029241284e-05,
+      "loss": 0.6721,
+      "step": 1898
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9523603541252128e-05,
+      "loss": 0.6684,
+      "step": 1899
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.952302679009142e-05,
+      "loss": 0.6912,
+      "step": 1900
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9522450038930706e-05,
+      "loss": 0.6779,
+      "step": 1901
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9521873287769994e-05,
+      "loss": 0.7695,
+      "step": 1902
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9521296536609282e-05,
+      "loss": 0.6975,
+      "step": 1903
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.952071978544857e-05,
+      "loss": 0.6149,
+      "step": 1904
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9520143034287857e-05,
+      "loss": 0.7404,
+      "step": 1905
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9519566283127148e-05,
+      "loss": 0.7254,
+      "step": 1906
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9518989531966436e-05,
+      "loss": 0.6212,
+      "step": 1907
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9518412780805724e-05,
+      "loss": 0.7008,
+      "step": 1908
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.951783602964501e-05,
+      "loss": 0.7047,
+      "step": 1909
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.95172592784843e-05,
+      "loss": 0.6255,
+      "step": 1910
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9516682527323587e-05,
+      "loss": 0.7504,
+      "step": 1911
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9516105776162878e-05,
+      "loss": 0.6615,
+      "step": 1912
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9515529025002165e-05,
+      "loss": 0.6344,
+      "step": 1913
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9514952273841453e-05,
+      "loss": 0.824,
+      "step": 1914
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.951437552268074e-05,
+      "loss": 0.6588,
+      "step": 1915
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.951379877152003e-05,
+      "loss": 0.5709,
+      "step": 1916
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9513222020359316e-05,
+      "loss": 0.6658,
+      "step": 1917
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9512645269198607e-05,
+      "loss": 0.7046,
+      "step": 1918
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9512068518037895e-05,
+      "loss": 0.6898,
+      "step": 1919
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9511491766877183e-05,
+      "loss": 0.7088,
+      "step": 1920
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.951091501571647e-05,
+      "loss": 0.8139,
+      "step": 1921
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9510338264555758e-05,
+      "loss": 0.7519,
+      "step": 1922
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9509761513395046e-05,
+      "loss": 0.6266,
+      "step": 1923
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9509184762234337e-05,
+      "loss": 0.6311,
+      "step": 1924
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9508608011073624e-05,
+      "loss": 0.8647,
+      "step": 1925
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9508031259912912e-05,
+      "loss": 0.7033,
+      "step": 1926
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.95074545087522e-05,
+      "loss": 0.7114,
+      "step": 1927
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950687775759149e-05,
+      "loss": 0.7322,
+      "step": 1928
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9506301006430775e-05,
+      "loss": 0.5621,
+      "step": 1929
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9505724255270063e-05,
+      "loss": 0.585,
+      "step": 1930
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9505147504109354e-05,
+      "loss": 0.6954,
+      "step": 1931
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950457075294864e-05,
+      "loss": 0.6438,
+      "step": 1932
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950399400178793e-05,
+      "loss": 0.6766,
+      "step": 1933
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950341725062722e-05,
+      "loss": 0.6429,
+      "step": 1934
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9502840499466508e-05,
+      "loss": 0.6969,
+      "step": 1935
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9502263748305796e-05,
+      "loss": 0.7144,
+      "step": 1936
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9501686997145083e-05,
+      "loss": 0.6457,
+      "step": 1937
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950111024598437e-05,
+      "loss": 0.6971,
+      "step": 1938
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.950053349482366e-05,
+      "loss": 0.6675,
+      "step": 1939
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.949995674366295e-05,
+      "loss": 0.6977,
+      "step": 1940
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9499379992502237e-05,
+      "loss": 0.609,
+      "step": 1941
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9498803241341525e-05,
+      "loss": 0.6401,
+      "step": 1942
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9498226490180813e-05,
+      "loss": 0.7597,
+      "step": 1943
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94976497390201e-05,
+      "loss": 0.6126,
+      "step": 1944
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9497072987859388e-05,
+      "loss": 0.6806,
+      "step": 1945
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.949649623669868e-05,
+      "loss": 0.7453,
+      "step": 1946
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9495919485537967e-05,
+      "loss": 0.7817,
+      "step": 1947
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9495342734377255e-05,
+      "loss": 0.7981,
+      "step": 1948
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9494765983216542e-05,
+      "loss": 0.6649,
+      "step": 1949
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.949418923205583e-05,
+      "loss": 0.7447,
+      "step": 1950
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9493612480895118e-05,
+      "loss": 0.7012,
+      "step": 1951
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.949303572973441e-05,
+      "loss": 0.7775,
+      "step": 1952
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9492458978573696e-05,
+      "loss": 0.629,
+      "step": 1953
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9491882227412984e-05,
+      "loss": 0.8037,
+      "step": 1954
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9491305476252275e-05,
+      "loss": 0.8247,
+      "step": 1955
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.949072872509156e-05,
+      "loss": 0.6555,
+      "step": 1956
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9490151973930847e-05,
+      "loss": 0.6637,
+      "step": 1957
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9489575222770138e-05,
+      "loss": 0.6323,
+      "step": 1958
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9488998471609426e-05,
+      "loss": 0.5558,
+      "step": 1959
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9488421720448714e-05,
+      "loss": 0.6494,
+      "step": 1960
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9487844969288e-05,
+      "loss": 0.6803,
+      "step": 1961
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9487268218127292e-05,
+      "loss": 0.7012,
+      "step": 1962
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.948669146696658e-05,
+      "loss": 0.74,
+      "step": 1963
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9486114715805868e-05,
+      "loss": 0.6652,
+      "step": 1964
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9485537964645155e-05,
+      "loss": 0.7035,
+      "step": 1965
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9484961213484443e-05,
+      "loss": 0.7515,
+      "step": 1966
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.948438446232373e-05,
+      "loss": 0.7493,
+      "step": 1967
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9483807711163022e-05,
+      "loss": 0.568,
+      "step": 1968
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.948323096000231e-05,
+      "loss": 0.7817,
+      "step": 1969
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9482654208841597e-05,
+      "loss": 0.7023,
+      "step": 1970
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9482077457680885e-05,
+      "loss": 0.6397,
+      "step": 1971
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9481500706520173e-05,
+      "loss": 0.736,
+      "step": 1972
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.948092395535946e-05,
+      "loss": 0.6258,
+      "step": 1973
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.948034720419875e-05,
+      "loss": 0.7489,
+      "step": 1974
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947977045303804e-05,
+      "loss": 0.5669,
+      "step": 1975
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9479193701877327e-05,
+      "loss": 0.5606,
+      "step": 1976
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9478616950716614e-05,
+      "loss": 0.6998,
+      "step": 1977
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9478040199555902e-05,
+      "loss": 0.7805,
+      "step": 1978
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947746344839519e-05,
+      "loss": 0.7695,
+      "step": 1979
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947688669723448e-05,
+      "loss": 0.6386,
+      "step": 1980
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947630994607377e-05,
+      "loss": 0.7115,
+      "step": 1981
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9475733194913056e-05,
+      "loss": 0.6032,
+      "step": 1982
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9475156443752344e-05,
+      "loss": 0.723,
+      "step": 1983
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947457969259163e-05,
+      "loss": 0.7229,
+      "step": 1984
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947400294143092e-05,
+      "loss": 0.6971,
+      "step": 1985
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947342619027021e-05,
+      "loss": 0.6102,
+      "step": 1986
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9472849439109498e-05,
+      "loss": 0.6033,
+      "step": 1987
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9472272687948786e-05,
+      "loss": 0.6157,
+      "step": 1988
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9471695936788077e-05,
+      "loss": 0.527,
+      "step": 1989
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9471119185627365e-05,
+      "loss": 0.6581,
+      "step": 1990
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.947054243446665e-05,
+      "loss": 0.6455,
+      "step": 1991
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.946996568330594e-05,
+      "loss": 0.5856,
+      "step": 1992
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9469388932145228e-05,
+      "loss": 0.636,
+      "step": 1993
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9468812180984515e-05,
+      "loss": 0.7506,
+      "step": 1994
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9468235429823806e-05,
+      "loss": 0.6874,
+      "step": 1995
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9467658678663094e-05,
+      "loss": 0.647,
+      "step": 1996
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9467081927502382e-05,
+      "loss": 0.6738,
+      "step": 1997
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.946650517634167e-05,
+      "loss": 0.6917,
+      "step": 1998
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9465928425180957e-05,
+      "loss": 0.7489,
+      "step": 1999
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9465351674020245e-05,
+      "loss": 0.6698,
+      "step": 2000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9464774922859532e-05,
+      "loss": 0.6954,
+      "step": 2001
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9464198171698824e-05,
+      "loss": 0.6414,
+      "step": 2002
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.946362142053811e-05,
+      "loss": 0.6644,
+      "step": 2003
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94630446693774e-05,
+      "loss": 0.6362,
+      "step": 2004
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9462467918216687e-05,
+      "loss": 0.6604,
+      "step": 2005
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9461891167055974e-05,
+      "loss": 0.6953,
+      "step": 2006
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9461314415895262e-05,
+      "loss": 0.7094,
+      "step": 2007
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9460737664734553e-05,
+      "loss": 0.6027,
+      "step": 2008
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.946016091357384e-05,
+      "loss": 0.7849,
+      "step": 2009
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945958416241313e-05,
+      "loss": 0.6782,
+      "step": 2010
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9459007411252416e-05,
+      "loss": 0.6543,
+      "step": 2011
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9458430660091704e-05,
+      "loss": 0.7252,
+      "step": 2012
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945785390893099e-05,
+      "loss": 0.6262,
+      "step": 2013
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9457277157770283e-05,
+      "loss": 0.6738,
+      "step": 2014
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945670040660957e-05,
+      "loss": 0.7092,
+      "step": 2015
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9456123655448858e-05,
+      "loss": 0.6117,
+      "step": 2016
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945554690428815e-05,
+      "loss": 0.7739,
+      "step": 2017
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9454970153127433e-05,
+      "loss": 0.7855,
+      "step": 2018
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945439340196672e-05,
+      "loss": 0.7073,
+      "step": 2019
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9453816650806012e-05,
+      "loss": 0.7519,
+      "step": 2020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94532398996453e-05,
+      "loss": 0.6076,
+      "step": 2021
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9452663148484587e-05,
+      "loss": 0.6382,
+      "step": 2022
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945208639732388e-05,
+      "loss": 0.7021,
+      "step": 2023
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9451509646163166e-05,
+      "loss": 0.6284,
+      "step": 2024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9450932895002454e-05,
+      "loss": 0.7254,
+      "step": 2025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.945035614384174e-05,
+      "loss": 0.6003,
+      "step": 2026
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944977939268103e-05,
+      "loss": 0.6737,
+      "step": 2027
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9449202641520317e-05,
+      "loss": 0.5995,
+      "step": 2028
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9448625890359608e-05,
+      "loss": 0.6023,
+      "step": 2029
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9448049139198896e-05,
+      "loss": 0.7522,
+      "step": 2030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9447472388038183e-05,
+      "loss": 0.6314,
+      "step": 2031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944689563687747e-05,
+      "loss": 0.7656,
+      "step": 2032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.944631888571676e-05,
+      "loss": 0.6147,
+      "step": 2033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9445742134556046e-05,
+      "loss": 0.6057,
+      "step": 2034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9445165383395334e-05,
+      "loss": 0.7175,
+      "step": 2035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9444588632234625e-05,
+      "loss": 0.6236,
+      "step": 2036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9444011881073913e-05,
+      "loss": 0.7539,
+      "step": 2037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94434351299132e-05,
+      "loss": 0.6799,
+      "step": 2038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9442858378752488e-05,
+      "loss": 0.7081,
+      "step": 2039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9442281627591776e-05,
+      "loss": 0.7157,
+      "step": 2040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9441704876431064e-05,
+      "loss": 0.708,
+      "step": 2041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9441128125270355e-05,
+      "loss": 0.7653,
+      "step": 2042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9440551374109642e-05,
+      "loss": 0.6357,
+      "step": 2043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943997462294893e-05,
+      "loss": 0.6657,
+      "step": 2044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9439397871788218e-05,
+      "loss": 0.7319,
+      "step": 2045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9438821120627505e-05,
+      "loss": 0.6891,
+      "step": 2046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9438244369466793e-05,
+      "loss": 0.716,
+      "step": 2047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9437667618306084e-05,
+      "loss": 0.5987,
+      "step": 2048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9437090867145372e-05,
+      "loss": 0.6042,
+      "step": 2049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943651411598466e-05,
+      "loss": 0.6916,
+      "step": 2050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.943593736482395e-05,
+      "loss": 0.7244,
+      "step": 2051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9435360613663238e-05,
+      "loss": 0.6018,
+      "step": 2052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9434783862502523e-05,
+      "loss": 0.6453,
+      "step": 2053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9434207111341814e-05,
+      "loss": 0.7096,
+      "step": 2054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.94336303601811e-05,
+      "loss": 0.6867,
+      "step": 2055
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.943305360902039e-05,
+      "loss": 0.6506,
+      "step": 2056
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.943247685785968e-05,
+      "loss": 0.6646,
+      "step": 2057
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9431900106698968e-05,
+      "loss": 0.6253,
+      "step": 2058
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9431323355538255e-05,
+      "loss": 0.8098,
+      "step": 2059
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9430746604377543e-05,
+      "loss": 0.7387,
+      "step": 2060
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.943016985321683e-05,
+      "loss": 0.8519,
+      "step": 2061
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942959310205612e-05,
+      "loss": 0.636,
+      "step": 2062
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942901635089541e-05,
+      "loss": 0.6775,
+      "step": 2063
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9428439599734697e-05,
+      "loss": 0.6351,
+      "step": 2064
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9427862848573985e-05,
+      "loss": 0.7558,
+      "step": 2065
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9427286097413273e-05,
+      "loss": 0.677,
+      "step": 2066
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942670934625256e-05,
+      "loss": 0.6098,
+      "step": 2067
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9426132595091848e-05,
+      "loss": 0.5472,
+      "step": 2068
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9425555843931136e-05,
+      "loss": 0.6949,
+      "step": 2069
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9424979092770427e-05,
+      "loss": 0.7038,
+      "step": 2070
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9424402341609714e-05,
+      "loss": 0.6812,
+      "step": 2071
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9423825590449002e-05,
+      "loss": 0.6126,
+      "step": 2072
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942324883928829e-05,
+      "loss": 0.6708,
+      "step": 2073
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9422672088127578e-05,
+      "loss": 0.5934,
+      "step": 2074
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9422095336966865e-05,
+      "loss": 0.7018,
+      "step": 2075
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9421518585806156e-05,
+      "loss": 0.7246,
+      "step": 2076
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9420941834645444e-05,
+      "loss": 0.6165,
+      "step": 2077
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.942036508348473e-05,
+      "loss": 0.6469,
+      "step": 2078
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9419788332324023e-05,
+      "loss": 0.6347,
+      "step": 2079
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9419211581163307e-05,
+      "loss": 0.6442,
+      "step": 2080
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9418634830002595e-05,
+      "loss": 0.6137,
+      "step": 2081
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9418058078841886e-05,
+      "loss": 0.7187,
+      "step": 2082
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9417481327681173e-05,
+      "loss": 0.836,
+      "step": 2083
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.941690457652046e-05,
+      "loss": 0.711,
+      "step": 2084
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9416327825359752e-05,
+      "loss": 0.7337,
+      "step": 2085
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.941575107419904e-05,
+      "loss": 0.6761,
+      "step": 2086
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9415174323038328e-05,
+      "loss": 0.7365,
+      "step": 2087
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9414597571877615e-05,
+      "loss": 0.6742,
+      "step": 2088
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9414020820716903e-05,
+      "loss": 0.6904,
+      "step": 2089
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.941344406955619e-05,
+      "loss": 0.6717,
+      "step": 2090
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9412867318395482e-05,
+      "loss": 0.6798,
+      "step": 2091
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.941229056723477e-05,
+      "loss": 0.7239,
+      "step": 2092
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9411713816074057e-05,
+      "loss": 0.5963,
+      "step": 2093
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9411137064913345e-05,
+      "loss": 0.6133,
+      "step": 2094
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9410560313752632e-05,
+      "loss": 0.7073,
+      "step": 2095
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940998356259192e-05,
+      "loss": 0.6793,
+      "step": 2096
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940940681143121e-05,
+      "loss": 0.6891,
+      "step": 2097
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.94088300602705e-05,
+      "loss": 0.6571,
+      "step": 2098
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9408253309109787e-05,
+      "loss": 0.6985,
+      "step": 2099
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9407676557949074e-05,
+      "loss": 0.6963,
+      "step": 2100
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9407099806788362e-05,
+      "loss": 0.6909,
+      "step": 2101
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940652305562765e-05,
+      "loss": 0.7397,
+      "step": 2102
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9405946304466937e-05,
+      "loss": 0.6182,
+      "step": 2103
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940536955330623e-05,
+      "loss": 0.6367,
+      "step": 2104
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9404792802145516e-05,
+      "loss": 0.6667,
+      "step": 2105
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9404216050984804e-05,
+      "loss": 0.7506,
+      "step": 2106
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940363929982409e-05,
+      "loss": 0.8152,
+      "step": 2107
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.940306254866338e-05,
+      "loss": 0.6874,
+      "step": 2108
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9402485797502667e-05,
+      "loss": 0.7324,
+      "step": 2109
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9401909046341958e-05,
+      "loss": 0.7161,
+      "step": 2110
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9401332295181246e-05,
+      "loss": 0.8413,
+      "step": 2111
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9400755544020533e-05,
+      "loss": 0.653,
+      "step": 2112
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9400178792859824e-05,
+      "loss": 0.6551,
+      "step": 2113
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9399602041699112e-05,
+      "loss": 0.8044,
+      "step": 2114
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9399025290538396e-05,
+      "loss": 0.6899,
+      "step": 2115
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9398448539377687e-05,
+      "loss": 0.8205,
+      "step": 2116
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9397871788216975e-05,
+      "loss": 0.6007,
+      "step": 2117
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9397295037056263e-05,
+      "loss": 0.8037,
+      "step": 2118
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9396718285895554e-05,
+      "loss": 0.7429,
+      "step": 2119
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.939614153473484e-05,
+      "loss": 0.8034,
+      "step": 2120
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.939556478357413e-05,
+      "loss": 0.5863,
+      "step": 2121
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9394988032413417e-05,
+      "loss": 0.664,
+      "step": 2122
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9394411281252705e-05,
+      "loss": 0.7195,
+      "step": 2123
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9393834530091992e-05,
+      "loss": 0.6916,
+      "step": 2124
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9393257778931283e-05,
+      "loss": 0.6194,
+      "step": 2125
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.939268102777057e-05,
+      "loss": 0.6875,
+      "step": 2126
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.939210427660986e-05,
+      "loss": 0.6692,
+      "step": 2127
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9391527525449146e-05,
+      "loss": 0.5508,
+      "step": 2128
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9390950774288434e-05,
+      "loss": 0.6768,
+      "step": 2129
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9390374023127722e-05,
+      "loss": 0.6662,
+      "step": 2130
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9389797271967013e-05,
+      "loss": 0.554,
+      "step": 2131
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93892205208063e-05,
+      "loss": 0.6408,
+      "step": 2132
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9388643769645588e-05,
+      "loss": 0.7338,
+      "step": 2133
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9388067018484876e-05,
+      "loss": 0.7115,
+      "step": 2134
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9387490267324164e-05,
+      "loss": 0.7518,
+      "step": 2135
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938691351616345e-05,
+      "loss": 0.658,
+      "step": 2136
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938633676500274e-05,
+      "loss": 0.7686,
+      "step": 2137
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938576001384203e-05,
+      "loss": 0.6793,
+      "step": 2138
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9385183262681318e-05,
+      "loss": 0.6837,
+      "step": 2139
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9384606511520605e-05,
+      "loss": 0.7236,
+      "step": 2140
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9384029760359896e-05,
+      "loss": 0.74,
+      "step": 2141
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938345300919918e-05,
+      "loss": 0.7266,
+      "step": 2142
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938287625803847e-05,
+      "loss": 0.6395,
+      "step": 2143
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.938229950687776e-05,
+      "loss": 0.7045,
+      "step": 2144
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9381722755717047e-05,
+      "loss": 0.5635,
+      "step": 2145
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9381146004556335e-05,
+      "loss": 0.682,
+      "step": 2146
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9380569253395626e-05,
+      "loss": 0.7194,
+      "step": 2147
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9379992502234914e-05,
+      "loss": 0.5388,
+      "step": 2148
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93794157510742e-05,
+      "loss": 0.7488,
+      "step": 2149
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.937883899991349e-05,
+      "loss": 0.6724,
+      "step": 2150
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9378262248752777e-05,
+      "loss": 0.6603,
+      "step": 2151
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9377685497592064e-05,
+      "loss": 0.708,
+      "step": 2152
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9377108746431355e-05,
+      "loss": 0.7188,
+      "step": 2153
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9376531995270643e-05,
+      "loss": 0.6531,
+      "step": 2154
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.937595524410993e-05,
+      "loss": 0.7431,
+      "step": 2155
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.937537849294922e-05,
+      "loss": 0.7478,
+      "step": 2156
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9374801741788506e-05,
+      "loss": 0.7392,
+      "step": 2157
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9374224990627794e-05,
+      "loss": 0.7369,
+      "step": 2158
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9373648239467085e-05,
+      "loss": 0.7683,
+      "step": 2159
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9373071488306373e-05,
+      "loss": 0.7288,
+      "step": 2160
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.937249473714566e-05,
+      "loss": 0.5625,
+      "step": 2161
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9371917985984948e-05,
+      "loss": 0.7188,
+      "step": 2162
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9371341234824236e-05,
+      "loss": 0.6723,
+      "step": 2163
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9370764483663523e-05,
+      "loss": 0.6464,
+      "step": 2164
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9370187732502814e-05,
+      "loss": 0.6789,
+      "step": 2165
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9369610981342102e-05,
+      "loss": 0.6599,
+      "step": 2166
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936903423018139e-05,
+      "loss": 0.6839,
+      "step": 2167
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9368457479020678e-05,
+      "loss": 0.7558,
+      "step": 2168
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9367880727859965e-05,
+      "loss": 0.654,
+      "step": 2169
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9367303976699253e-05,
+      "loss": 0.6996,
+      "step": 2170
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936672722553854e-05,
+      "loss": 0.6095,
+      "step": 2171
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936615047437783e-05,
+      "loss": 0.7862,
+      "step": 2172
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936557372321712e-05,
+      "loss": 0.6052,
+      "step": 2173
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9364996972056407e-05,
+      "loss": 0.63,
+      "step": 2174
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9364420220895698e-05,
+      "loss": 0.6901,
+      "step": 2175
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9363843469734986e-05,
+      "loss": 0.6819,
+      "step": 2176
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9363266718574273e-05,
+      "loss": 0.5424,
+      "step": 2177
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936268996741356e-05,
+      "loss": 0.6127,
+      "step": 2178
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.936211321625285e-05,
+      "loss": 0.5833,
+      "step": 2179
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9361536465092137e-05,
+      "loss": 0.6915,
+      "step": 2180
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9360959713931428e-05,
+      "loss": 0.5619,
+      "step": 2181
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9360382962770715e-05,
+      "loss": 0.6617,
+      "step": 2182
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9359806211610003e-05,
+      "loss": 0.6567,
+      "step": 2183
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935922946044929e-05,
+      "loss": 0.6389,
+      "step": 2184
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935865270928858e-05,
+      "loss": 0.5973,
+      "step": 2185
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9358075958127866e-05,
+      "loss": 0.6756,
+      "step": 2186
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9357499206967157e-05,
+      "loss": 0.7006,
+      "step": 2187
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9356922455806445e-05,
+      "loss": 0.6789,
+      "step": 2188
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9356345704645732e-05,
+      "loss": 0.5619,
+      "step": 2189
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935576895348502e-05,
+      "loss": 0.4746,
+      "step": 2190
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9355192202324308e-05,
+      "loss": 0.6247,
+      "step": 2191
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9354615451163596e-05,
+      "loss": 0.7712,
+      "step": 2192
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9354038700002887e-05,
+      "loss": 0.5653,
+      "step": 2193
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9353461948842174e-05,
+      "loss": 0.6458,
+      "step": 2194
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9352885197681462e-05,
+      "loss": 0.7428,
+      "step": 2195
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.935230844652075e-05,
+      "loss": 0.6371,
+      "step": 2196
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9351731695360037e-05,
+      "loss": 0.809,
+      "step": 2197
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9351154944199325e-05,
+      "loss": 0.646,
+      "step": 2198
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9350578193038616e-05,
+      "loss": 0.6827,
+      "step": 2199
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9350001441877904e-05,
+      "loss": 0.8092,
+      "step": 2200
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934942469071719e-05,
+      "loss": 0.7465,
+      "step": 2201
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9348847939556483e-05,
+      "loss": 0.6797,
+      "step": 2202
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934827118839577e-05,
+      "loss": 0.695,
+      "step": 2203
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9347694437235055e-05,
+      "loss": 0.5443,
+      "step": 2204
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9347117686074342e-05,
+      "loss": 0.5573,
+      "step": 2205
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9346540934913633e-05,
+      "loss": 0.6044,
+      "step": 2206
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934596418375292e-05,
+      "loss": 0.7761,
+      "step": 2207
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934538743259221e-05,
+      "loss": 0.6321,
+      "step": 2208
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.93448106814315e-05,
+      "loss": 0.7148,
+      "step": 2209
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9344233930270787e-05,
+      "loss": 0.6511,
+      "step": 2210
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9343657179110075e-05,
+      "loss": 0.7025,
+      "step": 2211
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9343080427949363e-05,
+      "loss": 0.7775,
+      "step": 2212
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934250367678865e-05,
+      "loss": 0.5702,
+      "step": 2213
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9341926925627938e-05,
+      "loss": 0.7851,
+      "step": 2214
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.934135017446723e-05,
+      "loss": 0.7808,
+      "step": 2215
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9340773423306517e-05,
+      "loss": 0.8001,
+      "step": 2216
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9340196672145805e-05,
+      "loss": 0.6875,
+      "step": 2217
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9339619920985092e-05,
+      "loss": 0.7373,
+      "step": 2218
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933904316982438e-05,
+      "loss": 0.7639,
+      "step": 2219
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9338466418663668e-05,
+      "loss": 0.6528,
+      "step": 2220
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933788966750296e-05,
+      "loss": 0.637,
+      "step": 2221
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9337312916342246e-05,
+      "loss": 0.6933,
+      "step": 2222
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9336736165181534e-05,
+      "loss": 0.7224,
+      "step": 2223
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9336159414020822e-05,
+      "loss": 0.5589,
+      "step": 2224
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933558266286011e-05,
+      "loss": 0.6991,
+      "step": 2225
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9335005911699397e-05,
+      "loss": 0.5877,
+      "step": 2226
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9334429160538688e-05,
+      "loss": 0.6665,
+      "step": 2227
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9333852409377976e-05,
+      "loss": 0.713,
+      "step": 2228
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9333275658217264e-05,
+      "loss": 0.6029,
+      "step": 2229
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933269890705655e-05,
+      "loss": 0.6874,
+      "step": 2230
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.933212215589584e-05,
+      "loss": 0.7924,
+      "step": 2231
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9331545404735127e-05,
+      "loss": 0.6219,
+      "step": 2232
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9330968653574418e-05,
+      "loss": 0.6722,
+      "step": 2233
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9330391902413705e-05,
+      "loss": 0.7263,
+      "step": 2234
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9329815151252993e-05,
+      "loss": 0.6118,
+      "step": 2235
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9329238400092284e-05,
+      "loss": 0.6507,
+      "step": 2236
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9328661648931572e-05,
+      "loss": 0.662,
+      "step": 2237
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932808489777086e-05,
+      "loss": 0.8441,
+      "step": 2238
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9327508146610147e-05,
+      "loss": 0.6236,
+      "step": 2239
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9326931395449435e-05,
+      "loss": 0.6268,
+      "step": 2240
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9326354644288723e-05,
+      "loss": 0.6686,
+      "step": 2241
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932577789312801e-05,
+      "loss": 0.6042,
+      "step": 2242
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.93252011419673e-05,
+      "loss": 0.5994,
+      "step": 2243
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932462439080659e-05,
+      "loss": 0.6729,
+      "step": 2244
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9324047639645877e-05,
+      "loss": 0.7266,
+      "step": 2245
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9323470888485164e-05,
+      "loss": 0.6409,
+      "step": 2246
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9322894137324452e-05,
+      "loss": 0.6187,
+      "step": 2247
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932231738616374e-05,
+      "loss": 0.5054,
+      "step": 2248
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932174063500303e-05,
+      "loss": 0.5806,
+      "step": 2249
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.932116388384232e-05,
+      "loss": 0.6917,
+      "step": 2250
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9320587132681606e-05,
+      "loss": 0.7467,
+      "step": 2251
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9320010381520894e-05,
+      "loss": 0.6625,
+      "step": 2252
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931943363036018e-05,
+      "loss": 0.6599,
+      "step": 2253
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931885687919947e-05,
+      "loss": 0.6361,
+      "step": 2254
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931828012803876e-05,
+      "loss": 0.6769,
+      "step": 2255
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9317703376878048e-05,
+      "loss": 0.768,
+      "step": 2256
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9317126625717336e-05,
+      "loss": 0.6522,
+      "step": 2257
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9316549874556623e-05,
+      "loss": 0.7123,
+      "step": 2258
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931597312339591e-05,
+      "loss": 0.6751,
+      "step": 2259
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.93153963722352e-05,
+      "loss": 0.6576,
+      "step": 2260
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931481962107449e-05,
+      "loss": 0.6873,
+      "step": 2261
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9314242869913778e-05,
+      "loss": 0.7056,
+      "step": 2262
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9313666118753065e-05,
+      "loss": 0.6725,
+      "step": 2263
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9313089367592356e-05,
+      "loss": 0.6863,
+      "step": 2264
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9312512616431644e-05,
+      "loss": 0.6526,
+      "step": 2265
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9311935865270928e-05,
+      "loss": 0.7517,
+      "step": 2266
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.931135911411022e-05,
+      "loss": 0.6656,
+      "step": 2267
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9310782362949507e-05,
+      "loss": 0.6733,
+      "step": 2268
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9310205611788795e-05,
+      "loss": 0.7514,
+      "step": 2269
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9309628860628086e-05,
+      "loss": 0.6241,
+      "step": 2270
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9309052109467373e-05,
+      "loss": 0.5476,
+      "step": 2271
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930847535830666e-05,
+      "loss": 0.6222,
+      "step": 2272
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930789860714595e-05,
+      "loss": 0.7051,
+      "step": 2273
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9307321855985237e-05,
+      "loss": 0.6182,
+      "step": 2274
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9306745104824524e-05,
+      "loss": 0.6274,
+      "step": 2275
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9306168353663812e-05,
+      "loss": 0.6814,
+      "step": 2276
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9305591602503103e-05,
+      "loss": 0.7458,
+      "step": 2277
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930501485134239e-05,
+      "loss": 0.6209,
+      "step": 2278
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930443810018168e-05,
+      "loss": 0.6835,
+      "step": 2279
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9303861349020966e-05,
+      "loss": 0.7593,
+      "step": 2280
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9303284597860254e-05,
+      "loss": 0.7349,
+      "step": 2281
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930270784669954e-05,
+      "loss": 0.648,
+      "step": 2282
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9302131095538832e-05,
+      "loss": 0.6845,
+      "step": 2283
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.930155434437812e-05,
+      "loss": 0.6378,
+      "step": 2284
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9300977593217408e-05,
+      "loss": 0.6351,
+      "step": 2285
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9300400842056696e-05,
+      "loss": 0.5949,
+      "step": 2286
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9299824090895983e-05,
+      "loss": 0.6431,
+      "step": 2287
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929924733973527e-05,
+      "loss": 0.7314,
+      "step": 2288
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9298670588574562e-05,
+      "loss": 0.6375,
+      "step": 2289
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929809383741385e-05,
+      "loss": 0.6872,
+      "step": 2290
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9297517086253137e-05,
+      "loss": 0.5285,
+      "step": 2291
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929694033509243e-05,
+      "loss": 0.6438,
+      "step": 2292
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9296363583931713e-05,
+      "loss": 0.7042,
+      "step": 2293
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9295786832771e-05,
+      "loss": 0.6389,
+      "step": 2294
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929521008161029e-05,
+      "loss": 0.7057,
+      "step": 2295
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929463333044958e-05,
+      "loss": 0.7091,
+      "step": 2296
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9294056579288867e-05,
+      "loss": 0.703,
+      "step": 2297
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9293479828128158e-05,
+      "loss": 0.8281,
+      "step": 2298
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9292903076967446e-05,
+      "loss": 0.7209,
+      "step": 2299
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9292326325806733e-05,
+      "loss": 0.7058,
+      "step": 2300
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929174957464602e-05,
+      "loss": 0.4239,
+      "step": 2301
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.929117282348531e-05,
+      "loss": 0.7083,
+      "step": 2302
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9290596072324596e-05,
+      "loss": 0.5927,
+      "step": 2303
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9290019321163887e-05,
+      "loss": 0.6348,
+      "step": 2304
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9289442570003175e-05,
+      "loss": 0.5826,
+      "step": 2305
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9288865818842463e-05,
+      "loss": 0.7481,
+      "step": 2306
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.928828906768175e-05,
+      "loss": 0.7213,
+      "step": 2307
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9287712316521038e-05,
+      "loss": 0.8046,
+      "step": 2308
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9287135565360326e-05,
+      "loss": 0.5263,
+      "step": 2309
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9286558814199613e-05,
+      "loss": 0.7133,
+      "step": 2310
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9285982063038905e-05,
+      "loss": 0.6407,
+      "step": 2311
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9285405311878192e-05,
+      "loss": 0.6739,
+      "step": 2312
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.928482856071748e-05,
+      "loss": 0.6031,
+      "step": 2313
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9284251809556768e-05,
+      "loss": 0.5764,
+      "step": 2314
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9283675058396055e-05,
+      "loss": 0.7426,
+      "step": 2315
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9283098307235343e-05,
+      "loss": 0.6199,
+      "step": 2316
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9282521556074634e-05,
+      "loss": 0.5413,
+      "step": 2317
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9281944804913922e-05,
+      "loss": 0.7256,
+      "step": 2318
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.928136805375321e-05,
+      "loss": 0.5157,
+      "step": 2319
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9280791302592497e-05,
+      "loss": 0.697,
+      "step": 2320
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9280214551431785e-05,
+      "loss": 0.6673,
+      "step": 2321
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9279637800271072e-05,
+      "loss": 0.6082,
+      "step": 2322
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9279061049110364e-05,
+      "loss": 0.5753,
+      "step": 2323
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927848429794965e-05,
+      "loss": 0.8167,
+      "step": 2324
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927790754678894e-05,
+      "loss": 0.5612,
+      "step": 2325
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927733079562823e-05,
+      "loss": 0.6621,
+      "step": 2326
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9276754044467518e-05,
+      "loss": 0.6374,
+      "step": 2327
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9276177293306802e-05,
+      "loss": 0.7662,
+      "step": 2328
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9275600542146093e-05,
+      "loss": 0.6896,
+      "step": 2329
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927502379098538e-05,
+      "loss": 0.8489,
+      "step": 2330
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927444703982467e-05,
+      "loss": 0.6759,
+      "step": 2331
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927387028866396e-05,
+      "loss": 0.4549,
+      "step": 2332
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9273293537503247e-05,
+      "loss": 0.6959,
+      "step": 2333
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9272716786342535e-05,
+      "loss": 0.6845,
+      "step": 2334
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9272140035181823e-05,
+      "loss": 0.6333,
+      "step": 2335
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927156328402111e-05,
+      "loss": 0.7966,
+      "step": 2336
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9270986532860398e-05,
+      "loss": 0.6378,
+      "step": 2337
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.927040978169969e-05,
+      "loss": 0.7368,
+      "step": 2338
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9269833030538977e-05,
+      "loss": 0.6342,
+      "step": 2339
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9269256279378264e-05,
+      "loss": 0.7298,
+      "step": 2340
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9268679528217552e-05,
+      "loss": 0.5951,
+      "step": 2341
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.926810277705684e-05,
+      "loss": 0.7836,
+      "step": 2342
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9267526025896127e-05,
+      "loss": 0.6979,
+      "step": 2343
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9266949274735415e-05,
+      "loss": 0.6675,
+      "step": 2344
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9266372523574706e-05,
+      "loss": 0.5708,
+      "step": 2345
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9265795772413994e-05,
+      "loss": 0.5873,
+      "step": 2346
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.926521902125328e-05,
+      "loss": 0.7024,
+      "step": 2347
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.926464227009257e-05,
+      "loss": 0.6899,
+      "step": 2348
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9264065518931857e-05,
+      "loss": 0.5698,
+      "step": 2349
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9263488767771145e-05,
+      "loss": 0.6321,
+      "step": 2350
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9262912016610436e-05,
+      "loss": 0.7549,
+      "step": 2351
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9262335265449723e-05,
+      "loss": 0.7457,
+      "step": 2352
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.926175851428901e-05,
+      "loss": 0.598,
+      "step": 2353
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9261181763128302e-05,
+      "loss": 0.6193,
+      "step": 2354
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9260605011967586e-05,
+      "loss": 0.6114,
+      "step": 2355
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9260028260806874e-05,
+      "loss": 0.6989,
+      "step": 2356
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9259451509646165e-05,
+      "loss": 0.6151,
+      "step": 2357
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9258874758485453e-05,
+      "loss": 0.7451,
+      "step": 2358
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925829800732474e-05,
+      "loss": 0.6906,
+      "step": 2359
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925772125616403e-05,
+      "loss": 0.549,
+      "step": 2360
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925714450500332e-05,
+      "loss": 0.6477,
+      "step": 2361
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9256567753842607e-05,
+      "loss": 0.7806,
+      "step": 2362
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9255991002681895e-05,
+      "loss": 0.7068,
+      "step": 2363
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9255414251521182e-05,
+      "loss": 0.6566,
+      "step": 2364
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925483750036047e-05,
+      "loss": 0.6324,
+      "step": 2365
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925426074919976e-05,
+      "loss": 0.5937,
+      "step": 2366
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925368399803905e-05,
+      "loss": 0.7338,
+      "step": 2367
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9253107246878337e-05,
+      "loss": 0.6828,
+      "step": 2368
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9252530495717624e-05,
+      "loss": 0.5275,
+      "step": 2369
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9251953744556912e-05,
+      "loss": 0.654,
+      "step": 2370
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.92513769933962e-05,
+      "loss": 0.6614,
+      "step": 2371
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925080024223549e-05,
+      "loss": 0.5563,
+      "step": 2372
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.925022349107478e-05,
+      "loss": 0.7331,
+      "step": 2373
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9249646739914066e-05,
+      "loss": 0.6471,
+      "step": 2374
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9249069988753354e-05,
+      "loss": 0.5977,
+      "step": 2375
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924849323759264e-05,
+      "loss": 0.6343,
+      "step": 2376
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924791648643193e-05,
+      "loss": 0.6235,
+      "step": 2377
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924733973527122e-05,
+      "loss": 0.6324,
+      "step": 2378
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9246762984110508e-05,
+      "loss": 0.6531,
+      "step": 2379
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9246186232949796e-05,
+      "loss": 0.6984,
+      "step": 2380
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9245609481789083e-05,
+      "loss": 0.8243,
+      "step": 2381
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924503273062837e-05,
+      "loss": 0.6175,
+      "step": 2382
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924445597946766e-05,
+      "loss": 0.5798,
+      "step": 2383
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9243879228306946e-05,
+      "loss": 0.5752,
+      "step": 2384
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9243302477146237e-05,
+      "loss": 0.6959,
+      "step": 2385
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9242725725985525e-05,
+      "loss": 0.707,
+      "step": 2386
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9242148974824813e-05,
+      "loss": 0.7457,
+      "step": 2387
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9241572223664104e-05,
+      "loss": 0.4878,
+      "step": 2388
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.924099547250339e-05,
+      "loss": 0.6046,
+      "step": 2389
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9240418721342676e-05,
+      "loss": 0.6084,
+      "step": 2390
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9239841970181967e-05,
+      "loss": 0.6969,
+      "step": 2391
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9239265219021254e-05,
+      "loss": 0.6963,
+      "step": 2392
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9238688467860542e-05,
+      "loss": 0.6953,
+      "step": 2393
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9238111716699833e-05,
+      "loss": 0.6141,
+      "step": 2394
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923753496553912e-05,
+      "loss": 0.7002,
+      "step": 2395
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923695821437841e-05,
+      "loss": 0.6204,
+      "step": 2396
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9236381463217696e-05,
+      "loss": 0.5932,
+      "step": 2397
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9235804712056984e-05,
+      "loss": 0.6007,
+      "step": 2398
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923522796089627e-05,
+      "loss": 0.5939,
+      "step": 2399
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9234651209735563e-05,
+      "loss": 0.6788,
+      "step": 2400
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923407445857485e-05,
+      "loss": 0.6912,
+      "step": 2401
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9233497707414138e-05,
+      "loss": 0.7072,
+      "step": 2402
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9232920956253426e-05,
+      "loss": 0.7983,
+      "step": 2403
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9232344205092713e-05,
+      "loss": 0.6494,
+      "step": 2404
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9231767453932e-05,
+      "loss": 0.6191,
+      "step": 2405
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9231190702771292e-05,
+      "loss": 0.6918,
+      "step": 2406
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.923061395161058e-05,
+      "loss": 0.7169,
+      "step": 2407
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9230037200449868e-05,
+      "loss": 0.7955,
+      "step": 2408
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9229460449289155e-05,
+      "loss": 0.6828,
+      "step": 2409
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9228883698128443e-05,
+      "loss": 0.6528,
+      "step": 2410
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922830694696773e-05,
+      "loss": 0.5764,
+      "step": 2411
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9227730195807022e-05,
+      "loss": 0.6505,
+      "step": 2412
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.922715344464631e-05,
+      "loss": 0.7849,
+      "step": 2413
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9226576693485597e-05,
+      "loss": 0.6878,
+      "step": 2414
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9225999942324885e-05,
+      "loss": 0.7788,
+      "step": 2415
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9225423191164176e-05,
+      "loss": 0.5785,
+      "step": 2416
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.922484644000346e-05,
+      "loss": 0.768,
+      "step": 2417
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9224269688842748e-05,
+      "loss": 0.6477,
+      "step": 2418
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.922369293768204e-05,
+      "loss": 0.6975,
+      "step": 2419
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9223116186521327e-05,
+      "loss": 0.6897,
+      "step": 2420
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9222539435360614e-05,
+      "loss": 0.6882,
+      "step": 2421
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9221962684199905e-05,
+      "loss": 0.8733,
+      "step": 2422
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9221385933039193e-05,
+      "loss": 0.6094,
+      "step": 2423
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.922080918187848e-05,
+      "loss": 0.6203,
+      "step": 2424
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.922023243071777e-05,
+      "loss": 0.6452,
+      "step": 2425
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9219655679557056e-05,
+      "loss": 0.6543,
+      "step": 2426
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9219078928396344e-05,
+      "loss": 0.8034,
+      "step": 2427
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9218502177235635e-05,
+      "loss": 0.6046,
+      "step": 2428
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9217925426074923e-05,
+      "loss": 0.5667,
+      "step": 2429
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.921734867491421e-05,
+      "loss": 0.7102,
+      "step": 2430
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9216771923753498e-05,
+      "loss": 0.6596,
+      "step": 2431
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9216195172592786e-05,
+      "loss": 0.6153,
+      "step": 2432
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9215618421432073e-05,
+      "loss": 0.7097,
+      "step": 2433
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9215041670271364e-05,
+      "loss": 0.5525,
+      "step": 2434
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9214464919110652e-05,
+      "loss": 0.6497,
+      "step": 2435
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.921388816794994e-05,
+      "loss": 0.6921,
+      "step": 2436
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9213311416789227e-05,
+      "loss": 0.6827,
+      "step": 2437
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9212734665628515e-05,
+      "loss": 0.6452,
+      "step": 2438
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9212157914467803e-05,
+      "loss": 0.6798,
+      "step": 2439
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9211581163307094e-05,
+      "loss": 0.5619,
+      "step": 2440
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.921100441214638e-05,
+      "loss": 0.7522,
+      "step": 2441
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.921042766098567e-05,
+      "loss": 0.6406,
+      "step": 2442
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9209850909824957e-05,
+      "loss": 0.6941,
+      "step": 2443
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9209274158664245e-05,
+      "loss": 0.5856,
+      "step": 2444
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9208697407503532e-05,
+      "loss": 0.6434,
+      "step": 2445
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9208120656342823e-05,
+      "loss": 0.491,
+      "step": 2446
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.920754390518211e-05,
+      "loss": 0.5676,
+      "step": 2447
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.92069671540214e-05,
+      "loss": 0.5654,
+      "step": 2448
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9206390402860686e-05,
+      "loss": 0.6842,
+      "step": 2449
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9205813651699978e-05,
+      "loss": 0.732,
+      "step": 2450
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9205236900539265e-05,
+      "loss": 0.6991,
+      "step": 2451
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.920466014937855e-05,
+      "loss": 0.5729,
+      "step": 2452
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.920408339821784e-05,
+      "loss": 0.6124,
+      "step": 2453
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9203506647057128e-05,
+      "loss": 0.6675,
+      "step": 2454
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9202929895896416e-05,
+      "loss": 0.6853,
+      "step": 2455
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9202353144735707e-05,
+      "loss": 0.5924,
+      "step": 2456
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9201776393574995e-05,
+      "loss": 0.7548,
+      "step": 2457
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9201199642414282e-05,
+      "loss": 0.5566,
+      "step": 2458
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.920062289125357e-05,
+      "loss": 0.7671,
+      "step": 2459
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9200046140092858e-05,
+      "loss": 0.7471,
+      "step": 2460
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9199469388932145e-05,
+      "loss": 0.6431,
+      "step": 2461
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9198892637771437e-05,
+      "loss": 0.5084,
+      "step": 2462
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9198315886610724e-05,
+      "loss": 0.7277,
+      "step": 2463
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9197739135450012e-05,
+      "loss": 0.5733,
+      "step": 2464
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91971623842893e-05,
+      "loss": 0.562,
+      "step": 2465
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9196585633128587e-05,
+      "loss": 0.5983,
+      "step": 2466
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9196008881967875e-05,
+      "loss": 0.5727,
+      "step": 2467
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9195432130807166e-05,
+      "loss": 0.6537,
+      "step": 2468
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9194855379646454e-05,
+      "loss": 0.6164,
+      "step": 2469
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.919427862848574e-05,
+      "loss": 0.5712,
+      "step": 2470
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.919370187732503e-05,
+      "loss": 0.4514,
+      "step": 2471
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9193125126164317e-05,
+      "loss": 0.5431,
+      "step": 2472
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9192548375003604e-05,
+      "loss": 0.72,
+      "step": 2473
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9191971623842895e-05,
+      "loss": 0.526,
+      "step": 2474
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9191394872682183e-05,
+      "loss": 0.747,
+      "step": 2475
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.919081812152147e-05,
+      "loss": 0.5957,
+      "step": 2476
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9190241370360762e-05,
+      "loss": 0.7761,
+      "step": 2477
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918966461920005e-05,
+      "loss": 0.6751,
+      "step": 2478
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9189087868039334e-05,
+      "loss": 0.6466,
+      "step": 2479
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9188511116878625e-05,
+      "loss": 0.6289,
+      "step": 2480
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9187934365717913e-05,
+      "loss": 0.4952,
+      "step": 2481
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91873576145572e-05,
+      "loss": 0.5847,
+      "step": 2482
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918678086339649e-05,
+      "loss": 0.653,
+      "step": 2483
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918620411223578e-05,
+      "loss": 0.641,
+      "step": 2484
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9185627361075067e-05,
+      "loss": 0.6184,
+      "step": 2485
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9185050609914354e-05,
+      "loss": 0.6959,
+      "step": 2486
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9184473858753642e-05,
+      "loss": 0.6154,
+      "step": 2487
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918389710759293e-05,
+      "loss": 0.7221,
+      "step": 2488
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9183320356432218e-05,
+      "loss": 0.7139,
+      "step": 2489
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918274360527151e-05,
+      "loss": 0.8201,
+      "step": 2490
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9182166854110796e-05,
+      "loss": 0.6415,
+      "step": 2491
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9181590102950084e-05,
+      "loss": 0.5339,
+      "step": 2492
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918101335178937e-05,
+      "loss": 0.6192,
+      "step": 2493
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.918043660062866e-05,
+      "loss": 0.5947,
+      "step": 2494
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9179859849467947e-05,
+      "loss": 0.5863,
+      "step": 2495
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9179283098307238e-05,
+      "loss": 0.613,
+      "step": 2496
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9178706347146526e-05,
+      "loss": 0.6816,
+      "step": 2497
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9178129595985813e-05,
+      "loss": 0.7222,
+      "step": 2498
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91775528448251e-05,
+      "loss": 0.6829,
+      "step": 2499
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.917697609366439e-05,
+      "loss": 0.5951,
+      "step": 2500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9176399342503677e-05,
+      "loss": 0.588,
+      "step": 2501
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9175822591342968e-05,
+      "loss": 0.6463,
+      "step": 2502
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9175245840182255e-05,
+      "loss": 0.7411,
+      "step": 2503
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9174669089021543e-05,
+      "loss": 0.5087,
+      "step": 2504
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.917409233786083e-05,
+      "loss": 0.711,
+      "step": 2505
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.917351558670012e-05,
+      "loss": 0.6717,
+      "step": 2506
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9172938835539406e-05,
+      "loss": 0.6595,
+      "step": 2507
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9172362084378697e-05,
+      "loss": 0.6069,
+      "step": 2508
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9171785333217985e-05,
+      "loss": 0.6732,
+      "step": 2509
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9171208582057272e-05,
+      "loss": 0.7112,
+      "step": 2510
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9170631830896564e-05,
+      "loss": 0.62,
+      "step": 2511
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.917005507973585e-05,
+      "loss": 0.7504,
+      "step": 2512
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916947832857514e-05,
+      "loss": 0.6249,
+      "step": 2513
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9168901577414427e-05,
+      "loss": 0.7647,
+      "step": 2514
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9168324826253714e-05,
+      "loss": 0.6128,
+      "step": 2515
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9167748075093002e-05,
+      "loss": 0.6515,
+      "step": 2516
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9167171323932293e-05,
+      "loss": 0.7176,
+      "step": 2517
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916659457277158e-05,
+      "loss": 0.5688,
+      "step": 2518
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916601782161087e-05,
+      "loss": 0.6544,
+      "step": 2519
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9165441070450156e-05,
+      "loss": 0.9292,
+      "step": 2520
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9164864319289444e-05,
+      "loss": 0.6008,
+      "step": 2521
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916428756812873e-05,
+      "loss": 0.6117,
+      "step": 2522
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916371081696802e-05,
+      "loss": 0.68,
+      "step": 2523
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916313406580731e-05,
+      "loss": 0.7124,
+      "step": 2524
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9162557314646598e-05,
+      "loss": 0.6745,
+      "step": 2525
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9161980563485886e-05,
+      "loss": 0.6705,
+      "step": 2526
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9161403812325173e-05,
+      "loss": 0.6557,
+      "step": 2527
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916082706116446e-05,
+      "loss": 0.6125,
+      "step": 2528
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.916025031000375e-05,
+      "loss": 0.5982,
+      "step": 2529
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.915967355884304e-05,
+      "loss": 0.683,
+      "step": 2530
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9159096807682327e-05,
+      "loss": 0.7343,
+      "step": 2531
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9158520056521615e-05,
+      "loss": 0.6924,
+      "step": 2532
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9157943305360903e-05,
+      "loss": 0.5431,
+      "step": 2533
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.915736655420019e-05,
+      "loss": 0.6814,
+      "step": 2534
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9156789803039478e-05,
+      "loss": 0.6069,
+      "step": 2535
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.915621305187877e-05,
+      "loss": 0.6928,
+      "step": 2536
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9155636300718057e-05,
+      "loss": 0.7315,
+      "step": 2537
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9155059549557345e-05,
+      "loss": 0.6461,
+      "step": 2538
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9154482798396636e-05,
+      "loss": 0.5791,
+      "step": 2539
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9153906047235923e-05,
+      "loss": 0.6314,
+      "step": 2540
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9153329296075208e-05,
+      "loss": 0.6707,
+      "step": 2541
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91527525449145e-05,
+      "loss": 0.7338,
+      "step": 2542
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9152175793753786e-05,
+      "loss": 0.6483,
+      "step": 2543
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9151599042593074e-05,
+      "loss": 0.578,
+      "step": 2544
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9151022291432365e-05,
+      "loss": 0.6588,
+      "step": 2545
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9150445540271653e-05,
+      "loss": 0.6289,
+      "step": 2546
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.914986878911094e-05,
+      "loss": 0.6667,
+      "step": 2547
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9149292037950228e-05,
+      "loss": 0.5646,
+      "step": 2548
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9148715286789516e-05,
+      "loss": 0.6383,
+      "step": 2549
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9148138535628804e-05,
+      "loss": 0.7068,
+      "step": 2550
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9147561784468095e-05,
+      "loss": 0.6097,
+      "step": 2551
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9146985033307382e-05,
+      "loss": 0.6321,
+      "step": 2552
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.914640828214667e-05,
+      "loss": 0.6917,
+      "step": 2553
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9145831530985958e-05,
+      "loss": 0.5722,
+      "step": 2554
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9145254779825245e-05,
+      "loss": 0.5723,
+      "step": 2555
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9144678028664533e-05,
+      "loss": 0.6823,
+      "step": 2556
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.914410127750382e-05,
+      "loss": 0.618,
+      "step": 2557
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9143524526343112e-05,
+      "loss": 0.6511,
+      "step": 2558
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91429477751824e-05,
+      "loss": 0.8225,
+      "step": 2559
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9142371024021687e-05,
+      "loss": 0.6123,
+      "step": 2560
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9141794272860975e-05,
+      "loss": 0.7256,
+      "step": 2561
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9141217521700263e-05,
+      "loss": 0.5541,
+      "step": 2562
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.914064077053955e-05,
+      "loss": 0.5901,
+      "step": 2563
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.914006401937884e-05,
+      "loss": 0.5796,
+      "step": 2564
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.913948726821813e-05,
+      "loss": 0.5801,
+      "step": 2565
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9138910517057417e-05,
+      "loss": 0.5936,
+      "step": 2566
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9138333765896704e-05,
+      "loss": 0.727,
+      "step": 2567
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9137757014735992e-05,
+      "loss": 0.6718,
+      "step": 2568
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.913718026357528e-05,
+      "loss": 0.638,
+      "step": 2569
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.913660351241457e-05,
+      "loss": 0.6819,
+      "step": 2570
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.913602676125386e-05,
+      "loss": 0.504,
+      "step": 2571
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9135450010093146e-05,
+      "loss": 0.5654,
+      "step": 2572
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9134873258932437e-05,
+      "loss": 0.5936,
+      "step": 2573
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9134296507771725e-05,
+      "loss": 0.5919,
+      "step": 2574
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9133719756611013e-05,
+      "loss": 0.5899,
+      "step": 2575
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.91331430054503e-05,
+      "loss": 0.6855,
+      "step": 2576
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9132566254289588e-05,
+      "loss": 0.6109,
+      "step": 2577
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9131989503128876e-05,
+      "loss": 0.6601,
+      "step": 2578
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9131412751968167e-05,
+      "loss": 0.7405,
+      "step": 2579
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9130836000807454e-05,
+      "loss": 0.6712,
+      "step": 2580
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9130259249646742e-05,
+      "loss": 0.6328,
+      "step": 2581
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.912968249848603e-05,
+      "loss": 0.6235,
+      "step": 2582
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9129105747325318e-05,
+      "loss": 0.705,
+      "step": 2583
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9128528996164605e-05,
+      "loss": 0.6092,
+      "step": 2584
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9127952245003896e-05,
+      "loss": 0.6006,
+      "step": 2585
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9127375493843184e-05,
+      "loss": 0.7746,
+      "step": 2586
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.912679874268247e-05,
+      "loss": 0.8284,
+      "step": 2587
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.912622199152176e-05,
+      "loss": 0.5692,
+      "step": 2588
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9125645240361047e-05,
+      "loss": 0.6755,
+      "step": 2589
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9125068489200335e-05,
+      "loss": 0.6216,
+      "step": 2590
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9124491738039622e-05,
+      "loss": 0.695,
+      "step": 2591
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9123914986878913e-05,
+      "loss": 0.6727,
+      "step": 2592
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.91233382357182e-05,
+      "loss": 0.7183,
+      "step": 2593
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.912276148455749e-05,
+      "loss": 0.8129,
+      "step": 2594
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9122184733396777e-05,
+      "loss": 0.6884,
+      "step": 2595
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9121607982236064e-05,
+      "loss": 0.6102,
+      "step": 2596
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9121031231075352e-05,
+      "loss": 0.5762,
+      "step": 2597
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9120454479914643e-05,
+      "loss": 0.6876,
+      "step": 2598
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911987772875393e-05,
+      "loss": 0.6131,
+      "step": 2599
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911930097759322e-05,
+      "loss": 0.6736,
+      "step": 2600
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911872422643251e-05,
+      "loss": 0.5991,
+      "step": 2601
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9118147475271797e-05,
+      "loss": 0.6,
+      "step": 2602
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911757072411108e-05,
+      "loss": 0.539,
+      "step": 2603
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9116993972950372e-05,
+      "loss": 0.7513,
+      "step": 2604
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911641722178966e-05,
+      "loss": 0.5934,
+      "step": 2605
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9115840470628948e-05,
+      "loss": 0.6626,
+      "step": 2606
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911526371946824e-05,
+      "loss": 0.5292,
+      "step": 2607
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9114686968307527e-05,
+      "loss": 0.6804,
+      "step": 2608
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9114110217146814e-05,
+      "loss": 0.7171,
+      "step": 2609
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9113533465986102e-05,
+      "loss": 0.7299,
+      "step": 2610
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911295671482539e-05,
+      "loss": 0.5819,
+      "step": 2611
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9112379963664677e-05,
+      "loss": 0.6248,
+      "step": 2612
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911180321250397e-05,
+      "loss": 0.6547,
+      "step": 2613
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9111226461343256e-05,
+      "loss": 0.6438,
+      "step": 2614
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9110649710182544e-05,
+      "loss": 0.6879,
+      "step": 2615
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.911007295902183e-05,
+      "loss": 0.582,
+      "step": 2616
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.910949620786112e-05,
+      "loss": 0.6239,
+      "step": 2617
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9108919456700407e-05,
+      "loss": 0.6549,
+      "step": 2618
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9108342705539698e-05,
+      "loss": 0.6421,
+      "step": 2619
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9107765954378986e-05,
+      "loss": 0.6614,
+      "step": 2620
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9107189203218273e-05,
+      "loss": 0.603,
+      "step": 2621
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.910661245205756e-05,
+      "loss": 0.6635,
+      "step": 2622
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.910603570089685e-05,
+      "loss": 0.6307,
+      "step": 2623
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9105458949736136e-05,
+      "loss": 0.5244,
+      "step": 2624
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9104882198575424e-05,
+      "loss": 0.6165,
+      "step": 2625
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9104305447414715e-05,
+      "loss": 0.658,
+      "step": 2626
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9103728696254003e-05,
+      "loss": 0.5962,
+      "step": 2627
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.910315194509329e-05,
+      "loss": 0.624,
+      "step": 2628
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.910257519393258e-05,
+      "loss": 0.6638,
+      "step": 2629
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9101998442771866e-05,
+      "loss": 0.6636,
+      "step": 2630
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9101421691611154e-05,
+      "loss": 0.7312,
+      "step": 2631
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9100844940450445e-05,
+      "loss": 0.6115,
+      "step": 2632
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9100268189289732e-05,
+      "loss": 0.5887,
+      "step": 2633
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909969143812902e-05,
+      "loss": 0.5574,
+      "step": 2634
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909911468696831e-05,
+      "loss": 0.6462,
+      "step": 2635
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90985379358076e-05,
+      "loss": 0.6561,
+      "step": 2636
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9097961184646886e-05,
+      "loss": 0.6442,
+      "step": 2637
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9097384433486174e-05,
+      "loss": 0.7014,
+      "step": 2638
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9096807682325462e-05,
+      "loss": 0.6186,
+      "step": 2639
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909623093116475e-05,
+      "loss": 0.6692,
+      "step": 2640
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909565418000404e-05,
+      "loss": 0.6817,
+      "step": 2641
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9095077428843328e-05,
+      "loss": 0.6298,
+      "step": 2642
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9094500677682616e-05,
+      "loss": 0.6806,
+      "step": 2643
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9093923926521904e-05,
+      "loss": 0.6532,
+      "step": 2644
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909334717536119e-05,
+      "loss": 0.6563,
+      "step": 2645
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909277042420048e-05,
+      "loss": 0.6625,
+      "step": 2646
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.909219367303977e-05,
+      "loss": 0.5862,
+      "step": 2647
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9091616921879058e-05,
+      "loss": 0.6367,
+      "step": 2648
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9091040170718345e-05,
+      "loss": 0.6687,
+      "step": 2649
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9090463419557633e-05,
+      "loss": 0.614,
+      "step": 2650
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.908988666839692e-05,
+      "loss": 0.6544,
+      "step": 2651
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.908930991723621e-05,
+      "loss": 0.6812,
+      "step": 2652
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90887331660755e-05,
+      "loss": 0.7782,
+      "step": 2653
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9088156414914787e-05,
+      "loss": 0.6306,
+      "step": 2654
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9087579663754075e-05,
+      "loss": 0.5274,
+      "step": 2655
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9087002912593363e-05,
+      "loss": 0.5886,
+      "step": 2656
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.908642616143265e-05,
+      "loss": 0.5502,
+      "step": 2657
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9085849410271938e-05,
+      "loss": 0.6393,
+      "step": 2658
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9085272659111226e-05,
+      "loss": 0.6342,
+      "step": 2659
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9084695907950517e-05,
+      "loss": 0.533,
+      "step": 2660
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9084119156789804e-05,
+      "loss": 0.7351,
+      "step": 2661
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9083542405629092e-05,
+      "loss": 0.6635,
+      "step": 2662
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9082965654468383e-05,
+      "loss": 0.5385,
+      "step": 2663
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.908238890330767e-05,
+      "loss": 0.6249,
+      "step": 2664
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9081812152146955e-05,
+      "loss": 0.6682,
+      "step": 2665
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9081235400986246e-05,
+      "loss": 0.5251,
+      "step": 2666
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9080658649825534e-05,
+      "loss": 0.7297,
+      "step": 2667
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.908008189866482e-05,
+      "loss": 0.623,
+      "step": 2668
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9079505147504113e-05,
+      "loss": 0.6535,
+      "step": 2669
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90789283963434e-05,
+      "loss": 0.6456,
+      "step": 2670
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9078351645182688e-05,
+      "loss": 0.6785,
+      "step": 2671
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9077774894021976e-05,
+      "loss": 0.7716,
+      "step": 2672
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9077198142861263e-05,
+      "loss": 0.7266,
+      "step": 2673
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.907662139170055e-05,
+      "loss": 0.6035,
+      "step": 2674
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9076044640539842e-05,
+      "loss": 0.6199,
+      "step": 2675
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.907546788937913e-05,
+      "loss": 0.5984,
+      "step": 2676
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9074891138218418e-05,
+      "loss": 0.5886,
+      "step": 2677
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9074314387057705e-05,
+      "loss": 0.6541,
+      "step": 2678
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9073737635896993e-05,
+      "loss": 0.5961,
+      "step": 2679
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.907316088473628e-05,
+      "loss": 0.5867,
+      "step": 2680
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.907258413357557e-05,
+      "loss": 0.7446,
+      "step": 2681
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.907200738241486e-05,
+      "loss": 0.6811,
+      "step": 2682
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9071430631254147e-05,
+      "loss": 0.6065,
+      "step": 2683
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9070853880093435e-05,
+      "loss": 0.6187,
+      "step": 2684
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9070277128932722e-05,
+      "loss": 0.6902,
+      "step": 2685
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.906970037777201e-05,
+      "loss": 0.6371,
+      "step": 2686
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.90691236266113e-05,
+      "loss": 0.7172,
+      "step": 2687
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.906854687545059e-05,
+      "loss": 0.5623,
+      "step": 2688
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9067970124289877e-05,
+      "loss": 0.6723,
+      "step": 2689
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9067393373129168e-05,
+      "loss": 0.6241,
+      "step": 2690
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9066816621968455e-05,
+      "loss": 0.7021,
+      "step": 2691
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.906623987080774e-05,
+      "loss": 0.6227,
+      "step": 2692
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9065663119647027e-05,
+      "loss": 0.531,
+      "step": 2693
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.906508636848632e-05,
+      "loss": 0.5467,
+      "step": 2694
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9064509617325606e-05,
+      "loss": 0.4997,
+      "step": 2695
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9063932866164894e-05,
+      "loss": 0.5998,
+      "step": 2696
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9063356115004185e-05,
+      "loss": 0.6888,
+      "step": 2697
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9062779363843472e-05,
+      "loss": 0.5715,
+      "step": 2698
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.906220261268276e-05,
+      "loss": 0.5492,
+      "step": 2699
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9061625861522048e-05,
+      "loss": 0.6726,
+      "step": 2700
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9061049110361336e-05,
+      "loss": 0.7067,
+      "step": 2701
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9060472359200623e-05,
+      "loss": 0.5987,
+      "step": 2702
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9059895608039914e-05,
+      "loss": 0.619,
+      "step": 2703
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9059318856879202e-05,
+      "loss": 0.7162,
+      "step": 2704
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905874210571849e-05,
+      "loss": 0.6108,
+      "step": 2705
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9058165354557777e-05,
+      "loss": 0.5903,
+      "step": 2706
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9057588603397065e-05,
+      "loss": 0.5759,
+      "step": 2707
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9057011852236353e-05,
+      "loss": 0.6096,
+      "step": 2708
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9056435101075644e-05,
+      "loss": 0.741,
+      "step": 2709
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905585834991493e-05,
+      "loss": 0.5811,
+      "step": 2710
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905528159875422e-05,
+      "loss": 0.614,
+      "step": 2711
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9054704847593507e-05,
+      "loss": 0.7519,
+      "step": 2712
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9054128096432795e-05,
+      "loss": 0.6418,
+      "step": 2713
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9053551345272082e-05,
+      "loss": 0.5175,
+      "step": 2714
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9052974594111373e-05,
+      "loss": 0.6315,
+      "step": 2715
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905239784295066e-05,
+      "loss": 0.5217,
+      "step": 2716
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.905182109178995e-05,
+      "loss": 0.6186,
+      "step": 2717
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9051244340629236e-05,
+      "loss": 0.5979,
+      "step": 2718
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9050667589468524e-05,
+      "loss": 0.7669,
+      "step": 2719
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9050090838307812e-05,
+      "loss": 0.6675,
+      "step": 2720
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9049514087147103e-05,
+      "loss": 0.6386,
+      "step": 2721
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904893733598639e-05,
+      "loss": 0.5176,
+      "step": 2722
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9048360584825678e-05,
+      "loss": 0.6524,
+      "step": 2723
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904778383366497e-05,
+      "loss": 0.5721,
+      "step": 2724
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9047207082504257e-05,
+      "loss": 0.6756,
+      "step": 2725
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9046630331343545e-05,
+      "loss": 0.5335,
+      "step": 2726
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904605358018283e-05,
+      "loss": 0.6164,
+      "step": 2727
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904547682902212e-05,
+      "loss": 0.6705,
+      "step": 2728
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9044900077861408e-05,
+      "loss": 0.6688,
+      "step": 2729
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9044323326700695e-05,
+      "loss": 0.8078,
+      "step": 2730
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9043746575539986e-05,
+      "loss": 0.6714,
+      "step": 2731
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9043169824379274e-05,
+      "loss": 0.5081,
+      "step": 2732
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9042593073218562e-05,
+      "loss": 0.6205,
+      "step": 2733
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.904201632205785e-05,
+      "loss": 0.6459,
+      "step": 2734
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9041439570897137e-05,
+      "loss": 0.5882,
+      "step": 2735
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9040862819736425e-05,
+      "loss": 0.667,
+      "step": 2736
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9040286068575716e-05,
+      "loss": 0.6424,
+      "step": 2737
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9039709317415004e-05,
+      "loss": 0.6061,
+      "step": 2738
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903913256625429e-05,
+      "loss": 0.7715,
+      "step": 2739
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903855581509358e-05,
+      "loss": 0.6302,
+      "step": 2740
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9037979063932867e-05,
+      "loss": 0.5998,
+      "step": 2741
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9037402312772154e-05,
+      "loss": 0.6166,
+      "step": 2742
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9036825561611445e-05,
+      "loss": 0.5535,
+      "step": 2743
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9036248810450733e-05,
+      "loss": 0.7159,
+      "step": 2744
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903567205929002e-05,
+      "loss": 0.57,
+      "step": 2745
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903509530812931e-05,
+      "loss": 0.586,
+      "step": 2746
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9034518556968596e-05,
+      "loss": 0.6376,
+      "step": 2747
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033941805807884e-05,
+      "loss": 0.6155,
+      "step": 2748
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9033365054647175e-05,
+      "loss": 0.6251,
+      "step": 2749
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9032788303486463e-05,
+      "loss": 0.6417,
+      "step": 2750
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903221155232575e-05,
+      "loss": 0.6014,
+      "step": 2751
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903163480116504e-05,
+      "loss": 0.6232,
+      "step": 2752
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.903105805000433e-05,
+      "loss": 0.6674,
+      "step": 2753
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9030481298843613e-05,
+      "loss": 0.6199,
+      "step": 2754
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029904547682904e-05,
+      "loss": 0.6551,
+      "step": 2755
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9029327796522192e-05,
+      "loss": 0.5012,
+      "step": 2756
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902875104536148e-05,
+      "loss": 0.7272,
+      "step": 2757
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902817429420077e-05,
+      "loss": 0.6901,
+      "step": 2758
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902759754304006e-05,
+      "loss": 0.5888,
+      "step": 2759
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9027020791879346e-05,
+      "loss": 0.6678,
+      "step": 2760
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9026444040718634e-05,
+      "loss": 0.6677,
+      "step": 2761
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902586728955792e-05,
+      "loss": 0.7583,
+      "step": 2762
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902529053839721e-05,
+      "loss": 0.581,
+      "step": 2763
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024713787236497e-05,
+      "loss": 0.6161,
+      "step": 2764
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9024137036075788e-05,
+      "loss": 0.6522,
+      "step": 2765
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9023560284915076e-05,
+      "loss": 0.5128,
+      "step": 2766
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9022983533754363e-05,
+      "loss": 0.8596,
+      "step": 2767
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902240678259365e-05,
+      "loss": 0.5397,
+      "step": 2768
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.902183003143294e-05,
+      "loss": 0.6812,
+      "step": 2769
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9021253280272226e-05,
+      "loss": 0.6704,
+      "step": 2770
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9020676529111518e-05,
+      "loss": 0.5345,
+      "step": 2771
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9020099777950805e-05,
+      "loss": 0.5247,
+      "step": 2772
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9019523026790093e-05,
+      "loss": 0.7744,
+      "step": 2773
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.901894627562938e-05,
+      "loss": 0.5063,
+      "step": 2774
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9018369524468668e-05,
+      "loss": 0.5938,
+      "step": 2775
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9017792773307956e-05,
+      "loss": 0.6662,
+      "step": 2776
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9017216022147247e-05,
+      "loss": 0.655,
+      "step": 2777
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9016639270986535e-05,
+      "loss": 0.6907,
+      "step": 2778
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9016062519825822e-05,
+      "loss": 0.7099,
+      "step": 2779
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.901548576866511e-05,
+      "loss": 0.7175,
+      "step": 2780
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9014909017504398e-05,
+      "loss": 0.7105,
+      "step": 2781
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9014332266343685e-05,
+      "loss": 0.65,
+      "step": 2782
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9013755515182977e-05,
+      "loss": 0.6132,
+      "step": 2783
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9013178764022264e-05,
+      "loss": 0.5219,
+      "step": 2784
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9012602012861552e-05,
+      "loss": 0.7321,
+      "step": 2785
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9012025261700843e-05,
+      "loss": 0.6788,
+      "step": 2786
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.901144851054013e-05,
+      "loss": 0.5123,
+      "step": 2787
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.901087175937942e-05,
+      "loss": 0.6718,
+      "step": 2788
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9010295008218706e-05,
+      "loss": 0.6006,
+      "step": 2789
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9009718257057994e-05,
+      "loss": 0.6016,
+      "step": 2790
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900914150589728e-05,
+      "loss": 0.5677,
+      "step": 2791
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9008564754736572e-05,
+      "loss": 0.6599,
+      "step": 2792
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900798800357586e-05,
+      "loss": 0.6152,
+      "step": 2793
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9007411252415148e-05,
+      "loss": 0.6272,
+      "step": 2794
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9006834501254436e-05,
+      "loss": 0.5838,
+      "step": 2795
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9006257750093723e-05,
+      "loss": 0.6966,
+      "step": 2796
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900568099893301e-05,
+      "loss": 0.6208,
+      "step": 2797
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90051042477723e-05,
+      "loss": 0.6583,
+      "step": 2798
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900452749661159e-05,
+      "loss": 0.5827,
+      "step": 2799
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9003950745450877e-05,
+      "loss": 0.5324,
+      "step": 2800
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9003373994290165e-05,
+      "loss": 0.6182,
+      "step": 2801
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9002797243129453e-05,
+      "loss": 0.7547,
+      "step": 2802
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900222049196874e-05,
+      "loss": 0.5655,
+      "step": 2803
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9001643740808028e-05,
+      "loss": 0.5926,
+      "step": 2804
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.900106698964732e-05,
+      "loss": 0.4872,
+      "step": 2805
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9000490238486607e-05,
+      "loss": 0.5473,
+      "step": 2806
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8999913487325895e-05,
+      "loss": 0.5827,
+      "step": 2807
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8999336736165182e-05,
+      "loss": 0.6056,
+      "step": 2808
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.899875998500447e-05,
+      "loss": 0.5966,
+      "step": 2809
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8998183233843758e-05,
+      "loss": 0.635,
+      "step": 2810
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.899760648268305e-05,
+      "loss": 0.6449,
+      "step": 2811
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8997029731522336e-05,
+      "loss": 0.6498,
+      "step": 2812
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8996452980361624e-05,
+      "loss": 0.7185,
+      "step": 2813
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8995876229200915e-05,
+      "loss": 0.5916,
+      "step": 2814
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8995299478040203e-05,
+      "loss": 0.6178,
+      "step": 2815
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8994722726879487e-05,
+      "loss": 0.6114,
+      "step": 2816
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8994145975718778e-05,
+      "loss": 0.5972,
+      "step": 2817
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8993569224558066e-05,
+      "loss": 0.6328,
+      "step": 2818
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8992992473397354e-05,
+      "loss": 0.6534,
+      "step": 2819
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8992415722236645e-05,
+      "loss": 0.5522,
+      "step": 2820
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8991838971075932e-05,
+      "loss": 0.6049,
+      "step": 2821
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.899126221991522e-05,
+      "loss": 0.6537,
+      "step": 2822
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8990685468754508e-05,
+      "loss": 0.7009,
+      "step": 2823
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8990108717593795e-05,
+      "loss": 0.5928,
+      "step": 2824
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8989531966433083e-05,
+      "loss": 0.6562,
+      "step": 2825
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8988955215272374e-05,
+      "loss": 0.7059,
+      "step": 2826
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8988378464111662e-05,
+      "loss": 0.6492,
+      "step": 2827
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898780171295095e-05,
+      "loss": 0.6041,
+      "step": 2828
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8987224961790237e-05,
+      "loss": 0.5275,
+      "step": 2829
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8986648210629525e-05,
+      "loss": 0.6241,
+      "step": 2830
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8986071459468813e-05,
+      "loss": 0.7912,
+      "step": 2831
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89854947083081e-05,
+      "loss": 0.6123,
+      "step": 2832
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898491795714739e-05,
+      "loss": 0.5732,
+      "step": 2833
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898434120598668e-05,
+      "loss": 0.5961,
+      "step": 2834
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8983764454825967e-05,
+      "loss": 0.6482,
+      "step": 2835
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8983187703665254e-05,
+      "loss": 0.5942,
+      "step": 2836
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8982610952504542e-05,
+      "loss": 0.5312,
+      "step": 2837
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898203420134383e-05,
+      "loss": 0.593,
+      "step": 2838
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898145745018312e-05,
+      "loss": 0.6282,
+      "step": 2839
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.898088069902241e-05,
+      "loss": 0.6778,
+      "step": 2840
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8980303947861696e-05,
+      "loss": 0.6662,
+      "step": 2841
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8979727196700984e-05,
+      "loss": 0.5336,
+      "step": 2842
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897915044554027e-05,
+      "loss": 0.5291,
+      "step": 2843
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897857369437956e-05,
+      "loss": 0.7456,
+      "step": 2844
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897799694321885e-05,
+      "loss": 0.6453,
+      "step": 2845
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8977420192058138e-05,
+      "loss": 0.6094,
+      "step": 2846
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8976843440897426e-05,
+      "loss": 0.5533,
+      "step": 2847
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8976266689736717e-05,
+      "loss": 0.5584,
+      "step": 2848
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8975689938576004e-05,
+      "loss": 0.585,
+      "step": 2849
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8975113187415292e-05,
+      "loss": 0.6731,
+      "step": 2850
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897453643625458e-05,
+      "loss": 0.5298,
+      "step": 2851
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8973959685093867e-05,
+      "loss": 0.6616,
+      "step": 2852
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8973382933933155e-05,
+      "loss": 0.7344,
+      "step": 2853
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8972806182772446e-05,
+      "loss": 0.673,
+      "step": 2854
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8972229431611734e-05,
+      "loss": 0.6851,
+      "step": 2855
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897165268045102e-05,
+      "loss": 0.6464,
+      "step": 2856
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.897107592929031e-05,
+      "loss": 0.5667,
+      "step": 2857
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8970499178129597e-05,
+      "loss": 0.7056,
+      "step": 2858
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8969922426968885e-05,
+      "loss": 0.5369,
+      "step": 2859
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8969345675808176e-05,
+      "loss": 0.6245,
+      "step": 2860
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8968768924647463e-05,
+      "loss": 0.63,
+      "step": 2861
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896819217348675e-05,
+      "loss": 0.5734,
+      "step": 2862
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896761542232604e-05,
+      "loss": 0.7298,
+      "step": 2863
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8967038671165326e-05,
+      "loss": 0.6249,
+      "step": 2864
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8966461920004614e-05,
+      "loss": 0.6364,
+      "step": 2865
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965885168843905e-05,
+      "loss": 0.7025,
+      "step": 2866
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8965308417683193e-05,
+      "loss": 0.6679,
+      "step": 2867
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896473166652248e-05,
+      "loss": 0.7732,
+      "step": 2868
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8964154915361768e-05,
+      "loss": 0.6378,
+      "step": 2869
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963578164201056e-05,
+      "loss": 0.7273,
+      "step": 2870
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8963001413040344e-05,
+      "loss": 0.6217,
+      "step": 2871
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896242466187963e-05,
+      "loss": 0.645,
+      "step": 2872
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8961847910718922e-05,
+      "loss": 0.5792,
+      "step": 2873
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896127115955821e-05,
+      "loss": 0.6343,
+      "step": 2874
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8960694408397498e-05,
+      "loss": 0.6383,
+      "step": 2875
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.896011765723679e-05,
+      "loss": 0.5233,
+      "step": 2876
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8959540906076077e-05,
+      "loss": 0.5789,
+      "step": 2877
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895896415491536e-05,
+      "loss": 0.7234,
+      "step": 2878
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8958387403754652e-05,
+      "loss": 0.6031,
+      "step": 2879
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895781065259394e-05,
+      "loss": 0.6569,
+      "step": 2880
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8957233901433227e-05,
+      "loss": 0.6922,
+      "step": 2881
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895665715027252e-05,
+      "loss": 0.5641,
+      "step": 2882
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8956080399111806e-05,
+      "loss": 0.6705,
+      "step": 2883
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8955503647951094e-05,
+      "loss": 0.5756,
+      "step": 2884
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895492689679038e-05,
+      "loss": 0.6132,
+      "step": 2885
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895435014562967e-05,
+      "loss": 0.6312,
+      "step": 2886
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953773394468957e-05,
+      "loss": 0.6847,
+      "step": 2887
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8953196643308248e-05,
+      "loss": 0.5513,
+      "step": 2888
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8952619892147536e-05,
+      "loss": 0.7527,
+      "step": 2889
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8952043140986823e-05,
+      "loss": 0.5817,
+      "step": 2890
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.895146638982611e-05,
+      "loss": 0.6439,
+      "step": 2891
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89508896386654e-05,
+      "loss": 0.6729,
+      "step": 2892
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8950312887504686e-05,
+      "loss": 0.655,
+      "step": 2893
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949736136343977e-05,
+      "loss": 0.6307,
+      "step": 2894
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8949159385183265e-05,
+      "loss": 0.7716,
+      "step": 2895
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8948582634022553e-05,
+      "loss": 0.6046,
+      "step": 2896
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894800588286184e-05,
+      "loss": 0.6696,
+      "step": 2897
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8947429131701128e-05,
+      "loss": 0.5987,
+      "step": 2898
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8946852380540416e-05,
+      "loss": 0.6687,
+      "step": 2899
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8946275629379707e-05,
+      "loss": 0.6482,
+      "step": 2900
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8945698878218995e-05,
+      "loss": 0.6216,
+      "step": 2901
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8945122127058282e-05,
+      "loss": 0.5893,
+      "step": 2902
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894454537589757e-05,
+      "loss": 0.6784,
+      "step": 2903
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943968624736858e-05,
+      "loss": 0.6778,
+      "step": 2904
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8943391873576145e-05,
+      "loss": 0.5846,
+      "step": 2905
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942815122415433e-05,
+      "loss": 0.5197,
+      "step": 2906
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8942238371254724e-05,
+      "loss": 0.6123,
+      "step": 2907
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8941661620094012e-05,
+      "loss": 0.5179,
+      "step": 2908
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89410848689333e-05,
+      "loss": 0.6332,
+      "step": 2909
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.894050811777259e-05,
+      "loss": 0.5932,
+      "step": 2910
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939931366611878e-05,
+      "loss": 0.5205,
+      "step": 2911
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8939354615451166e-05,
+      "loss": 0.5912,
+      "step": 2912
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8938777864290454e-05,
+      "loss": 0.5968,
+      "step": 2913
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893820111312974e-05,
+      "loss": 0.5994,
+      "step": 2914
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893762436196903e-05,
+      "loss": 0.4617,
+      "step": 2915
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893704761080832e-05,
+      "loss": 0.626,
+      "step": 2916
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8936470859647608e-05,
+      "loss": 0.747,
+      "step": 2917
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8935894108486895e-05,
+      "loss": 0.6223,
+      "step": 2918
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8935317357326183e-05,
+      "loss": 0.6411,
+      "step": 2919
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893474060616547e-05,
+      "loss": 0.6668,
+      "step": 2920
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893416385500476e-05,
+      "loss": 0.6157,
+      "step": 2921
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893358710384405e-05,
+      "loss": 0.7753,
+      "step": 2922
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8933010352683337e-05,
+      "loss": 0.5473,
+      "step": 2923
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8932433601522625e-05,
+      "loss": 0.6871,
+      "step": 2924
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8931856850361913e-05,
+      "loss": 0.6094,
+      "step": 2925
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89312800992012e-05,
+      "loss": 0.622,
+      "step": 2926
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8930703348040488e-05,
+      "loss": 0.4968,
+      "step": 2927
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.893012659687978e-05,
+      "loss": 0.7533,
+      "step": 2928
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8929549845719067e-05,
+      "loss": 0.6504,
+      "step": 2929
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928973094558354e-05,
+      "loss": 0.7715,
+      "step": 2930
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8928396343397642e-05,
+      "loss": 0.6325,
+      "step": 2931
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892781959223693e-05,
+      "loss": 0.6127,
+      "step": 2932
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8927242841076217e-05,
+      "loss": 0.5965,
+      "step": 2933
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892666608991551e-05,
+      "loss": 0.6025,
+      "step": 2934
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8926089338754796e-05,
+      "loss": 0.7439,
+      "step": 2935
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8925512587594084e-05,
+      "loss": 0.6126,
+      "step": 2936
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892493583643337e-05,
+      "loss": 0.6335,
+      "step": 2937
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8924359085272663e-05,
+      "loss": 0.4578,
+      "step": 2938
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892378233411195e-05,
+      "loss": 0.5444,
+      "step": 2939
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8923205582951235e-05,
+      "loss": 0.6745,
+      "step": 2940
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922628831790526e-05,
+      "loss": 0.7133,
+      "step": 2941
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8922052080629813e-05,
+      "loss": 0.6515,
+      "step": 2942
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.89214753294691e-05,
+      "loss": 0.486,
+      "step": 2943
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8920898578308392e-05,
+      "loss": 0.7319,
+      "step": 2944
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.892032182714768e-05,
+      "loss": 0.5929,
+      "step": 2945
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8919745075986967e-05,
+      "loss": 0.5302,
+      "step": 2946
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8919168324826255e-05,
+      "loss": 0.6374,
+      "step": 2947
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.8918591573665543e-05,
+      "loss": 0.5509,
+      "step": 2948
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.891801482250483e-05,
+      "loss": 0.6144,
+      "step": 2949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891743807134412e-05,
+      "loss": 0.6623,
+      "step": 2950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891686132018341e-05,
+      "loss": 0.5135,
+      "step": 2951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8916284569022697e-05,
+      "loss": 0.62,
+      "step": 2952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8915707817861985e-05,
+      "loss": 0.616,
+      "step": 2953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8915131066701272e-05,
+      "loss": 0.5838,
+      "step": 2954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891455431554056e-05,
+      "loss": 0.5989,
+      "step": 2955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891397756437985e-05,
+      "loss": 0.6686,
+      "step": 2956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891340081321914e-05,
+      "loss": 0.5074,
+      "step": 2957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8912824062058426e-05,
+      "loss": 0.7445,
+      "step": 2958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8912247310897714e-05,
+      "loss": 0.5512,
+      "step": 2959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8911670559737002e-05,
+      "loss": 0.5625,
+      "step": 2960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891109380857629e-05,
+      "loss": 0.66,
+      "step": 2961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891051705741558e-05,
+      "loss": 0.5773,
+      "step": 2962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8909940306254868e-05,
+      "loss": 0.6946,
+      "step": 2963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8909363555094156e-05,
+      "loss": 0.6051,
+      "step": 2964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8908786803933447e-05,
+      "loss": 0.6238,
+      "step": 2965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890821005277273e-05,
+      "loss": 0.5323,
+      "step": 2966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890763330161202e-05,
+      "loss": 0.7091,
+      "step": 2967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890705655045131e-05,
+      "loss": 0.591,
+      "step": 2968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8906479799290598e-05,
+      "loss": 0.5887,
+      "step": 2969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8905903048129885e-05,
+      "loss": 0.6793,
+      "step": 2970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8905326296969173e-05,
+      "loss": 0.6636,
+      "step": 2971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8904749545808464e-05,
+      "loss": 0.5272,
+      "step": 2972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8904172794647752e-05,
+      "loss": 0.5549,
+      "step": 2973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890359604348704e-05,
+      "loss": 0.4857,
+      "step": 2974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8903019292326327e-05,
+      "loss": 0.6894,
+      "step": 2975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8902442541165615e-05,
+      "loss": 0.6223,
+      "step": 2976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8901865790004903e-05,
+      "loss": 0.6439,
+      "step": 2977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8901289038844194e-05,
+      "loss": 0.5703,
+      "step": 2978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890071228768348e-05,
+      "loss": 0.6012,
+      "step": 2979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.890013553652277e-05,
+      "loss": 0.5069,
+      "step": 2980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8899558785362057e-05,
+      "loss": 0.6528,
+      "step": 2981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8898982034201344e-05,
+      "loss": 0.6382,
+      "step": 2982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8898405283040632e-05,
+      "loss": 0.561,
+      "step": 2983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8897828531879923e-05,
+      "loss": 0.6317,
+      "step": 2984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889725178071921e-05,
+      "loss": 0.5874,
+      "step": 2985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88966750295585e-05,
+      "loss": 0.6339,
+      "step": 2986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8896098278397786e-05,
+      "loss": 0.7107,
+      "step": 2987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8895521527237074e-05,
+      "loss": 0.6564,
+      "step": 2988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889494477607636e-05,
+      "loss": 0.7423,
+      "step": 2989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8894368024915653e-05,
+      "loss": 0.6596,
+      "step": 2990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889379127375494e-05,
+      "loss": 0.6756,
+      "step": 2991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8893214522594228e-05,
+      "loss": 0.6709,
+      "step": 2992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892637771433516e-05,
+      "loss": 0.7767,
+      "step": 2993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8892061020272803e-05,
+      "loss": 0.5217,
+      "step": 2994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889148426911209e-05,
+      "loss": 0.6546,
+      "step": 2995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8890907517951382e-05,
+      "loss": 0.5878,
+      "step": 2996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.889033076679067e-05,
+      "loss": 0.7083,
+      "step": 2997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8889754015629958e-05,
+      "loss": 0.5189,
+      "step": 2998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888917726446925e-05,
+      "loss": 0.6052,
+      "step": 2999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8888600513308536e-05,
+      "loss": 0.4818,
+      "step": 3000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8888023762147824e-05,
+      "loss": 0.5567,
+      "step": 3001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8887447010987112e-05,
+      "loss": 0.7137,
+      "step": 3002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88868702598264e-05,
+      "loss": 0.7196,
+      "step": 3003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8886293508665687e-05,
+      "loss": 0.6873,
+      "step": 3004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885716757504978e-05,
+      "loss": 0.5507,
+      "step": 3005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8885140006344266e-05,
+      "loss": 0.595,
+      "step": 3006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8884563255183554e-05,
+      "loss": 0.6486,
+      "step": 3007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888398650402284e-05,
+      "loss": 0.6438,
+      "step": 3008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888340975286213e-05,
+      "loss": 0.653,
+      "step": 3009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882833001701417e-05,
+      "loss": 0.6419,
+      "step": 3010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8882256250540704e-05,
+      "loss": 0.704,
+      "step": 3011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8881679499379995e-05,
+      "loss": 0.7462,
+      "step": 3012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8881102748219283e-05,
+      "loss": 0.6493,
+      "step": 3013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.888052599705857e-05,
+      "loss": 0.6375,
+      "step": 3014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887994924589786e-05,
+      "loss": 0.6664,
+      "step": 3015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8879372494737146e-05,
+      "loss": 0.5809,
+      "step": 3016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878795743576434e-05,
+      "loss": 0.7268,
+      "step": 3017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8878218992415725e-05,
+      "loss": 0.7913,
+      "step": 3018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8877642241255013e-05,
+      "loss": 0.5887,
+      "step": 3019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88770654900943e-05,
+      "loss": 0.5351,
+      "step": 3020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8876488738933588e-05,
+      "loss": 0.6018,
+      "step": 3021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875911987772876e-05,
+      "loss": 0.7561,
+      "step": 3022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8875335236612163e-05,
+      "loss": 0.7484,
+      "step": 3023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874758485451454e-05,
+      "loss": 0.6366,
+      "step": 3024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8874181734290742e-05,
+      "loss": 0.6248,
+      "step": 3025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887360498313003e-05,
+      "loss": 0.6174,
+      "step": 3026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887302823196932e-05,
+      "loss": 0.5062,
+      "step": 3027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887245148080861e-05,
+      "loss": 0.6295,
+      "step": 3028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8871874729647893e-05,
+      "loss": 0.6174,
+      "step": 3029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8871297978487184e-05,
+      "loss": 0.7517,
+      "step": 3030
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887072122732647e-05,
+      "loss": 0.7475,
+      "step": 3031
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.887014447616576e-05,
+      "loss": 0.6089,
+      "step": 3032
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886956772500505e-05,
+      "loss": 0.5068,
+      "step": 3033
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868990973844338e-05,
+      "loss": 0.5238,
+      "step": 3034
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8868414222683626e-05,
+      "loss": 0.6216,
+      "step": 3035
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8867837471522913e-05,
+      "loss": 0.6137,
+      "step": 3036
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88672607203622e-05,
+      "loss": 0.5497,
+      "step": 3037
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886668396920149e-05,
+      "loss": 0.7159,
+      "step": 3038
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886610721804078e-05,
+      "loss": 0.7361,
+      "step": 3039
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8865530466880067e-05,
+      "loss": 0.7707,
+      "step": 3040
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864953715719355e-05,
+      "loss": 0.6718,
+      "step": 3041
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8864376964558643e-05,
+      "loss": 0.6293,
+      "step": 3042
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886380021339793e-05,
+      "loss": 0.6744,
+      "step": 3043
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8863223462237218e-05,
+      "loss": 0.5197,
+      "step": 3044
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862646711076506e-05,
+      "loss": 0.5619,
+      "step": 3045
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8862069959915797e-05,
+      "loss": 0.6945,
+      "step": 3046
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8861493208755085e-05,
+      "loss": 0.5757,
+      "step": 3047
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8860916457594372e-05,
+      "loss": 0.6188,
+      "step": 3048
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.886033970643366e-05,
+      "loss": 0.5064,
+      "step": 3049
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8859762955272948e-05,
+      "loss": 0.543,
+      "step": 3050
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8859186204112235e-05,
+      "loss": 0.575,
+      "step": 3051
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858609452951526e-05,
+      "loss": 0.6252,
+      "step": 3052
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8858032701790814e-05,
+      "loss": 0.5661,
+      "step": 3053
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8857455950630102e-05,
+      "loss": 0.7245,
+      "step": 3054
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885687919946939e-05,
+      "loss": 0.6931,
+      "step": 3055
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8856302448308677e-05,
+      "loss": 0.6327,
+      "step": 3056
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855725697147965e-05,
+      "loss": 0.6531,
+      "step": 3057
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8855148945987256e-05,
+      "loss": 0.5459,
+      "step": 3058
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8854572194826544e-05,
+      "loss": 0.7003,
+      "step": 3059
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885399544366583e-05,
+      "loss": 0.6426,
+      "step": 3060
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8853418692505122e-05,
+      "loss": 0.7206,
+      "step": 3061
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885284194134441e-05,
+      "loss": 0.5887,
+      "step": 3062
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8852265190183698e-05,
+      "loss": 0.6524,
+      "step": 3063
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8851688439022985e-05,
+      "loss": 0.7122,
+      "step": 3064
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8851111687862273e-05,
+      "loss": 0.6145,
+      "step": 3065
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.885053493670156e-05,
+      "loss": 0.6184,
+      "step": 3066
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8849958185540852e-05,
+      "loss": 0.6286,
+      "step": 3067
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884938143438014e-05,
+      "loss": 0.7177,
+      "step": 3068
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8848804683219427e-05,
+      "loss": 0.6456,
+      "step": 3069
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8848227932058715e-05,
+      "loss": 0.7225,
+      "step": 3070
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8847651180898003e-05,
+      "loss": 0.666,
+      "step": 3071
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884707442973729e-05,
+      "loss": 0.6392,
+      "step": 3072
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884649767857658e-05,
+      "loss": 0.5654,
+      "step": 3073
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884592092741587e-05,
+      "loss": 0.638,
+      "step": 3074
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8845344176255157e-05,
+      "loss": 0.6333,
+      "step": 3075
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8844767425094444e-05,
+      "loss": 0.693,
+      "step": 3076
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8844190673933732e-05,
+      "loss": 0.5174,
+      "step": 3077
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884361392277302e-05,
+      "loss": 0.7455,
+      "step": 3078
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8843037171612308e-05,
+      "loss": 0.5996,
+      "step": 3079
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88424604204516e-05,
+      "loss": 0.6254,
+      "step": 3080
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8841883669290886e-05,
+      "loss": 0.6967,
+      "step": 3081
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8841306918130174e-05,
+      "loss": 0.6138,
+      "step": 3082
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884073016696946e-05,
+      "loss": 0.5488,
+      "step": 3083
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.884015341580875e-05,
+      "loss": 0.6103,
+      "step": 3084
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8839576664648037e-05,
+      "loss": 0.6642,
+      "step": 3085
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8838999913487328e-05,
+      "loss": 0.5905,
+      "step": 3086
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8838423162326616e-05,
+      "loss": 0.5896,
+      "step": 3087
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8837846411165903e-05,
+      "loss": 0.6758,
+      "step": 3088
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8837269660005195e-05,
+      "loss": 0.6471,
+      "step": 3089
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8836692908844482e-05,
+      "loss": 0.7392,
+      "step": 3090
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8836116157683767e-05,
+      "loss": 0.5454,
+      "step": 3091
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8835539406523058e-05,
+      "loss": 0.6491,
+      "step": 3092
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8834962655362345e-05,
+      "loss": 0.5508,
+      "step": 3093
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8834385904201633e-05,
+      "loss": 0.6327,
+      "step": 3094
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8833809153040924e-05,
+      "loss": 0.7641,
+      "step": 3095
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8833232401880212e-05,
+      "loss": 0.527,
+      "step": 3096
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88326556507195e-05,
+      "loss": 0.6861,
+      "step": 3097
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8832078899558787e-05,
+      "loss": 0.5225,
+      "step": 3098
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8831502148398075e-05,
+      "loss": 0.4898,
+      "step": 3099
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8830925397237362e-05,
+      "loss": 0.6523,
+      "step": 3100
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8830348646076654e-05,
+      "loss": 0.6297,
+      "step": 3101
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882977189491594e-05,
+      "loss": 0.6621,
+      "step": 3102
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882919514375523e-05,
+      "loss": 0.6045,
+      "step": 3103
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8828618392594517e-05,
+      "loss": 0.6374,
+      "step": 3104
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8828041641433804e-05,
+      "loss": 0.5339,
+      "step": 3105
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8827464890273092e-05,
+      "loss": 0.4724,
+      "step": 3106
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8826888139112383e-05,
+      "loss": 0.652,
+      "step": 3107
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882631138795167e-05,
+      "loss": 0.6009,
+      "step": 3108
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882573463679096e-05,
+      "loss": 0.6058,
+      "step": 3109
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8825157885630246e-05,
+      "loss": 0.7165,
+      "step": 3110
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8824581134469534e-05,
+      "loss": 0.6584,
+      "step": 3111
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882400438330882e-05,
+      "loss": 0.5809,
+      "step": 3112
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882342763214811e-05,
+      "loss": 0.6407,
+      "step": 3113
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.88228508809874e-05,
+      "loss": 0.6507,
+      "step": 3114
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8822274129826688e-05,
+      "loss": 0.7781,
+      "step": 3115
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8821697378665976e-05,
+      "loss": 0.5727,
+      "step": 3116
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8821120627505263e-05,
+      "loss": 0.6757,
+      "step": 3117
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.882054387634455e-05,
+      "loss": 0.6841,
+      "step": 3118
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.881996712518384e-05,
+      "loss": 0.5835,
+      "step": 3119
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.881939037402313e-05,
+      "loss": 0.6262,
+      "step": 3120
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8818813622862417e-05,
+      "loss": 0.6113,
+      "step": 3121
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8818236871701705e-05,
+      "loss": 0.7407,
+      "step": 3122
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8817660120540996e-05,
+      "loss": 0.701,
+      "step": 3123
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8817083369380284e-05,
+      "loss": 0.6555,
+      "step": 3124
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.881650661821957e-05,
+      "loss": 0.611,
+      "step": 3125
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.881592986705886e-05,
+      "loss": 0.5627,
+      "step": 3126
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8815353115898147e-05,
+      "loss": 0.6516,
+      "step": 3127
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8814776364737435e-05,
+      "loss": 0.5863,
+      "step": 3128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814199613576726e-05,
+      "loss": 0.5703,
+      "step": 3129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813622862416013e-05,
+      "loss": 0.6198,
+      "step": 3130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88130461112553e-05,
+      "loss": 0.7801,
+      "step": 3131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881246936009459e-05,
+      "loss": 0.6023,
+      "step": 3132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811892608933876e-05,
+      "loss": 0.6183,
+      "step": 3133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8811315857773164e-05,
+      "loss": 0.6988,
+      "step": 3134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810739106612455e-05,
+      "loss": 0.5332,
+      "step": 3135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810162355451743e-05,
+      "loss": 0.5851,
+      "step": 3136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880958560429103e-05,
+      "loss": 0.6494,
+      "step": 3137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809008853130318e-05,
+      "loss": 0.5766,
+      "step": 3138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808432101969606e-05,
+      "loss": 0.7212,
+      "step": 3139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807855350808894e-05,
+      "loss": 0.5901,
+      "step": 3140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807278599648185e-05,
+      "loss": 0.448,
+      "step": 3141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806701848487472e-05,
+      "loss": 0.6849,
+      "step": 3142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880612509732676e-05,
+      "loss": 0.6119,
+      "step": 3143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805548346166048e-05,
+      "loss": 0.7288,
+      "step": 3144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804971595005335e-05,
+      "loss": 0.756,
+      "step": 3145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804394843844623e-05,
+      "loss": 0.6594,
+      "step": 3146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880381809268391e-05,
+      "loss": 0.5176,
+      "step": 3147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803241341523202e-05,
+      "loss": 0.6793,
+      "step": 3148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880266459036249e-05,
+      "loss": 0.5334,
+      "step": 3149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802087839201777e-05,
+      "loss": 0.627,
+      "step": 3150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8801511088041068e-05,
+      "loss": 0.5625,
+      "step": 3151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800934336880356e-05,
+      "loss": 0.6899,
+      "step": 3152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880035758571964e-05,
+      "loss": 0.6741,
+      "step": 3153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879978083455893e-05,
+      "loss": 0.6514,
+      "step": 3154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879920408339822e-05,
+      "loss": 0.7364,
+      "step": 3155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798627332237507e-05,
+      "loss": 0.7188,
+      "step": 3156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798050581076798e-05,
+      "loss": 0.5811,
+      "step": 3157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797473829916085e-05,
+      "loss": 0.6205,
+      "step": 3158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796897078755373e-05,
+      "loss": 0.6397,
+      "step": 3159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879632032759466e-05,
+      "loss": 0.5664,
+      "step": 3160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879574357643395e-05,
+      "loss": 0.6401,
+      "step": 3161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8795166825273236e-05,
+      "loss": 0.6797,
+      "step": 3162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794590074112527e-05,
+      "loss": 0.5846,
+      "step": 3163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794013322951815e-05,
+      "loss": 0.5875,
+      "step": 3164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793436571791103e-05,
+      "loss": 0.527,
+      "step": 3165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879285982063039e-05,
+      "loss": 0.6083,
+      "step": 3166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792283069469678e-05,
+      "loss": 0.5605,
+      "step": 3167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791706318308966e-05,
+      "loss": 0.5247,
+      "step": 3168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8791129567148257e-05,
+      "loss": 0.6371,
+      "step": 3169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790552815987544e-05,
+      "loss": 0.5978,
+      "step": 3170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789976064826832e-05,
+      "loss": 0.4671,
+      "step": 3171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878939931366612e-05,
+      "loss": 0.6364,
+      "step": 3172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788822562505408e-05,
+      "loss": 0.6124,
+      "step": 3173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788245811344695e-05,
+      "loss": 0.5308,
+      "step": 3174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787669060183986e-05,
+      "loss": 0.6989,
+      "step": 3175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787092309023274e-05,
+      "loss": 0.448,
+      "step": 3176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878651555786256e-05,
+      "loss": 0.5588,
+      "step": 3177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785938806701853e-05,
+      "loss": 0.6127,
+      "step": 3178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785362055541137e-05,
+      "loss": 0.5495,
+      "step": 3179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784785304380425e-05,
+      "loss": 0.6145,
+      "step": 3180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784208553219712e-05,
+      "loss": 0.6034,
+      "step": 3181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783631802059003e-05,
+      "loss": 0.6877,
+      "step": 3182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878305505089829e-05,
+      "loss": 0.5544,
+      "step": 3183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878247829973758e-05,
+      "loss": 0.6615,
+      "step": 3184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878190154857687e-05,
+      "loss": 0.6944,
+      "step": 3185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781324797416158e-05,
+      "loss": 0.6531,
+      "step": 3186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780748046255445e-05,
+      "loss": 0.5253,
+      "step": 3187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780171295094733e-05,
+      "loss": 0.6156,
+      "step": 3188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877959454393402e-05,
+      "loss": 0.4589,
+      "step": 3189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877901779277331e-05,
+      "loss": 0.5569,
+      "step": 3190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87784410416126e-05,
+      "loss": 0.5726,
+      "step": 3191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777864290451887e-05,
+      "loss": 0.675,
+      "step": 3192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777287539291175e-05,
+      "loss": 0.6249,
+      "step": 3193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776710788130462e-05,
+      "loss": 0.6936,
+      "step": 3194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877613403696975e-05,
+      "loss": 0.6713,
+      "step": 3195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775557285809038e-05,
+      "loss": 0.5392,
+      "step": 3196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877498053464833e-05,
+      "loss": 0.6129,
+      "step": 3197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774403783487617e-05,
+      "loss": 0.497,
+      "step": 3198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773827032326904e-05,
+      "loss": 0.6854,
+      "step": 3199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8773250281166192e-05,
+      "loss": 0.7105,
+      "step": 3200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877267353000548e-05,
+      "loss": 0.5679,
+      "step": 3201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772096778844767e-05,
+      "loss": 0.6035,
+      "step": 3202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877152002768406e-05,
+      "loss": 0.6326,
+      "step": 3203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770943276523346e-05,
+      "loss": 0.7288,
+      "step": 3204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770366525362634e-05,
+      "loss": 0.7211,
+      "step": 3205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876978977420192e-05,
+      "loss": 0.7289,
+      "step": 3206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876921302304121e-05,
+      "loss": 0.6291,
+      "step": 3207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768636271880497e-05,
+      "loss": 0.5311,
+      "step": 3208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768059520719788e-05,
+      "loss": 0.6542,
+      "step": 3209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767482769559076e-05,
+      "loss": 0.6761,
+      "step": 3210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766906018398363e-05,
+      "loss": 0.6361,
+      "step": 3211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766329267237654e-05,
+      "loss": 0.837,
+      "step": 3212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8765752516076942e-05,
+      "loss": 0.6012,
+      "step": 3213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876517576491623e-05,
+      "loss": 0.7013,
+      "step": 3214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8764599013755514e-05,
+      "loss": 0.4937,
+      "step": 3215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8764022262594805e-05,
+      "loss": 0.5812,
+      "step": 3216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8763445511434093e-05,
+      "loss": 0.5161,
+      "step": 3217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876286876027338e-05,
+      "loss": 0.5734,
+      "step": 3218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876229200911267e-05,
+      "loss": 0.5597,
+      "step": 3219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876171525795196e-05,
+      "loss": 0.5725,
+      "step": 3220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8761138506791247e-05,
+      "loss": 0.6016,
+      "step": 3221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8760561755630535e-05,
+      "loss": 0.626,
+      "step": 3222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8759985004469822e-05,
+      "loss": 0.5861,
+      "step": 3223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875940825330911e-05,
+      "loss": 0.7012,
+      "step": 3224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87588315021484e-05,
+      "loss": 0.6573,
+      "step": 3225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875825475098769e-05,
+      "loss": 0.7959,
+      "step": 3226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8757677999826976e-05,
+      "loss": 0.636,
+      "step": 3227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8757101248666264e-05,
+      "loss": 0.7384,
+      "step": 3228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8756524497505552e-05,
+      "loss": 0.532,
+      "step": 3229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875594774634484e-05,
+      "loss": 0.6409,
+      "step": 3230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875537099518413e-05,
+      "loss": 0.5848,
+      "step": 3231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8754794244023418e-05,
+      "loss": 0.6852,
+      "step": 3232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8754217492862706e-05,
+      "loss": 0.7089,
+      "step": 3233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8753640741701994e-05,
+      "loss": 0.6074,
+      "step": 3234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875306399054128e-05,
+      "loss": 0.7712,
+      "step": 3235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875248723938057e-05,
+      "loss": 0.6504,
+      "step": 3236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.875191048821986e-05,
+      "loss": 0.6223,
+      "step": 3237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8751333737059148e-05,
+      "loss": 0.6668,
+      "step": 3238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8750756985898435e-05,
+      "loss": 0.6143,
+      "step": 3239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8750180234737726e-05,
+      "loss": 0.7055,
+      "step": 3240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874960348357701e-05,
+      "loss": 0.6166,
+      "step": 3241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87490267324163e-05,
+      "loss": 0.604,
+      "step": 3242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874844998125559e-05,
+      "loss": 0.7256,
+      "step": 3243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8747873230094877e-05,
+      "loss": 0.7071,
+      "step": 3244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8747296478934165e-05,
+      "loss": 0.6321,
+      "step": 3245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8746719727773456e-05,
+      "loss": 0.665,
+      "step": 3246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8746142976612744e-05,
+      "loss": 0.6774,
+      "step": 3247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874556622545203e-05,
+      "loss": 0.5984,
+      "step": 3248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874498947429132e-05,
+      "loss": 0.598,
+      "step": 3249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8744412723130607e-05,
+      "loss": 0.594,
+      "step": 3250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8743835971969894e-05,
+      "loss": 0.7347,
+      "step": 3251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8743259220809182e-05,
+      "loss": 0.5039,
+      "step": 3252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8742682469648473e-05,
+      "loss": 0.5399,
+      "step": 3253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874210571848776e-05,
+      "loss": 0.6499,
+      "step": 3254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.874152896732705e-05,
+      "loss": 0.7201,
+      "step": 3255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8740952216166336e-05,
+      "loss": 0.708,
+      "step": 3256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8740375465005624e-05,
+      "loss": 0.7028,
+      "step": 3257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873979871384491e-05,
+      "loss": 0.6882,
+      "step": 3258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8739221962684203e-05,
+      "loss": 0.6124,
+      "step": 3259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873864521152349e-05,
+      "loss": 0.6622,
+      "step": 3260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8738068460362778e-05,
+      "loss": 0.6699,
+      "step": 3261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8737491709202066e-05,
+      "loss": 0.6273,
+      "step": 3262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8736914958041353e-05,
+      "loss": 0.5822,
+      "step": 3263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873633820688064e-05,
+      "loss": 0.6357,
+      "step": 3264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8735761455719932e-05,
+      "loss": 0.5764,
+      "step": 3265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873518470455922e-05,
+      "loss": 0.6028,
+      "step": 3266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8734607953398508e-05,
+      "loss": 0.6755,
+      "step": 3267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8734031202237795e-05,
+      "loss": 0.7585,
+      "step": 3268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8733454451077083e-05,
+      "loss": 0.5995,
+      "step": 3269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873287769991637e-05,
+      "loss": 0.7173,
+      "step": 3270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873230094875566e-05,
+      "loss": 0.591,
+      "step": 3271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.873172419759495e-05,
+      "loss": 0.5164,
+      "step": 3272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8731147446434237e-05,
+      "loss": 0.5181,
+      "step": 3273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8730570695273528e-05,
+      "loss": 0.6275,
+      "step": 3274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8729993944112816e-05,
+      "loss": 0.6035,
+      "step": 3275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8729417192952103e-05,
+      "loss": 0.4889,
+      "step": 3276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.872884044179139e-05,
+      "loss": 0.7191,
+      "step": 3277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.872826369063068e-05,
+      "loss": 0.583,
+      "step": 3278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8727686939469967e-05,
+      "loss": 0.7036,
+      "step": 3279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8727110188309258e-05,
+      "loss": 0.5788,
+      "step": 3280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8726533437148545e-05,
+      "loss": 0.7288,
+      "step": 3281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8725956685987833e-05,
+      "loss": 0.634,
+      "step": 3282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.872537993482712e-05,
+      "loss": 0.6165,
+      "step": 3283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.872480318366641e-05,
+      "loss": 0.5417,
+      "step": 3284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8724226432505696e-05,
+      "loss": 0.6676,
+      "step": 3285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8723649681344984e-05,
+      "loss": 0.6999,
+      "step": 3286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8723072930184275e-05,
+      "loss": 0.6371,
+      "step": 3287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8722496179023562e-05,
+      "loss": 0.846,
+      "step": 3288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.872191942786285e-05,
+      "loss": 0.5891,
+      "step": 3289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8721342676702138e-05,
+      "loss": 0.5445,
+      "step": 3290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8720765925541426e-05,
+      "loss": 0.6579,
+      "step": 3291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8720189174380713e-05,
+      "loss": 0.6827,
+      "step": 3292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8719612423220004e-05,
+      "loss": 0.6932,
+      "step": 3293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8719035672059292e-05,
+      "loss": 0.6484,
+      "step": 3294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.871845892089858e-05,
+      "loss": 0.7862,
+      "step": 3295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8717882169737867e-05,
+      "loss": 0.5656,
+      "step": 3296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8717305418577155e-05,
+      "loss": 0.5333,
+      "step": 3297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8716728667416443e-05,
+      "loss": 0.4743,
+      "step": 3298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8716151916255734e-05,
+      "loss": 0.669,
+      "step": 3299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.871557516509502e-05,
+      "loss": 0.5154,
+      "step": 3300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.871499841393431e-05,
+      "loss": 0.6425,
+      "step": 3301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87144216627736e-05,
+      "loss": 0.6614,
+      "step": 3302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8713844911612885e-05,
+      "loss": 0.6032,
+      "step": 3303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8713268160452172e-05,
+      "loss": 0.7135,
+      "step": 3304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8712691409291463e-05,
+      "loss": 0.6698,
+      "step": 3305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.871211465813075e-05,
+      "loss": 0.6169,
+      "step": 3306
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871153790697004e-05,
+      "loss": 0.6404,
+      "step": 3307
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871096115580933e-05,
+      "loss": 0.6436,
+      "step": 3308
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710384404648617e-05,
+      "loss": 0.6204,
+      "step": 3309
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709807653487905e-05,
+      "loss": 0.6299,
+      "step": 3310
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709230902327193e-05,
+      "loss": 0.6479,
+      "step": 3311
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870865415116648e-05,
+      "loss": 0.4657,
+      "step": 3312
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708077400005768e-05,
+      "loss": 0.6534,
+      "step": 3313
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870750064884506e-05,
+      "loss": 0.6947,
+      "step": 3314
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706923897684347e-05,
+      "loss": 0.5267,
+      "step": 3315
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8706347146523635e-05,
+      "loss": 0.6106,
+      "step": 3316
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8705770395362922e-05,
+      "loss": 0.7166,
+      "step": 3317
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870519364420221e-05,
+      "loss": 0.5436,
+      "step": 3318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704616893041498e-05,
+      "loss": 0.722,
+      "step": 3319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704040141880785e-05,
+      "loss": 0.7044,
+      "step": 3320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703463390720076e-05,
+      "loss": 0.723,
+      "step": 3321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702886639559364e-05,
+      "loss": 0.6124,
+      "step": 3322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702309888398652e-05,
+      "loss": 0.5011,
+      "step": 3323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870173313723794e-05,
+      "loss": 0.5425,
+      "step": 3324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701156386077227e-05,
+      "loss": 0.5983,
+      "step": 3325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700579634916515e-05,
+      "loss": 0.7197,
+      "step": 3326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700002883755806e-05,
+      "loss": 0.6578,
+      "step": 3327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699426132595094e-05,
+      "loss": 0.6456,
+      "step": 3328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869884938143438e-05,
+      "loss": 0.7104,
+      "step": 3329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869827263027367e-05,
+      "loss": 0.5547,
+      "step": 3330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697695879112957e-05,
+      "loss": 0.5545,
+      "step": 3331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697119127952244e-05,
+      "loss": 0.7892,
+      "step": 3332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696542376791535e-05,
+      "loss": 0.6256,
+      "step": 3333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695965625630823e-05,
+      "loss": 0.5891,
+      "step": 3334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869538887447011e-05,
+      "loss": 0.6951,
+      "step": 3335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8694812123309402e-05,
+      "loss": 0.6843,
+      "step": 3336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869423537214869e-05,
+      "loss": 0.7419,
+      "step": 3337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693658620987977e-05,
+      "loss": 0.6194,
+      "step": 3338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693081869827265e-05,
+      "loss": 0.6052,
+      "step": 3339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692505118666553e-05,
+      "loss": 0.621,
+      "step": 3340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869192836750584e-05,
+      "loss": 0.6318,
+      "step": 3341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869135161634513e-05,
+      "loss": 0.6527,
+      "step": 3342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869077486518442e-05,
+      "loss": 0.4954,
+      "step": 3343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690198114023707e-05,
+      "loss": 0.6116,
+      "step": 3344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689621362862994e-05,
+      "loss": 0.7688,
+      "step": 3345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8689044611702282e-05,
+      "loss": 0.5196,
+      "step": 3346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868846786054157e-05,
+      "loss": 0.6879,
+      "step": 3347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868789110938086e-05,
+      "loss": 0.5028,
+      "step": 3348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868731435822015e-05,
+      "loss": 0.6874,
+      "step": 3349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686737607059436e-05,
+      "loss": 0.6282,
+      "step": 3350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686160855898724e-05,
+      "loss": 0.6853,
+      "step": 3351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868558410473801e-05,
+      "loss": 0.7388,
+      "step": 3352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86850073535773e-05,
+      "loss": 0.6652,
+      "step": 3353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8684430602416587e-05,
+      "loss": 0.5962,
+      "step": 3354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683853851255878e-05,
+      "loss": 0.5746,
+      "step": 3355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683277100095166e-05,
+      "loss": 0.5725,
+      "step": 3356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8682700348934453e-05,
+      "loss": 0.6061,
+      "step": 3357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868212359777374e-05,
+      "loss": 0.6468,
+      "step": 3358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868154684661303e-05,
+      "loss": 0.7432,
+      "step": 3359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680970095452316e-05,
+      "loss": 0.7432,
+      "step": 3360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8680393344291608e-05,
+      "loss": 0.5714,
+      "step": 3361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679816593130895e-05,
+      "loss": 0.5998,
+      "step": 3362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679239841970183e-05,
+      "loss": 0.5301,
+      "step": 3363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678663090809474e-05,
+      "loss": 0.6826,
+      "step": 3364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867808633964876e-05,
+      "loss": 0.5473,
+      "step": 3365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677509588488046e-05,
+      "loss": 0.4965,
+      "step": 3366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676932837327337e-05,
+      "loss": 0.6461,
+      "step": 3367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676356086166625e-05,
+      "loss": 0.5969,
+      "step": 3368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675779335005912e-05,
+      "loss": 0.5751,
+      "step": 3369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675202583845203e-05,
+      "loss": 0.5846,
+      "step": 3370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867462583268449e-05,
+      "loss": 0.6359,
+      "step": 3371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867404908152378e-05,
+      "loss": 0.5797,
+      "step": 3372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673472330363067e-05,
+      "loss": 0.6101,
+      "step": 3373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672895579202354e-05,
+      "loss": 0.439,
+      "step": 3374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672318828041642e-05,
+      "loss": 0.6059,
+      "step": 3375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671742076880933e-05,
+      "loss": 0.6474,
+      "step": 3376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867116532572022e-05,
+      "loss": 0.6799,
+      "step": 3377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867058857455951e-05,
+      "loss": 0.5518,
+      "step": 3378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670011823398796e-05,
+      "loss": 0.574,
+      "step": 3379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8669435072238084e-05,
+      "loss": 0.5298,
+      "step": 3380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866885832107737e-05,
+      "loss": 0.7044,
+      "step": 3381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8668281569916662e-05,
+      "loss": 0.6489,
+      "step": 3382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866770481875595e-05,
+      "loss": 0.6832,
+      "step": 3383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667128067595238e-05,
+      "loss": 0.7087,
+      "step": 3384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666551316434526e-05,
+      "loss": 0.5789,
+      "step": 3385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665974565273813e-05,
+      "loss": 0.6595,
+      "step": 3386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86653978141131e-05,
+      "loss": 0.6696,
+      "step": 3387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664821062952392e-05,
+      "loss": 0.629,
+      "step": 3388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866424431179168e-05,
+      "loss": 0.6601,
+      "step": 3389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663667560630967e-05,
+      "loss": 0.5401,
+      "step": 3390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663090809470255e-05,
+      "loss": 0.5438,
+      "step": 3391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662514058309543e-05,
+      "loss": 0.7213,
+      "step": 3392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866193730714883e-05,
+      "loss": 0.5901,
+      "step": 3393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661360555988118e-05,
+      "loss": 0.6434,
+      "step": 3394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866078380482741e-05,
+      "loss": 0.6846,
+      "step": 3395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660207053666697e-05,
+      "loss": 0.6367,
+      "step": 3396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659630302505985e-05,
+      "loss": 0.6179,
+      "step": 3397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659053551345276e-05,
+      "loss": 0.6533,
+      "step": 3398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658476800184563e-05,
+      "loss": 0.6661,
+      "step": 3399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865790004902385e-05,
+      "loss": 0.6397,
+      "step": 3400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865732329786314e-05,
+      "loss": 0.6005,
+      "step": 3401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656746546702426e-05,
+      "loss": 0.5587,
+      "step": 3402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656169795541714e-05,
+      "loss": 0.5491,
+      "step": 3403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655593044381005e-05,
+      "loss": 0.6686,
+      "step": 3404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655016293220293e-05,
+      "loss": 0.4671,
+      "step": 3405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865443954205958e-05,
+      "loss": 0.6323,
+      "step": 3406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653862790898868e-05,
+      "loss": 0.5164,
+      "step": 3407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653286039738156e-05,
+      "loss": 0.6193,
+      "step": 3408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652709288577443e-05,
+      "loss": 0.5238,
+      "step": 3409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652132537416735e-05,
+      "loss": 0.5914,
+      "step": 3410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651555786256022e-05,
+      "loss": 0.6504,
+      "step": 3411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865097903509531e-05,
+      "loss": 0.7466,
+      "step": 3412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650402283934598e-05,
+      "loss": 0.5879,
+      "step": 3413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649825532773885e-05,
+      "loss": 0.6205,
+      "step": 3414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649248781613173e-05,
+      "loss": 0.4801,
+      "step": 3415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648672030452464e-05,
+      "loss": 0.6613,
+      "step": 3416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648095279291752e-05,
+      "loss": 0.6278,
+      "step": 3417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864751852813104e-05,
+      "loss": 0.5743,
+      "step": 3418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646941776970327e-05,
+      "loss": 0.4381,
+      "step": 3419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8646365025809615e-05,
+      "loss": 0.582,
+      "step": 3420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645788274648902e-05,
+      "loss": 0.6525,
+      "step": 3421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645211523488194e-05,
+      "loss": 0.8564,
+      "step": 3422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864463477232748e-05,
+      "loss": 0.6075,
+      "step": 3423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864405802116677e-05,
+      "loss": 0.6191,
+      "step": 3424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8643481270006057e-05,
+      "loss": 0.64,
+      "step": 3425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642904518845348e-05,
+      "loss": 0.665,
+      "step": 3426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642327767684635e-05,
+      "loss": 0.6007,
+      "step": 3427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864175101652392e-05,
+      "loss": 0.6636,
+      "step": 3428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864117426536321e-05,
+      "loss": 0.5001,
+      "step": 3429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86405975142025e-05,
+      "loss": 0.6906,
+      "step": 3430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640020763041786e-05,
+      "loss": 0.6177,
+      "step": 3431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639444011881077e-05,
+      "loss": 0.626,
+      "step": 3432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638867260720365e-05,
+      "loss": 0.669,
+      "step": 3433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638290509559653e-05,
+      "loss": 0.6639,
+      "step": 3434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863771375839894e-05,
+      "loss": 0.602,
+      "step": 3435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637137007238228e-05,
+      "loss": 0.4091,
+      "step": 3436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636560256077516e-05,
+      "loss": 0.6056,
+      "step": 3437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635983504916807e-05,
+      "loss": 0.6596,
+      "step": 3438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8635406753756094e-05,
+      "loss": 0.5601,
+      "step": 3439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634830002595382e-05,
+      "loss": 0.5624,
+      "step": 3440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863425325143467e-05,
+      "loss": 0.6013,
+      "step": 3441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633676500273957e-05,
+      "loss": 0.5229,
+      "step": 3442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633099749113245e-05,
+      "loss": 0.6871,
+      "step": 3443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632522997952536e-05,
+      "loss": 0.5088,
+      "step": 3444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8631946246791824e-05,
+      "loss": 0.7234,
+      "step": 3445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863136949563111e-05,
+      "loss": 0.578,
+      "step": 3446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86307927444704e-05,
+      "loss": 0.5901,
+      "step": 3447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630215993309687e-05,
+      "loss": 0.5847,
+      "step": 3448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629639242148975e-05,
+      "loss": 0.6869,
+      "step": 3449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629062490988266e-05,
+      "loss": 0.6511,
+      "step": 3450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628485739827553e-05,
+      "loss": 0.6432,
+      "step": 3451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862790898866684e-05,
+      "loss": 0.4555,
+      "step": 3452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8627332237506132e-05,
+      "loss": 0.619,
+      "step": 3453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626755486345416e-05,
+      "loss": 0.5528,
+      "step": 3454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626178735184704e-05,
+      "loss": 0.706,
+      "step": 3455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625601984023995e-05,
+      "loss": 0.5525,
+      "step": 3456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625025232863283e-05,
+      "loss": 0.621,
+      "step": 3457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862444848170257e-05,
+      "loss": 0.6562,
+      "step": 3458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8623871730541858e-05,
+      "loss": 0.5887,
+      "step": 3459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862329497938115e-05,
+      "loss": 0.3786,
+      "step": 3460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622718228220437e-05,
+      "loss": 0.5841,
+      "step": 3461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622141477059725e-05,
+      "loss": 0.5221,
+      "step": 3462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621564725899012e-05,
+      "loss": 0.5773,
+      "step": 3463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86209879747383e-05,
+      "loss": 0.556,
+      "step": 3464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620411223577588e-05,
+      "loss": 0.5447,
+      "step": 3465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861983447241688e-05,
+      "loss": 0.5981,
+      "step": 3466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8619257721256167e-05,
+      "loss": 0.4172,
+      "step": 3467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618680970095454e-05,
+      "loss": 0.5922,
+      "step": 3468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618104218934742e-05,
+      "loss": 0.4848,
+      "step": 3469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861752746777403e-05,
+      "loss": 0.715,
+      "step": 3470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616950716613317e-05,
+      "loss": 0.5271,
+      "step": 3471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861637396545261e-05,
+      "loss": 0.5807,
+      "step": 3472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615797214291896e-05,
+      "loss": 0.5132,
+      "step": 3473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8615220463131184e-05,
+      "loss": 0.6739,
+      "step": 3474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861464371197047e-05,
+      "loss": 0.613,
+      "step": 3475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861406696080976e-05,
+      "loss": 0.5805,
+      "step": 3476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8613490209649047e-05,
+      "loss": 0.6065,
+      "step": 3477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612913458488338e-05,
+      "loss": 0.7024,
+      "step": 3478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612336707327626e-05,
+      "loss": 0.6913,
+      "step": 3479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8611759956166913e-05,
+      "loss": 0.5273,
+      "step": 3480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86111832050062e-05,
+      "loss": 0.5579,
+      "step": 3481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861060645384549e-05,
+      "loss": 0.707,
+      "step": 3482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610029702684776e-05,
+      "loss": 0.6344,
+      "step": 3483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609452951524067e-05,
+      "loss": 0.6081,
+      "step": 3484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8608876200363355e-05,
+      "loss": 0.5354,
+      "step": 3485
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8608299449202643e-05,
+      "loss": 0.6281,
+      "step": 3486
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8607722698041934e-05,
+      "loss": 0.6056,
+      "step": 3487
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860714594688122e-05,
+      "loss": 0.6813,
+      "step": 3488
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860656919572051e-05,
+      "loss": 0.6661,
+      "step": 3489
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605992444559797e-05,
+      "loss": 0.5699,
+      "step": 3490
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605415693399085e-05,
+      "loss": 0.7285,
+      "step": 3491
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604838942238372e-05,
+      "loss": 0.5202,
+      "step": 3492
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860426219107766e-05,
+      "loss": 0.573,
+      "step": 3493
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860368543991695e-05,
+      "loss": 0.5803,
+      "step": 3494
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860310868875624e-05,
+      "loss": 0.6551,
+      "step": 3495
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602531937595526e-05,
+      "loss": 0.6332,
+      "step": 3496
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8601955186434814e-05,
+      "loss": 0.5566,
+      "step": 3497
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.86013784352741e-05,
+      "loss": 0.7223,
+      "step": 3498
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860080168411339e-05,
+      "loss": 0.5265,
+      "step": 3499
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860022493295268e-05,
+      "loss": 0.6081,
+      "step": 3500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599648181791968e-05,
+      "loss": 0.6088,
+      "step": 3501
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8599071430631256e-05,
+      "loss": 0.6584,
+      "step": 3502
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598494679470543e-05,
+      "loss": 0.5985,
+      "step": 3503
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859791792830983e-05,
+      "loss": 0.5732,
+      "step": 3504
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859734117714912e-05,
+      "loss": 0.6591,
+      "step": 3505
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859676442598841e-05,
+      "loss": 0.5277,
+      "step": 3506
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596187674827698e-05,
+      "loss": 0.6432,
+      "step": 3507
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595610923666985e-05,
+      "loss": 0.6145,
+      "step": 3508
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595034172506273e-05,
+      "loss": 0.5777,
+      "step": 3509
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859445742134556e-05,
+      "loss": 0.7232,
+      "step": 3510
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859388067018485e-05,
+      "loss": 0.5478,
+      "step": 3511
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859330391902414e-05,
+      "loss": 0.5462,
+      "step": 3512
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8592727167863427e-05,
+      "loss": 0.5619,
+      "step": 3513
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8592150416702715e-05,
+      "loss": 0.5892,
+      "step": 3514
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591573665542006e-05,
+      "loss": 0.4929,
+      "step": 3515
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859099691438129e-05,
+      "loss": 0.7269,
+      "step": 3516
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590420163220578e-05,
+      "loss": 0.5929,
+      "step": 3517
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858984341205987e-05,
+      "loss": 0.6328,
+      "step": 3518
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589266660899157e-05,
+      "loss": 0.4934,
+      "step": 3519
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588689909738444e-05,
+      "loss": 0.6,
+      "step": 3520
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8588113158577735e-05,
+      "loss": 0.5696,
+      "step": 3521
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8587536407417023e-05,
+      "loss": 0.647,
+      "step": 3522
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858695965625631e-05,
+      "loss": 0.6299,
+      "step": 3523
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85863829050956e-05,
+      "loss": 0.5748,
+      "step": 3524
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585806153934886e-05,
+      "loss": 0.5759,
+      "step": 3525
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585229402774174e-05,
+      "loss": 0.6344,
+      "step": 3526
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584652651613465e-05,
+      "loss": 0.5971,
+      "step": 3527
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584075900452753e-05,
+      "loss": 0.5251,
+      "step": 3528
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858349914929204e-05,
+      "loss": 0.5798,
+      "step": 3529
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582922398131328e-05,
+      "loss": 0.6237,
+      "step": 3530
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582345646970616e-05,
+      "loss": 0.7331,
+      "step": 3531
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581768895809903e-05,
+      "loss": 0.5492,
+      "step": 3532
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858119214464919e-05,
+      "loss": 0.6839,
+      "step": 3533
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580615393488482e-05,
+      "loss": 0.5079,
+      "step": 3534
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858003864232777e-05,
+      "loss": 0.6942,
+      "step": 3535
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579461891167057e-05,
+      "loss": 0.6195,
+      "step": 3536
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578885140006345e-05,
+      "loss": 0.7189,
+      "step": 3537
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8578308388845633e-05,
+      "loss": 0.6509,
+      "step": 3538
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857773163768492e-05,
+      "loss": 0.6107,
+      "step": 3539
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857715488652421e-05,
+      "loss": 0.5986,
+      "step": 3540
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85765781353635e-05,
+      "loss": 0.5527,
+      "step": 3541
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576001384202787e-05,
+      "loss": 0.6834,
+      "step": 3542
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575424633042075e-05,
+      "loss": 0.5681,
+      "step": 3543
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574847881881362e-05,
+      "loss": 0.6394,
+      "step": 3544
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857427113072065e-05,
+      "loss": 0.7012,
+      "step": 3545
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857369437955994e-05,
+      "loss": 0.6986,
+      "step": 3546
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857311762839923e-05,
+      "loss": 0.4581,
+      "step": 3547
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572540877238516e-05,
+      "loss": 0.6288,
+      "step": 3548
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571964126077808e-05,
+      "loss": 0.6664,
+      "step": 3549
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571387374917095e-05,
+      "loss": 0.7123,
+      "step": 3550
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8570810623756383e-05,
+      "loss": 0.6456,
+      "step": 3551
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857023387259567e-05,
+      "loss": 0.5669,
+      "step": 3552
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569657121434958e-05,
+      "loss": 0.5431,
+      "step": 3553
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569080370274246e-05,
+      "loss": 0.5196,
+      "step": 3554
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568503619113537e-05,
+      "loss": 0.6002,
+      "step": 3555
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567926867952825e-05,
+      "loss": 0.6174,
+      "step": 3556
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567350116792112e-05,
+      "loss": 0.6701,
+      "step": 3557
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85667733656314e-05,
+      "loss": 0.6393,
+      "step": 3558
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566196614470688e-05,
+      "loss": 0.5571,
+      "step": 3559
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565619863309975e-05,
+      "loss": 0.597,
+      "step": 3560
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565043112149267e-05,
+      "loss": 0.6095,
+      "step": 3561
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564466360988554e-05,
+      "loss": 0.6332,
+      "step": 3562
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8563889609827842e-05,
+      "loss": 0.7151,
+      "step": 3563
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856331285866713e-05,
+      "loss": 0.4591,
+      "step": 3564
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562736107506417e-05,
+      "loss": 0.5336,
+      "step": 3565
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562159356345705e-05,
+      "loss": 0.5119,
+      "step": 3566
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561582605184993e-05,
+      "loss": 0.5377,
+      "step": 3567
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561005854024284e-05,
+      "loss": 0.6433,
+      "step": 3568
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856042910286357e-05,
+      "loss": 0.7758,
+      "step": 3569
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855985235170286e-05,
+      "loss": 0.516,
+      "step": 3570
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559275600542147e-05,
+      "loss": 0.5171,
+      "step": 3571
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558698849381434e-05,
+      "loss": 0.6544,
+      "step": 3572
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558122098220722e-05,
+      "loss": 0.5982,
+      "step": 3573
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557545347060013e-05,
+      "loss": 0.5636,
+      "step": 3574
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85569685958993e-05,
+      "loss": 0.5375,
+      "step": 3575
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855639184473859e-05,
+      "loss": 0.6349,
+      "step": 3576
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855581509357788e-05,
+      "loss": 0.6238,
+      "step": 3577
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555238342417164e-05,
+      "loss": 0.7158,
+      "step": 3578
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855466159125645e-05,
+      "loss": 0.5326,
+      "step": 3579
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554084840095743e-05,
+      "loss": 0.688,
+      "step": 3580
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855350808893503e-05,
+      "loss": 0.586,
+      "step": 3581
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8552931337774318e-05,
+      "loss": 0.5095,
+      "step": 3582
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855235458661361e-05,
+      "loss": 0.5635,
+      "step": 3583
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551777835452897e-05,
+      "loss": 0.5731,
+      "step": 3584
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551201084292184e-05,
+      "loss": 0.5637,
+      "step": 3585
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550624333131472e-05,
+      "loss": 0.5691,
+      "step": 3586
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855004758197076e-05,
+      "loss": 0.628,
+      "step": 3587
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549470830810048e-05,
+      "loss": 0.4962,
+      "step": 3588
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854889407964934e-05,
+      "loss": 0.7942,
+      "step": 3589
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8548317328488626e-05,
+      "loss": 0.6844,
+      "step": 3590
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547740577327914e-05,
+      "loss": 0.5333,
+      "step": 3591
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85471638261672e-05,
+      "loss": 0.6904,
+      "step": 3592
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854658707500649e-05,
+      "loss": 0.7404,
+      "step": 3593
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546010323845777e-05,
+      "loss": 0.6968,
+      "step": 3594
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8545433572685068e-05,
+      "loss": 0.7084,
+      "step": 3595
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544856821524356e-05,
+      "loss": 0.5479,
+      "step": 3596
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544280070363643e-05,
+      "loss": 0.6842,
+      "step": 3597
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854370331920293e-05,
+      "loss": 0.5968,
+      "step": 3598
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854312656804222e-05,
+      "loss": 0.7168,
+      "step": 3599
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8542549816881507e-05,
+      "loss": 0.6328,
+      "step": 3600
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541973065720794e-05,
+      "loss": 0.6123,
+      "step": 3601
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541396314560085e-05,
+      "loss": 0.6161,
+      "step": 3602
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540819563399373e-05,
+      "loss": 0.5198,
+      "step": 3603
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854024281223866e-05,
+      "loss": 0.6672,
+      "step": 3604
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853966606107795e-05,
+      "loss": 0.4945,
+      "step": 3605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539089309917236e-05,
+      "loss": 0.5147,
+      "step": 3606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538512558756524e-05,
+      "loss": 0.6234,
+      "step": 3607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537935807595815e-05,
+      "loss": 0.6297,
+      "step": 3608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537359056435102e-05,
+      "loss": 0.5226,
+      "step": 3609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853678230527439e-05,
+      "loss": 0.6274,
+      "step": 3610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853620555411368e-05,
+      "loss": 0.7051,
+      "step": 3611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853562880295297e-05,
+      "loss": 0.5001,
+      "step": 3612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8535052051792257e-05,
+      "loss": 0.6027,
+      "step": 3613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8534475300631544e-05,
+      "loss": 0.6656,
+      "step": 3614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533898549470832e-05,
+      "loss": 0.6321,
+      "step": 3615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853332179831012e-05,
+      "loss": 0.6015,
+      "step": 3616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853274504714941e-05,
+      "loss": 0.5334,
+      "step": 3617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85321682959887e-05,
+      "loss": 0.4844,
+      "step": 3618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531591544827986e-05,
+      "loss": 0.6754,
+      "step": 3619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531014793667274e-05,
+      "loss": 0.6715,
+      "step": 3620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853043804250656e-05,
+      "loss": 0.601,
+      "step": 3621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852986129134585e-05,
+      "loss": 0.4858,
+      "step": 3622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852928454018514e-05,
+      "loss": 0.4438,
+      "step": 3623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528707789024428e-05,
+      "loss": 0.5956,
+      "step": 3624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528131037863716e-05,
+      "loss": 0.641,
+      "step": 3625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527554286703003e-05,
+      "loss": 0.5951,
+      "step": 3626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852697753554229e-05,
+      "loss": 0.6462,
+      "step": 3627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852640078438158e-05,
+      "loss": 0.5836,
+      "step": 3628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852582403322087e-05,
+      "loss": 0.5658,
+      "step": 3629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525247282060157e-05,
+      "loss": 0.6888,
+      "step": 3630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524670530899445e-05,
+      "loss": 0.4982,
+      "step": 3631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524093779738733e-05,
+      "loss": 0.5479,
+      "step": 3632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852351702857802e-05,
+      "loss": 0.6697,
+      "step": 3633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522940277417308e-05,
+      "loss": 0.6813,
+      "step": 3634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522363526256596e-05,
+      "loss": 0.5599,
+      "step": 3635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521786775095887e-05,
+      "loss": 0.4948,
+      "step": 3636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521210023935175e-05,
+      "loss": 0.5355,
+      "step": 3637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520633272774462e-05,
+      "loss": 0.4405,
+      "step": 3638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520056521613753e-05,
+      "loss": 0.5883,
+      "step": 3639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519479770453038e-05,
+      "loss": 0.6696,
+      "step": 3640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518903019292325e-05,
+      "loss": 0.5742,
+      "step": 3641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8518326268131616e-05,
+      "loss": 0.5674,
+      "step": 3642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517749516970904e-05,
+      "loss": 0.5825,
+      "step": 3643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517172765810192e-05,
+      "loss": 0.6016,
+      "step": 3644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8516596014649483e-05,
+      "loss": 0.4837,
+      "step": 3645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851601926348877e-05,
+      "loss": 0.6183,
+      "step": 3646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515442512328058e-05,
+      "loss": 0.5903,
+      "step": 3647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514865761167346e-05,
+      "loss": 0.5815,
+      "step": 3648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514289010006634e-05,
+      "loss": 0.5112,
+      "step": 3649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851371225884592e-05,
+      "loss": 0.5198,
+      "step": 3650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513135507685212e-05,
+      "loss": 0.6188,
+      "step": 3651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85125587565245e-05,
+      "loss": 0.6251,
+      "step": 3652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511982005363788e-05,
+      "loss": 0.5944,
+      "step": 3653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511405254203075e-05,
+      "loss": 0.5797,
+      "step": 3654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510828503042363e-05,
+      "loss": 0.4918,
+      "step": 3655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851025175188165e-05,
+      "loss": 0.5122,
+      "step": 3656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509675000720942e-05,
+      "loss": 0.61,
+      "step": 3657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850909824956023e-05,
+      "loss": 0.5949,
+      "step": 3658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508521498399517e-05,
+      "loss": 0.5483,
+      "step": 3659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507944747238805e-05,
+      "loss": 0.5596,
+      "step": 3660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507367996078093e-05,
+      "loss": 0.5803,
+      "step": 3661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850679124491738e-05,
+      "loss": 0.6184,
+      "step": 3662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850621449375667e-05,
+      "loss": 0.6917,
+      "step": 3663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850563774259596e-05,
+      "loss": 0.5908,
+      "step": 3664
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8505060991435247e-05,
+      "loss": 0.6868,
+      "step": 3665
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8504484240274538e-05,
+      "loss": 0.6407,
+      "step": 3666
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8503907489113822e-05,
+      "loss": 0.627,
+      "step": 3667
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.850333073795311e-05,
+      "loss": 0.6468,
+      "step": 3668
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8502753986792397e-05,
+      "loss": 0.604,
+      "step": 3669
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.850217723563169e-05,
+      "loss": 0.764,
+      "step": 3670
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8501600484470976e-05,
+      "loss": 0.7124,
+      "step": 3671
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8501023733310264e-05,
+      "loss": 0.5723,
+      "step": 3672
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8500446982149555e-05,
+      "loss": 0.6986,
+      "step": 3673
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8499870230988843e-05,
+      "loss": 0.6801,
+      "step": 3674
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.849929347982813e-05,
+      "loss": 0.6613,
+      "step": 3675
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8498716728667418e-05,
+      "loss": 0.6356,
+      "step": 3676
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8498139977506706e-05,
+      "loss": 0.5976,
+      "step": 3677
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8497563226345993e-05,
+      "loss": 0.5373,
+      "step": 3678
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8496986475185284e-05,
+      "loss": 0.5248,
+      "step": 3679
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8496409724024572e-05,
+      "loss": 0.7032,
+      "step": 3680
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.849583297286386e-05,
+      "loss": 0.6642,
+      "step": 3681
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8495256221703148e-05,
+      "loss": 0.7077,
+      "step": 3682
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8494679470542435e-05,
+      "loss": 0.6289,
+      "step": 3683
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8494102719381723e-05,
+      "loss": 0.4665,
+      "step": 3684
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8493525968221014e-05,
+      "loss": 0.5949,
+      "step": 3685
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84929492170603e-05,
+      "loss": 0.7516,
+      "step": 3686
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.849237246589959e-05,
+      "loss": 0.7766,
+      "step": 3687
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8491795714738877e-05,
+      "loss": 0.7255,
+      "step": 3688
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8491218963578165e-05,
+      "loss": 0.6128,
+      "step": 3689
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8490642212417452e-05,
+      "loss": 0.6641,
+      "step": 3690
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8490065461256743e-05,
+      "loss": 0.5462,
+      "step": 3691
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848948871009603e-05,
+      "loss": 0.5902,
+      "step": 3692
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848891195893532e-05,
+      "loss": 0.5518,
+      "step": 3693
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8488335207774607e-05,
+      "loss": 0.5513,
+      "step": 3694
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8487758456613894e-05,
+      "loss": 0.6128,
+      "step": 3695
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8487181705453182e-05,
+      "loss": 0.5363,
+      "step": 3696
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8486604954292473e-05,
+      "loss": 0.615,
+      "step": 3697
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848602820313176e-05,
+      "loss": 0.5383,
+      "step": 3698
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848545145197105e-05,
+      "loss": 0.6685,
+      "step": 3699
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848487470081034e-05,
+      "loss": 0.6534,
+      "step": 3700
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8484297949649627e-05,
+      "loss": 0.5246,
+      "step": 3701
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848372119848891e-05,
+      "loss": 0.5188,
+      "step": 3702
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84831444473282e-05,
+      "loss": 0.6495,
+      "step": 3703
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.848256769616749e-05,
+      "loss": 0.6518,
+      "step": 3704
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8481990945006778e-05,
+      "loss": 0.6575,
+      "step": 3705
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8481414193846066e-05,
+      "loss": 0.466,
+      "step": 3706
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8480837442685357e-05,
+      "loss": 0.591,
+      "step": 3707
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8480260691524644e-05,
+      "loss": 0.7354,
+      "step": 3708
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8479683940363932e-05,
+      "loss": 0.5646,
+      "step": 3709
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.847910718920322e-05,
+      "loss": 0.6009,
+      "step": 3710
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8478530438042507e-05,
+      "loss": 0.6223,
+      "step": 3711
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8477953686881795e-05,
+      "loss": 0.6137,
+      "step": 3712
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8477376935721086e-05,
+      "loss": 0.6856,
+      "step": 3713
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8476800184560374e-05,
+      "loss": 0.5461,
+      "step": 3714
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.847622343339966e-05,
+      "loss": 0.6031,
+      "step": 3715
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.847564668223895e-05,
+      "loss": 0.658,
+      "step": 3716
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8475069931078237e-05,
+      "loss": 0.6288,
+      "step": 3717
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8474493179917525e-05,
+      "loss": 0.5521,
+      "step": 3718
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8473916428756816e-05,
+      "loss": 0.5786,
+      "step": 3719
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8473339677596103e-05,
+      "loss": 0.641,
+      "step": 3720
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.847276292643539e-05,
+      "loss": 0.5392,
+      "step": 3721
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.847218617527468e-05,
+      "loss": 0.6516,
+      "step": 3722
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8471609424113966e-05,
+      "loss": 0.496,
+      "step": 3723
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8471032672953254e-05,
+      "loss": 0.7295,
+      "step": 3724
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8470455921792545e-05,
+      "loss": 0.5614,
+      "step": 3725
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8469879170631833e-05,
+      "loss": 0.6629,
+      "step": 3726
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846930241947112e-05,
+      "loss": 0.6694,
+      "step": 3727
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846872566831041e-05,
+      "loss": 0.4588,
+      "step": 3728
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8468148917149696e-05,
+      "loss": 0.6188,
+      "step": 3729
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8467572165988984e-05,
+      "loss": 0.5729,
+      "step": 3730
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8466995414828275e-05,
+      "loss": 0.6095,
+      "step": 3731
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8466418663667562e-05,
+      "loss": 0.5647,
+      "step": 3732
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846584191250685e-05,
+      "loss": 0.7288,
+      "step": 3733
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846526516134614e-05,
+      "loss": 0.7165,
+      "step": 3734
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846468841018543e-05,
+      "loss": 0.5458,
+      "step": 3735
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8464111659024716e-05,
+      "loss": 0.5869,
+      "step": 3736
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8463534907864004e-05,
+      "loss": 0.6903,
+      "step": 3737
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8462958156703292e-05,
+      "loss": 0.5085,
+      "step": 3738
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.846238140554258e-05,
+      "loss": 0.5708,
+      "step": 3739
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8461804654381867e-05,
+      "loss": 0.5682,
+      "step": 3740
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8461227903221158e-05,
+      "loss": 0.6518,
+      "step": 3741
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8460651152060446e-05,
+      "loss": 0.7248,
+      "step": 3742
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8460074400899734e-05,
+      "loss": 0.714,
+      "step": 3743
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.845949764973902e-05,
+      "loss": 0.6715,
+      "step": 3744
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.845892089857831e-05,
+      "loss": 0.5546,
+      "step": 3745
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8458344147417597e-05,
+      "loss": 0.4528,
+      "step": 3746
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8457767396256888e-05,
+      "loss": 0.5793,
+      "step": 3747
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8457190645096175e-05,
+      "loss": 0.4984,
+      "step": 3748
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8456613893935463e-05,
+      "loss": 0.5745,
+      "step": 3749
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.845603714277475e-05,
+      "loss": 0.7648,
+      "step": 3750
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.845546039161404e-05,
+      "loss": 0.6433,
+      "step": 3751
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8454883640453326e-05,
+      "loss": 0.5351,
+      "step": 3752
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8454306889292617e-05,
+      "loss": 0.8205,
+      "step": 3753
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8453730138131905e-05,
+      "loss": 0.6564,
+      "step": 3754
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8453153386971193e-05,
+      "loss": 0.5478,
+      "step": 3755
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.845257663581048e-05,
+      "loss": 0.6761,
+      "step": 3756
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8451999884649768e-05,
+      "loss": 0.545,
+      "step": 3757
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8451423133489056e-05,
+      "loss": 0.5209,
+      "step": 3758
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8450846382328347e-05,
+      "loss": 0.5869,
+      "step": 3759
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8450269631167634e-05,
+      "loss": 0.5416,
+      "step": 3760
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8449692880006922e-05,
+      "loss": 0.5592,
+      "step": 3761
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8449116128846213e-05,
+      "loss": 0.6739,
+      "step": 3762
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84485393776855e-05,
+      "loss": 0.6395,
+      "step": 3763
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844796262652479e-05,
+      "loss": 0.6615,
+      "step": 3764
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8447385875364076e-05,
+      "loss": 0.6161,
+      "step": 3765
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8446809124203364e-05,
+      "loss": 0.5122,
+      "step": 3766
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844623237304265e-05,
+      "loss": 0.6631,
+      "step": 3767
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8445655621881943e-05,
+      "loss": 0.5464,
+      "step": 3768
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844507887072123e-05,
+      "loss": 0.7147,
+      "step": 3769
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8444502119560518e-05,
+      "loss": 0.7054,
+      "step": 3770
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8443925368399806e-05,
+      "loss": 0.5754,
+      "step": 3771
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8443348617239093e-05,
+      "loss": 0.5071,
+      "step": 3772
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844277186607838e-05,
+      "loss": 0.5771,
+      "step": 3773
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844219511491767e-05,
+      "loss": 0.4889,
+      "step": 3774
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.844161836375696e-05,
+      "loss": 0.6106,
+      "step": 3775
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8441041612596248e-05,
+      "loss": 0.5971,
+      "step": 3776
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8440464861435535e-05,
+      "loss": 0.7031,
+      "step": 3777
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8439888110274823e-05,
+      "loss": 0.5684,
+      "step": 3778
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843931135911411e-05,
+      "loss": 0.5842,
+      "step": 3779
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8438734607953398e-05,
+      "loss": 0.6843,
+      "step": 3780
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843815785679269e-05,
+      "loss": 0.5061,
+      "step": 3781
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8437581105631977e-05,
+      "loss": 0.5785,
+      "step": 3782
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8437004354471265e-05,
+      "loss": 0.4489,
+      "step": 3783
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8436427603310552e-05,
+      "loss": 0.5686,
+      "step": 3784
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843585085214984e-05,
+      "loss": 0.5709,
+      "step": 3785
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8435274100989128e-05,
+      "loss": 0.5795,
+      "step": 3786
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843469734982842e-05,
+      "loss": 0.7253,
+      "step": 3787
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434120598667707e-05,
+      "loss": 0.6259,
+      "step": 3788
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433543847506994e-05,
+      "loss": 0.509,
+      "step": 3789
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432967096346285e-05,
+      "loss": 0.5472,
+      "step": 3790
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843239034518557e-05,
+      "loss": 0.5831,
+      "step": 3791
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8431813594024857e-05,
+      "loss": 0.5357,
+      "step": 3792
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843123684286415e-05,
+      "loss": 0.7032,
+      "step": 3793
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430660091703436e-05,
+      "loss": 0.4919,
+      "step": 3794
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430083340542724e-05,
+      "loss": 0.6547,
+      "step": 3795
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8429506589382015e-05,
+      "loss": 0.4814,
+      "step": 3796
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428929838221302e-05,
+      "loss": 0.7338,
+      "step": 3797
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842835308706059e-05,
+      "loss": 0.6175,
+      "step": 3798
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427776335899878e-05,
+      "loss": 0.6067,
+      "step": 3799
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427199584739166e-05,
+      "loss": 0.6637,
+      "step": 3800
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426622833578453e-05,
+      "loss": 0.5141,
+      "step": 3801
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426046082417744e-05,
+      "loss": 0.5991,
+      "step": 3802
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425469331257032e-05,
+      "loss": 0.6335,
+      "step": 3803
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842489258009632e-05,
+      "loss": 0.5119,
+      "step": 3804
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8424315828935607e-05,
+      "loss": 0.5509,
+      "step": 3805
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423739077774895e-05,
+      "loss": 0.6407,
+      "step": 3806
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423162326614183e-05,
+      "loss": 0.398,
+      "step": 3807
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842258557545347e-05,
+      "loss": 0.7324,
+      "step": 3808
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842200882429276e-05,
+      "loss": 0.6481,
+      "step": 3809
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842143207313205e-05,
+      "loss": 0.5967,
+      "step": 3810
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420855321971337e-05,
+      "loss": 0.5808,
+      "step": 3811
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420278570810625e-05,
+      "loss": 0.7102,
+      "step": 3812
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419701819649912e-05,
+      "loss": 0.6993,
+      "step": 3813
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84191250684892e-05,
+      "loss": 0.5798,
+      "step": 3814
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841854831732849e-05,
+      "loss": 0.6481,
+      "step": 3815
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841797156616778e-05,
+      "loss": 0.6087,
+      "step": 3816
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417394815007066e-05,
+      "loss": 0.5724,
+      "step": 3817
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416818063846354e-05,
+      "loss": 0.5248,
+      "step": 3818
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8416241312685642e-05,
+      "loss": 0.5949,
+      "step": 3819
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841566456152493e-05,
+      "loss": 0.6887,
+      "step": 3820
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841508781036422e-05,
+      "loss": 0.5007,
+      "step": 3821
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414511059203508e-05,
+      "loss": 0.6567,
+      "step": 3822
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413934308042796e-05,
+      "loss": 0.6322,
+      "step": 3823
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413357556882087e-05,
+      "loss": 0.4539,
+      "step": 3824
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412780805721375e-05,
+      "loss": 0.6368,
+      "step": 3825
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412204054560662e-05,
+      "loss": 0.803,
+      "step": 3826
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841162730339995e-05,
+      "loss": 0.6055,
+      "step": 3827
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411050552239238e-05,
+      "loss": 0.5228,
+      "step": 3828
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8410473801078525e-05,
+      "loss": 0.57,
+      "step": 3829
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409897049917816e-05,
+      "loss": 0.6732,
+      "step": 3830
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409320298757104e-05,
+      "loss": 0.6436,
+      "step": 3831
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408743547596392e-05,
+      "loss": 0.5036,
+      "step": 3832
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840816679643568e-05,
+      "loss": 0.627,
+      "step": 3833
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407590045274967e-05,
+      "loss": 0.6628,
+      "step": 3834
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407013294114255e-05,
+      "loss": 0.5687,
+      "step": 3835
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406436542953546e-05,
+      "loss": 0.5486,
+      "step": 3836
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405859791792834e-05,
+      "loss": 0.6593,
+      "step": 3837
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840528304063212e-05,
+      "loss": 0.6643,
+      "step": 3838
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840470628947141e-05,
+      "loss": 0.5163,
+      "step": 3839
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404129538310697e-05,
+      "loss": 0.6292,
+      "step": 3840
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403552787149984e-05,
+      "loss": 0.6245,
+      "step": 3841
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402976035989272e-05,
+      "loss": 0.742,
+      "step": 3842
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402399284828563e-05,
+      "loss": 0.4985,
+      "step": 3843
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.840182253366785e-05,
+      "loss": 0.5707,
+      "step": 3844
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.840124578250714e-05,
+      "loss": 0.553,
+      "step": 3845
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8400669031346426e-05,
+      "loss": 0.5121,
+      "step": 3846
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8400092280185714e-05,
+      "loss": 0.6087,
+      "step": 3847
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8399515529025e-05,
+      "loss": 0.6822,
+      "step": 3848
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8398938777864293e-05,
+      "loss": 0.5615,
+      "step": 3849
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839836202670358e-05,
+      "loss": 0.4275,
+      "step": 3850
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8397785275542868e-05,
+      "loss": 0.6005,
+      "step": 3851
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839720852438216e-05,
+      "loss": 0.5709,
+      "step": 3852
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8396631773221443e-05,
+      "loss": 0.4516,
+      "step": 3853
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839605502206073e-05,
+      "loss": 0.5451,
+      "step": 3854
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8395478270900022e-05,
+      "loss": 0.4892,
+      "step": 3855
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839490151973931e-05,
+      "loss": 0.6478,
+      "step": 3856
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8394324768578597e-05,
+      "loss": 0.5081,
+      "step": 3857
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839374801741789e-05,
+      "loss": 0.5284,
+      "step": 3858
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8393171266257176e-05,
+      "loss": 0.6338,
+      "step": 3859
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8392594515096464e-05,
+      "loss": 0.5452,
+      "step": 3860
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839201776393575e-05,
+      "loss": 0.6587,
+      "step": 3861
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.839144101277504e-05,
+      "loss": 0.5034,
+      "step": 3862
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8390864261614327e-05,
+      "loss": 0.6005,
+      "step": 3863
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8390287510453618e-05,
+      "loss": 0.586,
+      "step": 3864
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8389710759292906e-05,
+      "loss": 0.5389,
+      "step": 3865
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8389134008132193e-05,
+      "loss": 0.6592,
+      "step": 3866
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.838855725697148e-05,
+      "loss": 0.597,
+      "step": 3867
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.838798050581077e-05,
+      "loss": 0.4589,
+      "step": 3868
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8387403754650056e-05,
+      "loss": 0.5773,
+      "step": 3869
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8386827003489348e-05,
+      "loss": 0.4354,
+      "step": 3870
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8386250252328635e-05,
+      "loss": 0.5359,
+      "step": 3871
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8385673501167923e-05,
+      "loss": 0.673,
+      "step": 3872
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.838509675000721e-05,
+      "loss": 0.712,
+      "step": 3873
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8384519998846498e-05,
+      "loss": 0.4642,
+      "step": 3874
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8383943247685786e-05,
+      "loss": 0.6,
+      "step": 3875
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8383366496525074e-05,
+      "loss": 0.5745,
+      "step": 3876
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8382789745364365e-05,
+      "loss": 0.5746,
+      "step": 3877
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8382212994203652e-05,
+      "loss": 0.5831,
+      "step": 3878
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.838163624304294e-05,
+      "loss": 0.5157,
+      "step": 3879
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8381059491882228e-05,
+      "loss": 0.593,
+      "step": 3880
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8380482740721515e-05,
+      "loss": 0.6129,
+      "step": 3881
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8379905989560803e-05,
+      "loss": 0.7286,
+      "step": 3882
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8379329238400094e-05,
+      "loss": 0.5835,
+      "step": 3883
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8378752487239382e-05,
+      "loss": 0.6088,
+      "step": 3884
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837817573607867e-05,
+      "loss": 0.4508,
+      "step": 3885
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837759898491796e-05,
+      "loss": 0.5269,
+      "step": 3886
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837702223375725e-05,
+      "loss": 0.6212,
+      "step": 3887
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8376445482596536e-05,
+      "loss": 0.4242,
+      "step": 3888
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8375868731435824e-05,
+      "loss": 0.4998,
+      "step": 3889
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837529198027511e-05,
+      "loss": 0.4535,
+      "step": 3890
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83747152291144e-05,
+      "loss": 0.5048,
+      "step": 3891
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837413847795369e-05,
+      "loss": 0.6761,
+      "step": 3892
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8373561726792978e-05,
+      "loss": 0.4837,
+      "step": 3893
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8372984975632266e-05,
+      "loss": 0.6227,
+      "step": 3894
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8372408224471553e-05,
+      "loss": 0.6013,
+      "step": 3895
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837183147331084e-05,
+      "loss": 0.589,
+      "step": 3896
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837125472215013e-05,
+      "loss": 0.5807,
+      "step": 3897
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.837067797098942e-05,
+      "loss": 0.5938,
+      "step": 3898
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8370101219828707e-05,
+      "loss": 0.588,
+      "step": 3899
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8369524468667995e-05,
+      "loss": 0.5962,
+      "step": 3900
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8368947717507283e-05,
+      "loss": 0.7187,
+      "step": 3901
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.836837096634657e-05,
+      "loss": 0.5913,
+      "step": 3902
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8367794215185858e-05,
+      "loss": 0.5813,
+      "step": 3903
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.836721746402515e-05,
+      "loss": 0.5996,
+      "step": 3904
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8366640712864437e-05,
+      "loss": 0.6021,
+      "step": 3905
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8366063961703725e-05,
+      "loss": 0.5231,
+      "step": 3906
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8365487210543012e-05,
+      "loss": 0.6294,
+      "step": 3907
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83649104593823e-05,
+      "loss": 0.6143,
+      "step": 3908
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8364333708221588e-05,
+      "loss": 0.6001,
+      "step": 3909
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.836375695706088e-05,
+      "loss": 0.5534,
+      "step": 3910
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8363180205900166e-05,
+      "loss": 0.5311,
+      "step": 3911
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8362603454739454e-05,
+      "loss": 0.6655,
+      "step": 3912
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8362026703578742e-05,
+      "loss": 0.5585,
+      "step": 3913
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8361449952418033e-05,
+      "loss": 0.5419,
+      "step": 3914
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8360873201257317e-05,
+      "loss": 0.6595,
+      "step": 3915
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8360296450096605e-05,
+      "loss": 0.6281,
+      "step": 3916
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8359719698935896e-05,
+      "loss": 0.6444,
+      "step": 3917
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8359142947775184e-05,
+      "loss": 0.6198,
+      "step": 3918
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835856619661447e-05,
+      "loss": 0.5256,
+      "step": 3919
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8357989445453762e-05,
+      "loss": 0.474,
+      "step": 3920
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835741269429305e-05,
+      "loss": 0.6817,
+      "step": 3921
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8356835943132338e-05,
+      "loss": 0.6943,
+      "step": 3922
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8356259191971625e-05,
+      "loss": 0.4162,
+      "step": 3923
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8355682440810913e-05,
+      "loss": 0.6643,
+      "step": 3924
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83551056896502e-05,
+      "loss": 0.6163,
+      "step": 3925
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8354528938489492e-05,
+      "loss": 0.6428,
+      "step": 3926
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835395218732878e-05,
+      "loss": 0.6726,
+      "step": 3927
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8353375436168067e-05,
+      "loss": 0.5409,
+      "step": 3928
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8352798685007355e-05,
+      "loss": 0.6455,
+      "step": 3929
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8352221933846643e-05,
+      "loss": 0.6022,
+      "step": 3930
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835164518268593e-05,
+      "loss": 0.5161,
+      "step": 3931
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835106843152522e-05,
+      "loss": 0.6286,
+      "step": 3932
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.835049168036451e-05,
+      "loss": 0.6517,
+      "step": 3933
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8349914929203797e-05,
+      "loss": 0.5738,
+      "step": 3934
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8349338178043084e-05,
+      "loss": 0.4787,
+      "step": 3935
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8348761426882372e-05,
+      "loss": 0.6797,
+      "step": 3936
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834818467572166e-05,
+      "loss": 0.731,
+      "step": 3937
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834760792456095e-05,
+      "loss": 0.6329,
+      "step": 3938
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834703117340024e-05,
+      "loss": 0.6333,
+      "step": 3939
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8346454422239526e-05,
+      "loss": 0.6151,
+      "step": 3940
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8345877671078817e-05,
+      "loss": 0.5821,
+      "step": 3941
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83453009199181e-05,
+      "loss": 0.6377,
+      "step": 3942
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834472416875739e-05,
+      "loss": 0.676,
+      "step": 3943
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834414741759668e-05,
+      "loss": 0.6048,
+      "step": 3944
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8343570666435968e-05,
+      "loss": 0.6163,
+      "step": 3945
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8342993915275256e-05,
+      "loss": 0.4979,
+      "step": 3946
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8342417164114543e-05,
+      "loss": 0.6601,
+      "step": 3947
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8341840412953834e-05,
+      "loss": 0.4932,
+      "step": 3948
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8341263661793122e-05,
+      "loss": 0.6597,
+      "step": 3949
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.834068691063241e-05,
+      "loss": 0.5606,
+      "step": 3950
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8340110159471697e-05,
+      "loss": 0.6794,
+      "step": 3951
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8339533408310985e-05,
+      "loss": 0.686,
+      "step": 3952
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8338956657150273e-05,
+      "loss": 0.647,
+      "step": 3953
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8338379905989564e-05,
+      "loss": 0.626,
+      "step": 3954
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.833780315482885e-05,
+      "loss": 0.6145,
+      "step": 3955
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.833722640366814e-05,
+      "loss": 0.5883,
+      "step": 3956
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8336649652507427e-05,
+      "loss": 0.6824,
+      "step": 3957
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8336072901346715e-05,
+      "loss": 0.6187,
+      "step": 3958
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8335496150186002e-05,
+      "loss": 0.4173,
+      "step": 3959
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8334919399025293e-05,
+      "loss": 0.6423,
+      "step": 3960
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.833434264786458e-05,
+      "loss": 0.4306,
+      "step": 3961
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.833376589670387e-05,
+      "loss": 0.5133,
+      "step": 3962
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8333189145543156e-05,
+      "loss": 0.7458,
+      "step": 3963
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8332612394382444e-05,
+      "loss": 0.5967,
+      "step": 3964
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8332035643221732e-05,
+      "loss": 0.8463,
+      "step": 3965
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8331458892061023e-05,
+      "loss": 0.5125,
+      "step": 3966
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.833088214090031e-05,
+      "loss": 0.5953,
+      "step": 3967
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8330305389739598e-05,
+      "loss": 0.6465,
+      "step": 3968
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8329728638578886e-05,
+      "loss": 0.7652,
+      "step": 3969
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8329151887418174e-05,
+      "loss": 0.6397,
+      "step": 3970
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832857513625746e-05,
+      "loss": 0.6551,
+      "step": 3971
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8327998385096752e-05,
+      "loss": 0.6326,
+      "step": 3972
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832742163393604e-05,
+      "loss": 0.7226,
+      "step": 3973
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8326844882775328e-05,
+      "loss": 0.6708,
+      "step": 3974
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832626813161462e-05,
+      "loss": 0.6899,
+      "step": 3975
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8325691380453907e-05,
+      "loss": 0.5845,
+      "step": 3976
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832511462929319e-05,
+      "loss": 0.5891,
+      "step": 3977
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8324537878132482e-05,
+      "loss": 0.5897,
+      "step": 3978
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832396112697177e-05,
+      "loss": 0.4947,
+      "step": 3979
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8323384375811057e-05,
+      "loss": 0.6288,
+      "step": 3980
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8322807624650345e-05,
+      "loss": 0.5821,
+      "step": 3981
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8322230873489636e-05,
+      "loss": 0.6697,
+      "step": 3982
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8321654122328924e-05,
+      "loss": 0.6711,
+      "step": 3983
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.832107737116821e-05,
+      "loss": 0.5572,
+      "step": 3984
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83205006200075e-05,
+      "loss": 0.6932,
+      "step": 3985
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8319923868846787e-05,
+      "loss": 0.6626,
+      "step": 3986
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8319347117686074e-05,
+      "loss": 0.5952,
+      "step": 3987
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8318770366525366e-05,
+      "loss": 0.6631,
+      "step": 3988
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8318193615364653e-05,
+      "loss": 0.6768,
+      "step": 3989
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.831761686420394e-05,
+      "loss": 0.6226,
+      "step": 3990
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.831704011304323e-05,
+      "loss": 0.5442,
+      "step": 3991
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8316463361882516e-05,
+      "loss": 0.5816,
+      "step": 3992
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8315886610721804e-05,
+      "loss": 0.6363,
+      "step": 3993
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8315309859561095e-05,
+      "loss": 0.4478,
+      "step": 3994
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8314733108400383e-05,
+      "loss": 0.7071,
+      "step": 3995
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.831415635723967e-05,
+      "loss": 0.5595,
+      "step": 3996
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8313579606078958e-05,
+      "loss": 0.6027,
+      "step": 3997
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8313002854918246e-05,
+      "loss": 0.5218,
+      "step": 3998
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8312426103757533e-05,
+      "loss": 0.5396,
+      "step": 3999
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8311849352596825e-05,
+      "loss": 0.6822,
+      "step": 4000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8311272601436112e-05,
+      "loss": 0.6057,
+      "step": 4001
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83106958502754e-05,
+      "loss": 0.5681,
+      "step": 4002
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.831011909911469e-05,
+      "loss": 0.5312,
+      "step": 4003
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8309542347953975e-05,
+      "loss": 0.5772,
+      "step": 4004
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8308965596793263e-05,
+      "loss": 0.5193,
+      "step": 4005
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8308388845632554e-05,
+      "loss": 0.6401,
+      "step": 4006
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8307812094471842e-05,
+      "loss": 0.5065,
+      "step": 4007
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.830723534331113e-05,
+      "loss": 0.6775,
+      "step": 4008
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.830665859215042e-05,
+      "loss": 0.4671,
+      "step": 4009
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8306081840989708e-05,
+      "loss": 0.56,
+      "step": 4010
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8305505089828996e-05,
+      "loss": 0.6737,
+      "step": 4011
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8304928338668284e-05,
+      "loss": 0.7372,
+      "step": 4012
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.830435158750757e-05,
+      "loss": 0.6309,
+      "step": 4013
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.830377483634686e-05,
+      "loss": 0.5943,
+      "step": 4014
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8303198085186147e-05,
+      "loss": 0.6037,
+      "step": 4015
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8302621334025438e-05,
+      "loss": 0.6855,
+      "step": 4016
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8302044582864725e-05,
+      "loss": 0.5591,
+      "step": 4017
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8301467831704013e-05,
+      "loss": 0.5444,
+      "step": 4018
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.83008910805433e-05,
+      "loss": 0.5765,
+      "step": 4019
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.830031432938259e-05,
+      "loss": 0.525,
+      "step": 4020
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8299737578221876e-05,
+      "loss": 0.7146,
+      "step": 4021
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8299160827061167e-05,
+      "loss": 0.6159,
+      "step": 4022
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8298584075900455e-05,
+      "loss": 0.6759,
+      "step": 4023
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8298007324739743e-05,
+      "loss": 0.516,
+      "step": 4024
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.829743057357903e-05,
+      "loss": 0.6295,
+      "step": 4025
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8296853822418318e-05,
+      "loss": 0.6159,
+      "step": 4026
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8296277071257606e-05,
+      "loss": 0.575,
+      "step": 4027
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8295700320096897e-05,
+      "loss": 0.6837,
+      "step": 4028
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8295123568936184e-05,
+      "loss": 0.7132,
+      "step": 4029
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8294546817775472e-05,
+      "loss": 0.6215,
+      "step": 4030
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.829397006661476e-05,
+      "loss": 0.6543,
+      "step": 4031
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8293393315454047e-05,
+      "loss": 0.5573,
+      "step": 4032
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8292816564293335e-05,
+      "loss": 0.5137,
+      "step": 4033
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8292239813132626e-05,
+      "loss": 0.6568,
+      "step": 4034
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8291663061971914e-05,
+      "loss": 0.5024,
+      "step": 4035
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82910863108112e-05,
+      "loss": 0.5411,
+      "step": 4036
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8290509559650493e-05,
+      "loss": 0.7038,
+      "step": 4037
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828993280848978e-05,
+      "loss": 0.6789,
+      "step": 4038
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8289356057329065e-05,
+      "loss": 0.6961,
+      "step": 4039
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8288779306168356e-05,
+      "loss": 0.6923,
+      "step": 4040
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8288202555007643e-05,
+      "loss": 0.6236,
+      "step": 4041
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828762580384693e-05,
+      "loss": 0.5591,
+      "step": 4042
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8287049052686222e-05,
+      "loss": 0.5368,
+      "step": 4043
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828647230152551e-05,
+      "loss": 0.611,
+      "step": 4044
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8285895550364797e-05,
+      "loss": 0.5697,
+      "step": 4045
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8285318799204085e-05,
+      "loss": 0.5563,
+      "step": 4046
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8284742048043373e-05,
+      "loss": 0.7237,
+      "step": 4047
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828416529688266e-05,
+      "loss": 0.6081,
+      "step": 4048
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828358854572195e-05,
+      "loss": 0.635,
+      "step": 4049
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828301179456124e-05,
+      "loss": 0.5841,
+      "step": 4050
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8282435043400527e-05,
+      "loss": 0.6441,
+      "step": 4051
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8281858292239815e-05,
+      "loss": 0.555,
+      "step": 4052
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8281281541079102e-05,
+      "loss": 0.4435,
+      "step": 4053
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.828070478991839e-05,
+      "loss": 0.6573,
+      "step": 4054
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8280128038757678e-05,
+      "loss": 0.5908,
+      "step": 4055
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.827955128759697e-05,
+      "loss": 0.5112,
+      "step": 4056
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8278974536436256e-05,
+      "loss": 0.5913,
+      "step": 4057
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8278397785275544e-05,
+      "loss": 0.5697,
+      "step": 4058
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8277821034114832e-05,
+      "loss": 0.592,
+      "step": 4059
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.827724428295412e-05,
+      "loss": 0.6096,
+      "step": 4060
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8276667531793407e-05,
+      "loss": 0.4567,
+      "step": 4061
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8276090780632698e-05,
+      "loss": 0.6211,
+      "step": 4062
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8275514029471986e-05,
+      "loss": 0.5482,
+      "step": 4063
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8274937278311274e-05,
+      "loss": 0.622,
+      "step": 4064
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8274360527150565e-05,
+      "loss": 0.7693,
+      "step": 4065
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.827378377598985e-05,
+      "loss": 0.5172,
+      "step": 4066
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8273207024829137e-05,
+      "loss": 0.5214,
+      "step": 4067
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8272630273668428e-05,
+      "loss": 0.6675,
+      "step": 4068
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8272053522507715e-05,
+      "loss": 0.5056,
+      "step": 4069
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8271476771347003e-05,
+      "loss": 0.5843,
+      "step": 4070
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8270900020186294e-05,
+      "loss": 0.5568,
+      "step": 4071
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8270323269025582e-05,
+      "loss": 0.6992,
+      "step": 4072
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826974651786487e-05,
+      "loss": 0.5627,
+      "step": 4073
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8269169766704157e-05,
+      "loss": 0.5238,
+      "step": 4074
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8268593015543445e-05,
+      "loss": 0.6688,
+      "step": 4075
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8268016264382733e-05,
+      "loss": 0.603,
+      "step": 4076
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8267439513222024e-05,
+      "loss": 0.5865,
+      "step": 4077
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826686276206131e-05,
+      "loss": 0.5455,
+      "step": 4078
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82662860109006e-05,
+      "loss": 0.4951,
+      "step": 4079
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8265709259739887e-05,
+      "loss": 0.6095,
+      "step": 4080
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8265132508579174e-05,
+      "loss": 0.6137,
+      "step": 4081
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8264555757418462e-05,
+      "loss": 0.5749,
+      "step": 4082
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8263979006257753e-05,
+      "loss": 0.563,
+      "step": 4083
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826340225509704e-05,
+      "loss": 0.7318,
+      "step": 4084
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826282550393633e-05,
+      "loss": 0.6981,
+      "step": 4085
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8262248752775616e-05,
+      "loss": 0.7479,
+      "step": 4086
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8261672001614904e-05,
+      "loss": 0.6428,
+      "step": 4087
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826109525045419e-05,
+      "loss": 0.5402,
+      "step": 4088
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.826051849929348e-05,
+      "loss": 0.508,
+      "step": 4089
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.825994174813277e-05,
+      "loss": 0.6035,
+      "step": 4090
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8259364996972058e-05,
+      "loss": 0.5995,
+      "step": 4091
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8258788245811346e-05,
+      "loss": 0.532,
+      "step": 4092
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8258211494650633e-05,
+      "loss": 0.4809,
+      "step": 4093
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.825763474348992e-05,
+      "loss": 0.5839,
+      "step": 4094
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.825705799232921e-05,
+      "loss": 0.6374,
+      "step": 4095
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82564812411685e-05,
+      "loss": 0.6397,
+      "step": 4096
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8255904490007788e-05,
+      "loss": 0.5575,
+      "step": 4097
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8255327738847075e-05,
+      "loss": 0.5773,
+      "step": 4098
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8254750987686366e-05,
+      "loss": 0.5039,
+      "step": 4099
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8254174236525654e-05,
+      "loss": 0.5389,
+      "step": 4100
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8253597485364942e-05,
+      "loss": 0.7279,
+      "step": 4101
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.825302073420423e-05,
+      "loss": 0.6753,
+      "step": 4102
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8252443983043517e-05,
+      "loss": 0.5031,
+      "step": 4103
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8251867231882805e-05,
+      "loss": 0.5899,
+      "step": 4104
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8251290480722096e-05,
+      "loss": 0.5725,
+      "step": 4105
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8250713729561384e-05,
+      "loss": 0.6462,
+      "step": 4106
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.825013697840067e-05,
+      "loss": 0.6026,
+      "step": 4107
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.824956022723996e-05,
+      "loss": 0.651,
+      "step": 4108
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8248983476079247e-05,
+      "loss": 0.5567,
+      "step": 4109
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8248406724918534e-05,
+      "loss": 0.6773,
+      "step": 4110
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8247829973757825e-05,
+      "loss": 0.6749,
+      "step": 4111
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8247253222597113e-05,
+      "loss": 0.6385,
+      "step": 4112
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82466764714364e-05,
+      "loss": 0.6408,
+      "step": 4113
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.824609972027569e-05,
+      "loss": 0.5154,
+      "step": 4114
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8245522969114976e-05,
+      "loss": 0.5381,
+      "step": 4115
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8244946217954264e-05,
+      "loss": 0.6431,
+      "step": 4116
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8244369466793555e-05,
+      "loss": 0.5823,
+      "step": 4117
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8243792715632843e-05,
+      "loss": 0.5306,
+      "step": 4118
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.824321596447213e-05,
+      "loss": 0.5755,
+      "step": 4119
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8242639213311418e-05,
+      "loss": 0.5367,
+      "step": 4120
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8242062462150706e-05,
+      "loss": 0.6077,
+      "step": 4121
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8241485710989993e-05,
+      "loss": 0.5422,
+      "step": 4122
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.824090895982928e-05,
+      "loss": 0.6258,
+      "step": 4123
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8240332208668572e-05,
+      "loss": 0.5896,
+      "step": 4124
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823975545750786e-05,
+      "loss": 0.537,
+      "step": 4125
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8239178706347147e-05,
+      "loss": 0.5856,
+      "step": 4126
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823860195518644e-05,
+      "loss": 0.5443,
+      "step": 4127
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8238025204025723e-05,
+      "loss": 0.4664,
+      "step": 4128
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823744845286501e-05,
+      "loss": 0.5447,
+      "step": 4129
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82368717017043e-05,
+      "loss": 0.6118,
+      "step": 4130
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823629495054359e-05,
+      "loss": 0.5786,
+      "step": 4131
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8235718199382877e-05,
+      "loss": 0.5692,
+      "step": 4132
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8235141448222168e-05,
+      "loss": 0.5958,
+      "step": 4133
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8234564697061456e-05,
+      "loss": 0.606,
+      "step": 4134
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8233987945900743e-05,
+      "loss": 0.5053,
+      "step": 4135
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823341119474003e-05,
+      "loss": 0.5731,
+      "step": 4136
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.823283444357932e-05,
+      "loss": 0.6758,
+      "step": 4137
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8232257692418606e-05,
+      "loss": 0.5928,
+      "step": 4138
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8231680941257897e-05,
+      "loss": 0.6652,
+      "step": 4139
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8231104190097185e-05,
+      "loss": 0.5559,
+      "step": 4140
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8230527438936473e-05,
+      "loss": 0.5323,
+      "step": 4141
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.822995068777576e-05,
+      "loss": 0.5749,
+      "step": 4142
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8229373936615048e-05,
+      "loss": 0.5841,
+      "step": 4143
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8228797185454336e-05,
+      "loss": 0.6212,
+      "step": 4144
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8228220434293627e-05,
+      "loss": 0.5803,
+      "step": 4145
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8227643683132915e-05,
+      "loss": 0.5442,
+      "step": 4146
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8227066931972202e-05,
+      "loss": 0.5906,
+      "step": 4147
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.822649018081149e-05,
+      "loss": 0.5148,
+      "step": 4148
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8225913429650778e-05,
+      "loss": 0.5689,
+      "step": 4149
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8225336678490065e-05,
+      "loss": 0.5596,
+      "step": 4150
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8224759927329356e-05,
+      "loss": 0.5959,
+      "step": 4151
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8224183176168644e-05,
+      "loss": 0.5721,
+      "step": 4152
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8223606425007932e-05,
+      "loss": 0.6145,
+      "step": 4153
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.822302967384722e-05,
+      "loss": 0.6145,
+      "step": 4154
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8222452922686507e-05,
+      "loss": 0.5195,
+      "step": 4155
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8221876171525795e-05,
+      "loss": 0.6142,
+      "step": 4156
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8221299420365083e-05,
+      "loss": 0.6872,
+      "step": 4157
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8220722669204374e-05,
+      "loss": 0.7303,
+      "step": 4158
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.822014591804366e-05,
+      "loss": 0.5439,
+      "step": 4159
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821956916688295e-05,
+      "loss": 0.5329,
+      "step": 4160
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821899241572224e-05,
+      "loss": 0.6146,
+      "step": 4161
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8218415664561528e-05,
+      "loss": 0.6832,
+      "step": 4162
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8217838913400815e-05,
+      "loss": 0.5269,
+      "step": 4163
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8217262162240103e-05,
+      "loss": 0.5045,
+      "step": 4164
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821668541107939e-05,
+      "loss": 0.662,
+      "step": 4165
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821610865991868e-05,
+      "loss": 0.5836,
+      "step": 4166
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821553190875797e-05,
+      "loss": 0.608,
+      "step": 4167
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8214955157597257e-05,
+      "loss": 0.5353,
+      "step": 4168
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8214378406436545e-05,
+      "loss": 0.769,
+      "step": 4169
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8213801655275833e-05,
+      "loss": 0.5932,
+      "step": 4170
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.821322490411512e-05,
+      "loss": 0.5201,
+      "step": 4171
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8212648152954408e-05,
+      "loss": 0.5937,
+      "step": 4172
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.82120714017937e-05,
+      "loss": 0.7473,
+      "step": 4173
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8211494650632987e-05,
+      "loss": 0.5084,
+      "step": 4174
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8210917899472274e-05,
+      "loss": 0.5684,
+      "step": 4175
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8210341148311562e-05,
+      "loss": 0.4578,
+      "step": 4176
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.820976439715085e-05,
+      "loss": 0.6447,
+      "step": 4177
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8209187645990138e-05,
+      "loss": 0.6332,
+      "step": 4178
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.820861089482943e-05,
+      "loss": 0.6216,
+      "step": 4179
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8208034143668716e-05,
+      "loss": 0.7274,
+      "step": 4180
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8207457392508004e-05,
+      "loss": 0.696,
+      "step": 4181
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.820688064134729e-05,
+      "loss": 0.6702,
+      "step": 4182
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.820630389018658e-05,
+      "loss": 0.6672,
+      "step": 4183
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8205727139025867e-05,
+      "loss": 0.5748,
+      "step": 4184
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8205150387865158e-05,
+      "loss": 0.6475,
+      "step": 4185
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8204573636704446e-05,
+      "loss": 0.7098,
+      "step": 4186
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8203996885543733e-05,
+      "loss": 0.7211,
+      "step": 4187
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8203420134383025e-05,
+      "loss": 0.6304,
+      "step": 4188
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8202843383222312e-05,
+      "loss": 0.6546,
+      "step": 4189
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8202266632061597e-05,
+      "loss": 0.6438,
+      "step": 4190
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8201689880900884e-05,
+      "loss": 0.6798,
+      "step": 4191
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8201113129740175e-05,
+      "loss": 0.4748,
+      "step": 4192
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8200536378579463e-05,
+      "loss": 0.5487,
+      "step": 4193
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.819995962741875e-05,
+      "loss": 0.5258,
+      "step": 4194
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8199382876258042e-05,
+      "loss": 0.5378,
+      "step": 4195
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.819880612509733e-05,
+      "loss": 0.6473,
+      "step": 4196
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8198229373936617e-05,
+      "loss": 0.7479,
+      "step": 4197
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8197652622775905e-05,
+      "loss": 0.7299,
+      "step": 4198
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8197075871615192e-05,
+      "loss": 0.6135,
+      "step": 4199
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.819649912045448e-05,
+      "loss": 0.6002,
+      "step": 4200
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.819592236929377e-05,
+      "loss": 0.5289,
+      "step": 4201
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.819534561813306e-05,
+      "loss": 0.5214,
+      "step": 4202
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8194768866972347e-05,
+      "loss": 0.6583,
+      "step": 4203
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8194192115811634e-05,
+      "loss": 0.5127,
+      "step": 4204
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8193615364650922e-05,
+      "loss": 0.6573,
+      "step": 4205
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.819303861349021e-05,
+      "loss": 0.5709,
+      "step": 4206
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81924618623295e-05,
+      "loss": 0.4798,
+      "step": 4207
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.819188511116879e-05,
+      "loss": 0.6578,
+      "step": 4208
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8191308360008076e-05,
+      "loss": 0.6446,
+      "step": 4209
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8190731608847364e-05,
+      "loss": 0.6435,
+      "step": 4210
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.819015485768665e-05,
+      "loss": 0.6004,
+      "step": 4211
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818957810652594e-05,
+      "loss": 0.4877,
+      "step": 4212
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818900135536523e-05,
+      "loss": 0.6606,
+      "step": 4213
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8188424604204518e-05,
+      "loss": 0.6607,
+      "step": 4214
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8187847853043806e-05,
+      "loss": 0.7366,
+      "step": 4215
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8187271101883097e-05,
+      "loss": 0.6562,
+      "step": 4216
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818669435072238e-05,
+      "loss": 0.6974,
+      "step": 4217
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818611759956167e-05,
+      "loss": 0.5996,
+      "step": 4218
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818554084840096e-05,
+      "loss": 0.6119,
+      "step": 4219
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8184964097240247e-05,
+      "loss": 0.6001,
+      "step": 4220
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8184387346079535e-05,
+      "loss": 0.569,
+      "step": 4221
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8183810594918826e-05,
+      "loss": 0.5958,
+      "step": 4222
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8183233843758114e-05,
+      "loss": 0.61,
+      "step": 4223
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81826570925974e-05,
+      "loss": 0.607,
+      "step": 4224
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.818208034143669e-05,
+      "loss": 0.499,
+      "step": 4225
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8181503590275977e-05,
+      "loss": 0.5814,
+      "step": 4226
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8180926839115265e-05,
+      "loss": 0.5367,
+      "step": 4227
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8180350087954552e-05,
+      "loss": 0.6549,
+      "step": 4228
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8179773336793843e-05,
+      "loss": 0.6653,
+      "step": 4229
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.817919658563313e-05,
+      "loss": 0.4916,
+      "step": 4230
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.817861983447242e-05,
+      "loss": 0.6406,
+      "step": 4231
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8178043083311706e-05,
+      "loss": 0.6839,
+      "step": 4232
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8177466332150994e-05,
+      "loss": 0.7408,
+      "step": 4233
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8176889580990282e-05,
+      "loss": 0.5259,
+      "step": 4234
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8176312829829573e-05,
+      "loss": 0.5252,
+      "step": 4235
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.817573607866886e-05,
+      "loss": 0.5474,
+      "step": 4236
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8175159327508148e-05,
+      "loss": 0.5824,
+      "step": 4237
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8174582576347436e-05,
+      "loss": 0.5918,
+      "step": 4238
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8174005825186724e-05,
+      "loss": 0.4814,
+      "step": 4239
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.817342907402601e-05,
+      "loss": 0.5835,
+      "step": 4240
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8172852322865302e-05,
+      "loss": 0.5196,
+      "step": 4241
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.817227557170459e-05,
+      "loss": 0.6498,
+      "step": 4242
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8171698820543878e-05,
+      "loss": 0.6342,
+      "step": 4243
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8171122069383165e-05,
+      "loss": 0.5677,
+      "step": 4244
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8170545318222453e-05,
+      "loss": 0.551,
+      "step": 4245
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816996856706174e-05,
+      "loss": 0.612,
+      "step": 4246
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8169391815901032e-05,
+      "loss": 0.6327,
+      "step": 4247
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816881506474032e-05,
+      "loss": 0.5096,
+      "step": 4248
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8168238313579607e-05,
+      "loss": 0.6204,
+      "step": 4249
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8167661562418898e-05,
+      "loss": 0.6922,
+      "step": 4250
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8167084811258186e-05,
+      "loss": 0.6482,
+      "step": 4251
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816650806009747e-05,
+      "loss": 0.7138,
+      "step": 4252
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816593130893676e-05,
+      "loss": 0.5017,
+      "step": 4253
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816535455777605e-05,
+      "loss": 0.6258,
+      "step": 4254
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8164777806615337e-05,
+      "loss": 0.6155,
+      "step": 4255
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8164201055454628e-05,
+      "loss": 0.6247,
+      "step": 4256
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8163624304293915e-05,
+      "loss": 0.6416,
+      "step": 4257
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8163047553133203e-05,
+      "loss": 0.6325,
+      "step": 4258
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816247080197249e-05,
+      "loss": 0.5146,
+      "step": 4259
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.816189405081178e-05,
+      "loss": 0.5821,
+      "step": 4260
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8161317299651066e-05,
+      "loss": 0.6661,
+      "step": 4261
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8160740548490354e-05,
+      "loss": 0.6517,
+      "step": 4262
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8160163797329645e-05,
+      "loss": 0.5709,
+      "step": 4263
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8159587046168933e-05,
+      "loss": 0.5417,
+      "step": 4264
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.815901029500822e-05,
+      "loss": 0.5131,
+      "step": 4265
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8158433543847508e-05,
+      "loss": 0.5796,
+      "step": 4266
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8157856792686796e-05,
+      "loss": 0.5817,
+      "step": 4267
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8157280041526083e-05,
+      "loss": 0.6286,
+      "step": 4268
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8156703290365374e-05,
+      "loss": 0.6502,
+      "step": 4269
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8156126539204662e-05,
+      "loss": 0.4201,
+      "step": 4270
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.815554978804395e-05,
+      "loss": 0.529,
+      "step": 4271
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8154973036883238e-05,
+      "loss": 0.6305,
+      "step": 4272
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8154396285722525e-05,
+      "loss": 0.6122,
+      "step": 4273
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8153819534561813e-05,
+      "loss": 0.626,
+      "step": 4274
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8153242783401104e-05,
+      "loss": 0.6219,
+      "step": 4275
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.815266603224039e-05,
+      "loss": 0.5138,
+      "step": 4276
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.815208928107968e-05,
+      "loss": 0.5143,
+      "step": 4277
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.815151252991897e-05,
+      "loss": 0.631,
+      "step": 4278
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8150935778758255e-05,
+      "loss": 0.6154,
+      "step": 4279
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8150359027597542e-05,
+      "loss": 0.578,
+      "step": 4280
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8149782276436833e-05,
+      "loss": 0.6104,
+      "step": 4281
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814920552527612e-05,
+      "loss": 0.6021,
+      "step": 4282
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814862877411541e-05,
+      "loss": 0.5084,
+      "step": 4283
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81480520229547e-05,
+      "loss": 0.6412,
+      "step": 4284
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8147475271793988e-05,
+      "loss": 0.6427,
+      "step": 4285
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8146898520633275e-05,
+      "loss": 0.5808,
+      "step": 4286
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8146321769472563e-05,
+      "loss": 0.5212,
+      "step": 4287
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814574501831185e-05,
+      "loss": 0.6135,
+      "step": 4288
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814516826715114e-05,
+      "loss": 0.6602,
+      "step": 4289
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814459151599043e-05,
+      "loss": 0.5627,
+      "step": 4290
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8144014764829717e-05,
+      "loss": 0.56,
+      "step": 4291
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8143438013669005e-05,
+      "loss": 0.7515,
+      "step": 4292
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8142861262508292e-05,
+      "loss": 0.4954,
+      "step": 4293
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.814228451134758e-05,
+      "loss": 0.6472,
+      "step": 4294
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8141707760186868e-05,
+      "loss": 0.5808,
+      "step": 4295
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8141131009026156e-05,
+      "loss": 0.5631,
+      "step": 4296
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8140554257865447e-05,
+      "loss": 0.5593,
+      "step": 4297
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8139977506704734e-05,
+      "loss": 0.5345,
+      "step": 4298
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8139400755544022e-05,
+      "loss": 0.6802,
+      "step": 4299
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.813882400438331e-05,
+      "loss": 0.5466,
+      "step": 4300
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8138247253222597e-05,
+      "loss": 0.5479,
+      "step": 4301
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8137670502061885e-05,
+      "loss": 0.6468,
+      "step": 4302
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8137093750901176e-05,
+      "loss": 0.5463,
+      "step": 4303
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8136516999740464e-05,
+      "loss": 0.5517,
+      "step": 4304
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.813594024857975e-05,
+      "loss": 0.5038,
+      "step": 4305
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.813536349741904e-05,
+      "loss": 0.5365,
+      "step": 4306
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8134786746258327e-05,
+      "loss": 0.6223,
+      "step": 4307
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8134209995097615e-05,
+      "loss": 0.4855,
+      "step": 4308
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8133633243936906e-05,
+      "loss": 0.6776,
+      "step": 4309
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8133056492776193e-05,
+      "loss": 0.5439,
+      "step": 4310
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.813247974161548e-05,
+      "loss": 0.6432,
+      "step": 4311
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8131902990454772e-05,
+      "loss": 0.6125,
+      "step": 4312
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.813132623929406e-05,
+      "loss": 0.6866,
+      "step": 4313
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8130749488133344e-05,
+      "loss": 0.6707,
+      "step": 4314
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8130172736972635e-05,
+      "loss": 0.6446,
+      "step": 4315
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8129595985811923e-05,
+      "loss": 0.6716,
+      "step": 4316
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812901923465121e-05,
+      "loss": 0.6147,
+      "step": 4317
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81284424834905e-05,
+      "loss": 0.5782,
+      "step": 4318
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812786573232979e-05,
+      "loss": 0.4742,
+      "step": 4319
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8127288981169077e-05,
+      "loss": 0.6476,
+      "step": 4320
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8126712230008365e-05,
+      "loss": 0.5329,
+      "step": 4321
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8126135478847652e-05,
+      "loss": 0.6508,
+      "step": 4322
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812555872768694e-05,
+      "loss": 0.7595,
+      "step": 4323
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812498197652623e-05,
+      "loss": 0.4666,
+      "step": 4324
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812440522536552e-05,
+      "loss": 0.542,
+      "step": 4325
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8123828474204806e-05,
+      "loss": 0.5997,
+      "step": 4326
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8123251723044094e-05,
+      "loss": 0.6239,
+      "step": 4327
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8122674971883382e-05,
+      "loss": 0.6116,
+      "step": 4328
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.812209822072267e-05,
+      "loss": 0.5925,
+      "step": 4329
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8121521469561957e-05,
+      "loss": 0.5122,
+      "step": 4330
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8120944718401248e-05,
+      "loss": 0.6203,
+      "step": 4331
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8120367967240536e-05,
+      "loss": 0.5591,
+      "step": 4332
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8119791216079824e-05,
+      "loss": 0.6214,
+      "step": 4333
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.811921446491911e-05,
+      "loss": 0.7043,
+      "step": 4334
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.81186377137584e-05,
+      "loss": 0.5297,
+      "step": 4335
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8118060962597687e-05,
+      "loss": 0.6831,
+      "step": 4336
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8117484211436978e-05,
+      "loss": 0.6759,
+      "step": 4337
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8116907460276265e-05,
+      "loss": 0.5491,
+      "step": 4338
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8116330709115553e-05,
+      "loss": 0.5789,
+      "step": 4339
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8115753957954844e-05,
+      "loss": 0.5726,
+      "step": 4340
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.811517720679413e-05,
+      "loss": 0.6605,
+      "step": 4341
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8114600455633416e-05,
+      "loss": 0.6446,
+      "step": 4342
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8114023704472707e-05,
+      "loss": 0.5101,
+      "step": 4343
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8113446953311995e-05,
+      "loss": 0.4804,
+      "step": 4344
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8112870202151283e-05,
+      "loss": 0.7035,
+      "step": 4345
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8112293450990574e-05,
+      "loss": 0.4699,
+      "step": 4346
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.811171669982986e-05,
+      "loss": 0.6787,
+      "step": 4347
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.811113994866915e-05,
+      "loss": 0.6352,
+      "step": 4348
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8110563197508437e-05,
+      "loss": 0.5898,
+      "step": 4349
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8109986446347724e-05,
+      "loss": 0.5792,
+      "step": 4350
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8109409695187012e-05,
+      "loss": 0.5174,
+      "step": 4351
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8108832944026303e-05,
+      "loss": 0.6301,
+      "step": 4352
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810825619286559e-05,
+      "loss": 0.5517,
+      "step": 4353
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810767944170488e-05,
+      "loss": 0.6705,
+      "step": 4354
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8107102690544166e-05,
+      "loss": 0.6376,
+      "step": 4355
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8106525939383454e-05,
+      "loss": 0.6071,
+      "step": 4356
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810594918822274e-05,
+      "loss": 0.6784,
+      "step": 4357
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8105372437062033e-05,
+      "loss": 0.6014,
+      "step": 4358
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810479568590132e-05,
+      "loss": 0.5492,
+      "step": 4359
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8104218934740608e-05,
+      "loss": 0.6282,
+      "step": 4360
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8103642183579896e-05,
+      "loss": 0.5263,
+      "step": 4361
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8103065432419183e-05,
+      "loss": 0.4451,
+      "step": 4362
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810248868125847e-05,
+      "loss": 0.6519,
+      "step": 4363
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810191193009776e-05,
+      "loss": 0.4954,
+      "step": 4364
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.810133517893705e-05,
+      "loss": 0.5416,
+      "step": 4365
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8100758427776338e-05,
+      "loss": 0.5752,
+      "step": 4366
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8100181676615625e-05,
+      "loss": 0.5112,
+      "step": 4367
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8099604925454913e-05,
+      "loss": 0.5884,
+      "step": 4368
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.80990281742942e-05,
+      "loss": 0.6538,
+      "step": 4369
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8098451423133488e-05,
+      "loss": 0.4961,
+      "step": 4370
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.809787467197278e-05,
+      "loss": 0.614,
+      "step": 4371
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8097297920812067e-05,
+      "loss": 0.6074,
+      "step": 4372
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8096721169651355e-05,
+      "loss": 0.5578,
+      "step": 4373
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8096144418490646e-05,
+      "loss": 0.5588,
+      "step": 4374
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8095567667329933e-05,
+      "loss": 0.5682,
+      "step": 4375
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8094990916169218e-05,
+      "loss": 0.5566,
+      "step": 4376
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.809441416500851e-05,
+      "loss": 0.6336,
+      "step": 4377
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8093837413847797e-05,
+      "loss": 0.5434,
+      "step": 4378
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.8093260662687084e-05,
+      "loss": 0.7214,
+      "step": 4379
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8092683911526375e-05,
+      "loss": 0.5472,
+      "step": 4380
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8092107160365663e-05,
+      "loss": 0.6086,
+      "step": 4381
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.809153040920495e-05,
+      "loss": 0.6031,
+      "step": 4382
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.809095365804424e-05,
+      "loss": 0.5524,
+      "step": 4383
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8090376906883526e-05,
+      "loss": 0.5929,
+      "step": 4384
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8089800155722814e-05,
+      "loss": 0.5533,
+      "step": 4385
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8089223404562105e-05,
+      "loss": 0.5657,
+      "step": 4386
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8088646653401392e-05,
+      "loss": 0.6248,
+      "step": 4387
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.808806990224068e-05,
+      "loss": 0.5989,
+      "step": 4388
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8087493151079968e-05,
+      "loss": 0.498,
+      "step": 4389
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8086916399919256e-05,
+      "loss": 0.6028,
+      "step": 4390
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8086339648758543e-05,
+      "loss": 0.7065,
+      "step": 4391
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8085762897597834e-05,
+      "loss": 0.6177,
+      "step": 4392
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8085186146437122e-05,
+      "loss": 0.7335,
+      "step": 4393
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.808460939527641e-05,
+      "loss": 0.6204,
+      "step": 4394
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8084032644115697e-05,
+      "loss": 0.4757,
+      "step": 4395
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8083455892954985e-05,
+      "loss": 0.6674,
+      "step": 4396
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8082879141794273e-05,
+      "loss": 0.68,
+      "step": 4397
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.808230239063356e-05,
+      "loss": 0.5039,
+      "step": 4398
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.808172563947285e-05,
+      "loss": 0.6415,
+      "step": 4399
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.808114888831214e-05,
+      "loss": 0.5115,
+      "step": 4400
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8080572137151427e-05,
+      "loss": 0.6229,
+      "step": 4401
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8079995385990718e-05,
+      "loss": 0.5363,
+      "step": 4402
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8079418634830002e-05,
+      "loss": 0.6513,
+      "step": 4403
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807884188366929e-05,
+      "loss": 0.5823,
+      "step": 4404
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807826513250858e-05,
+      "loss": 0.5791,
+      "step": 4405
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807768838134787e-05,
+      "loss": 0.646,
+      "step": 4406
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8077111630187156e-05,
+      "loss": 0.6383,
+      "step": 4407
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8076534879026447e-05,
+      "loss": 0.5299,
+      "step": 4408
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8075958127865735e-05,
+      "loss": 0.5568,
+      "step": 4409
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8075381376705023e-05,
+      "loss": 0.5609,
+      "step": 4410
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807480462554431e-05,
+      "loss": 0.4834,
+      "step": 4411
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8074227874383598e-05,
+      "loss": 0.7224,
+      "step": 4412
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8073651123222886e-05,
+      "loss": 0.6156,
+      "step": 4413
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8073074372062177e-05,
+      "loss": 0.6208,
+      "step": 4414
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8072497620901465e-05,
+      "loss": 0.5319,
+      "step": 4415
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8071920869740752e-05,
+      "loss": 0.5405,
+      "step": 4416
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.807134411858004e-05,
+      "loss": 0.4853,
+      "step": 4417
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8070767367419328e-05,
+      "loss": 0.5948,
+      "step": 4418
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8070190616258615e-05,
+      "loss": 0.6475,
+      "step": 4419
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8069613865097906e-05,
+      "loss": 0.5901,
+      "step": 4420
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8069037113937194e-05,
+      "loss": 0.617,
+      "step": 4421
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8068460362776482e-05,
+      "loss": 0.5602,
+      "step": 4422
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806788361161577e-05,
+      "loss": 0.6231,
+      "step": 4423
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8067306860455057e-05,
+      "loss": 0.5489,
+      "step": 4424
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8066730109294345e-05,
+      "loss": 0.539,
+      "step": 4425
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8066153358133636e-05,
+      "loss": 0.5876,
+      "step": 4426
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8065576606972924e-05,
+      "loss": 0.5401,
+      "step": 4427
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806499985581221e-05,
+      "loss": 0.4981,
+      "step": 4428
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.80644231046515e-05,
+      "loss": 0.6153,
+      "step": 4429
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8063846353490787e-05,
+      "loss": 0.4924,
+      "step": 4430
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8063269602330074e-05,
+      "loss": 0.5339,
+      "step": 4431
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8062692851169365e-05,
+      "loss": 0.6534,
+      "step": 4432
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8062116100008653e-05,
+      "loss": 0.5998,
+      "step": 4433
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806153934884794e-05,
+      "loss": 0.5426,
+      "step": 4434
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806096259768723e-05,
+      "loss": 0.5719,
+      "step": 4435
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.806038584652652e-05,
+      "loss": 0.7349,
+      "step": 4436
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8059809095365807e-05,
+      "loss": 0.5521,
+      "step": 4437
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805923234420509e-05,
+      "loss": 0.6045,
+      "step": 4438
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8058655593044383e-05,
+      "loss": 0.7028,
+      "step": 4439
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805807884188367e-05,
+      "loss": 0.5398,
+      "step": 4440
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8057502090722958e-05,
+      "loss": 0.5378,
+      "step": 4441
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805692533956225e-05,
+      "loss": 0.6468,
+      "step": 4442
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8056348588401537e-05,
+      "loss": 0.5651,
+      "step": 4443
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8055771837240824e-05,
+      "loss": 0.5759,
+      "step": 4444
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8055195086080112e-05,
+      "loss": 0.6538,
+      "step": 4445
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.80546183349194e-05,
+      "loss": 0.6927,
+      "step": 4446
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8054041583758687e-05,
+      "loss": 0.5006,
+      "step": 4447
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805346483259798e-05,
+      "loss": 0.5195,
+      "step": 4448
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8052888081437266e-05,
+      "loss": 0.6631,
+      "step": 4449
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8052311330276554e-05,
+      "loss": 0.4438,
+      "step": 4450
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805173457911584e-05,
+      "loss": 0.5789,
+      "step": 4451
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.805115782795513e-05,
+      "loss": 0.5906,
+      "step": 4452
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8050581076794417e-05,
+      "loss": 0.5543,
+      "step": 4453
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8050004325633708e-05,
+      "loss": 0.6412,
+      "step": 4454
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8049427574472996e-05,
+      "loss": 0.6095,
+      "step": 4455
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8048850823312283e-05,
+      "loss": 0.5527,
+      "step": 4456
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804827407215157e-05,
+      "loss": 0.5336,
+      "step": 4457
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804769732099086e-05,
+      "loss": 0.7893,
+      "step": 4458
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8047120569830146e-05,
+      "loss": 0.6581,
+      "step": 4459
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8046543818669438e-05,
+      "loss": 0.6967,
+      "step": 4460
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8045967067508725e-05,
+      "loss": 0.6942,
+      "step": 4461
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8045390316348013e-05,
+      "loss": 0.5529,
+      "step": 4462
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8044813565187304e-05,
+      "loss": 0.5686,
+      "step": 4463
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804423681402659e-05,
+      "loss": 0.5266,
+      "step": 4464
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8043660062865876e-05,
+      "loss": 0.6369,
+      "step": 4465
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8043083311705167e-05,
+      "loss": 0.4839,
+      "step": 4466
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8042506560544455e-05,
+      "loss": 0.553,
+      "step": 4467
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8041929809383742e-05,
+      "loss": 0.3816,
+      "step": 4468
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804135305822303e-05,
+      "loss": 0.4918,
+      "step": 4469
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804077630706232e-05,
+      "loss": 0.6604,
+      "step": 4470
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.804019955590161e-05,
+      "loss": 0.4351,
+      "step": 4471
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8039622804740897e-05,
+      "loss": 0.6405,
+      "step": 4472
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8039046053580184e-05,
+      "loss": 0.6972,
+      "step": 4473
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8038469302419472e-05,
+      "loss": 0.5776,
+      "step": 4474
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803789255125876e-05,
+      "loss": 0.7651,
+      "step": 4475
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803731580009805e-05,
+      "loss": 0.566,
+      "step": 4476
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803673904893734e-05,
+      "loss": 0.5336,
+      "step": 4477
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8036162297776626e-05,
+      "loss": 0.6685,
+      "step": 4478
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8035585546615914e-05,
+      "loss": 0.5163,
+      "step": 4479
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.80350087954552e-05,
+      "loss": 0.5072,
+      "step": 4480
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803443204429449e-05,
+      "loss": 0.5518,
+      "step": 4481
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803385529313378e-05,
+      "loss": 0.6143,
+      "step": 4482
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8033278541973068e-05,
+      "loss": 0.5735,
+      "step": 4483
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8032701790812356e-05,
+      "loss": 0.6461,
+      "step": 4484
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8032125039651643e-05,
+      "loss": 0.5955,
+      "step": 4485
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803154828849093e-05,
+      "loss": 0.5631,
+      "step": 4486
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803097153733022e-05,
+      "loss": 0.6348,
+      "step": 4487
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.803039478616951e-05,
+      "loss": 0.5367,
+      "step": 4488
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8029818035008797e-05,
+      "loss": 0.5638,
+      "step": 4489
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8029241283848085e-05,
+      "loss": 0.5469,
+      "step": 4490
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8028664532687373e-05,
+      "loss": 0.7203,
+      "step": 4491
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802808778152666e-05,
+      "loss": 0.6484,
+      "step": 4492
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8027511030365948e-05,
+      "loss": 0.6491,
+      "step": 4493
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802693427920524e-05,
+      "loss": 0.5118,
+      "step": 4494
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8026357528044527e-05,
+      "loss": 0.6523,
+      "step": 4495
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8025780776883815e-05,
+      "loss": 0.7412,
+      "step": 4496
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8025204025723106e-05,
+      "loss": 0.7416,
+      "step": 4497
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8024627274562393e-05,
+      "loss": 0.6311,
+      "step": 4498
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802405052340168e-05,
+      "loss": 0.536,
+      "step": 4499
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802347377224097e-05,
+      "loss": 0.515,
+      "step": 4500
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8022897021080256e-05,
+      "loss": 0.6434,
+      "step": 4501
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8022320269919544e-05,
+      "loss": 0.6666,
+      "step": 4502
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802174351875883e-05,
+      "loss": 0.6161,
+      "step": 4503
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8021166767598123e-05,
+      "loss": 0.5346,
+      "step": 4504
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.802059001643741e-05,
+      "loss": 0.5571,
+      "step": 4505
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8020013265276698e-05,
+      "loss": 0.5278,
+      "step": 4506
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8019436514115986e-05,
+      "loss": 0.7271,
+      "step": 4507
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8018859762955274e-05,
+      "loss": 0.6105,
+      "step": 4508
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801828301179456e-05,
+      "loss": 0.5908,
+      "step": 4509
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8017706260633852e-05,
+      "loss": 0.5154,
+      "step": 4510
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801712950947314e-05,
+      "loss": 0.5687,
+      "step": 4511
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8016552758312428e-05,
+      "loss": 0.6717,
+      "step": 4512
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8015976007151715e-05,
+      "loss": 0.6868,
+      "step": 4513
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8015399255991003e-05,
+      "loss": 0.6085,
+      "step": 4514
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801482250483029e-05,
+      "loss": 0.606,
+      "step": 4515
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8014245753669582e-05,
+      "loss": 0.4872,
+      "step": 4516
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801366900250887e-05,
+      "loss": 0.5172,
+      "step": 4517
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8013092251348157e-05,
+      "loss": 0.6226,
+      "step": 4518
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8012515500187445e-05,
+      "loss": 0.4823,
+      "step": 4519
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8011938749026732e-05,
+      "loss": 0.5771,
+      "step": 4520
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801136199786602e-05,
+      "loss": 0.4737,
+      "step": 4521
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.801078524670531e-05,
+      "loss": 0.6493,
+      "step": 4522
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.80102084955446e-05,
+      "loss": 0.69,
+      "step": 4523
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8009631744383887e-05,
+      "loss": 0.4863,
+      "step": 4524
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8009054993223178e-05,
+      "loss": 0.5652,
+      "step": 4525
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8008478242062465e-05,
+      "loss": 0.706,
+      "step": 4526
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800790149090175e-05,
+      "loss": 0.7094,
+      "step": 4527
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800732473974104e-05,
+      "loss": 0.5901,
+      "step": 4528
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800674798858033e-05,
+      "loss": 0.5671,
+      "step": 4529
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8006171237419616e-05,
+      "loss": 0.485,
+      "step": 4530
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8005594486258907e-05,
+      "loss": 0.663,
+      "step": 4531
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8005017735098195e-05,
+      "loss": 0.4914,
+      "step": 4532
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8004440983937483e-05,
+      "loss": 0.5155,
+      "step": 4533
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.800386423277677e-05,
+      "loss": 0.6273,
+      "step": 4534
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8003287481616058e-05,
+      "loss": 0.5585,
+      "step": 4535
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8002710730455346e-05,
+      "loss": 0.5711,
+      "step": 4536
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8002133979294637e-05,
+      "loss": 0.6487,
+      "step": 4537
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8001557228133924e-05,
+      "loss": 0.7001,
+      "step": 4538
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.8000980476973212e-05,
+      "loss": 0.6377,
+      "step": 4539
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.80004037258125e-05,
+      "loss": 0.5807,
+      "step": 4540
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7999826974651787e-05,
+      "loss": 0.5716,
+      "step": 4541
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7999250223491075e-05,
+      "loss": 0.6228,
+      "step": 4542
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7998673472330363e-05,
+      "loss": 0.6812,
+      "step": 4543
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7998096721169654e-05,
+      "loss": 0.5664,
+      "step": 4544
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.799751997000894e-05,
+      "loss": 0.4061,
+      "step": 4545
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.799694321884823e-05,
+      "loss": 0.6302,
+      "step": 4546
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7996366467687517e-05,
+      "loss": 0.4957,
+      "step": 4547
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7995789716526805e-05,
+      "loss": 0.5866,
+      "step": 4548
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7995212965366092e-05,
+      "loss": 0.6237,
+      "step": 4549
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7994636214205383e-05,
+      "loss": 0.594,
+      "step": 4550
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.799405946304467e-05,
+      "loss": 0.6401,
+      "step": 4551
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.799348271188396e-05,
+      "loss": 0.5656,
+      "step": 4552
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.799290596072325e-05,
+      "loss": 0.6318,
+      "step": 4553
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7992329209562534e-05,
+      "loss": 0.578,
+      "step": 4554
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7991752458401822e-05,
+      "loss": 0.5702,
+      "step": 4555
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7991175707241113e-05,
+      "loss": 0.6216,
+      "step": 4556
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.79905989560804e-05,
+      "loss": 0.615,
+      "step": 4557
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7990022204919688e-05,
+      "loss": 0.5457,
+      "step": 4558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798944545375898e-05,
+      "loss": 0.6816,
+      "step": 4559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7988868702598267e-05,
+      "loss": 0.5988,
+      "step": 4560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7988291951437555e-05,
+      "loss": 0.6312,
+      "step": 4561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7987715200276842e-05,
+      "loss": 0.5372,
+      "step": 4562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798713844911613e-05,
+      "loss": 0.6278,
+      "step": 4563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7986561697955418e-05,
+      "loss": 0.5742,
+      "step": 4564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798598494679471e-05,
+      "loss": 0.4242,
+      "step": 4565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7985408195633997e-05,
+      "loss": 0.5046,
+      "step": 4566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7984831444473284e-05,
+      "loss": 0.5928,
+      "step": 4567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7984254693312572e-05,
+      "loss": 0.6138,
+      "step": 4568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798367794215186e-05,
+      "loss": 0.6933,
+      "step": 4569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7983101190991147e-05,
+      "loss": 0.5755,
+      "step": 4570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798252443983044e-05,
+      "loss": 0.5461,
+      "step": 4571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7981947688669726e-05,
+      "loss": 0.7854,
+      "step": 4572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7981370937509014e-05,
+      "loss": 0.5828,
+      "step": 4573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.79807941863483e-05,
+      "loss": 0.6121,
+      "step": 4574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.798021743518759e-05,
+      "loss": 0.5854,
+      "step": 4575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7979640684026877e-05,
+      "loss": 0.7338,
+      "step": 4576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7979063932866164e-05,
+      "loss": 0.5414,
+      "step": 4577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7978487181705456e-05,
+      "loss": 0.6779,
+      "step": 4578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7977910430544743e-05,
+      "loss": 0.5699,
+      "step": 4579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797733367938403e-05,
+      "loss": 0.6275,
+      "step": 4580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797675692822332e-05,
+      "loss": 0.7001,
+      "step": 4581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7976180177062606e-05,
+      "loss": 0.6447,
+      "step": 4582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7975603425901894e-05,
+      "loss": 0.6667,
+      "step": 4583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7975026674741185e-05,
+      "loss": 0.5436,
+      "step": 4584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7974449923580473e-05,
+      "loss": 0.724,
+      "step": 4585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797387317241976e-05,
+      "loss": 0.6473,
+      "step": 4586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797329642125905e-05,
+      "loss": 0.5984,
+      "step": 4587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797271967009834e-05,
+      "loss": 0.6145,
+      "step": 4588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7972142918937623e-05,
+      "loss": 0.5468,
+      "step": 4589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7971566167776915e-05,
+      "loss": 0.6643,
+      "step": 4590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7970989416616202e-05,
+      "loss": 0.6523,
+      "step": 4591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.797041266545549e-05,
+      "loss": 0.5085,
+      "step": 4592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796983591429478e-05,
+      "loss": 0.6684,
+      "step": 4593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796925916313407e-05,
+      "loss": 0.5563,
+      "step": 4594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7968682411973356e-05,
+      "loss": 0.4729,
+      "step": 4595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7968105660812644e-05,
+      "loss": 0.5815,
+      "step": 4596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796752890965193e-05,
+      "loss": 0.6999,
+      "step": 4597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796695215849122e-05,
+      "loss": 0.5244,
+      "step": 4598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796637540733051e-05,
+      "loss": 0.5657,
+      "step": 4599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7965798656169798e-05,
+      "loss": 0.5615,
+      "step": 4600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7965221905009086e-05,
+      "loss": 0.6454,
+      "step": 4601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7964645153848373e-05,
+      "loss": 0.47,
+      "step": 4602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796406840268766e-05,
+      "loss": 0.5952,
+      "step": 4603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796349165152695e-05,
+      "loss": 0.5421,
+      "step": 4604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796291490036624e-05,
+      "loss": 0.523,
+      "step": 4605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7962338149205528e-05,
+      "loss": 0.4824,
+      "step": 4606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7961761398044815e-05,
+      "loss": 0.656,
+      "step": 4607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7961184646884103e-05,
+      "loss": 0.5828,
+      "step": 4608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796060789572339e-05,
+      "loss": 0.6084,
+      "step": 4609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.796003114456268e-05,
+      "loss": 0.6267,
+      "step": 4610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7959454393401966e-05,
+      "loss": 0.6401,
+      "step": 4611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7958877642241257e-05,
+      "loss": 0.5458,
+      "step": 4612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7958300891080545e-05,
+      "loss": 0.6156,
+      "step": 4613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7957724139919832e-05,
+      "loss": 0.6459,
+      "step": 4614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7957147388759124e-05,
+      "loss": 0.6496,
+      "step": 4615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7956570637598408e-05,
+      "loss": 0.5559,
+      "step": 4616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7955993886437696e-05,
+      "loss": 0.6473,
+      "step": 4617
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7955417135276987e-05,
+      "loss": 0.6646,
+      "step": 4618
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7954840384116274e-05,
+      "loss": 0.5788,
+      "step": 4619
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7954263632955562e-05,
+      "loss": 0.5148,
+      "step": 4620
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7953686881794853e-05,
+      "loss": 0.554,
+      "step": 4621
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.795311013063414e-05,
+      "loss": 0.6791,
+      "step": 4622
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.795253337947343e-05,
+      "loss": 0.6487,
+      "step": 4623
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7951956628312716e-05,
+      "loss": 0.5146,
+      "step": 4624
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7951379877152004e-05,
+      "loss": 0.6086,
+      "step": 4625
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.795080312599129e-05,
+      "loss": 0.5739,
+      "step": 4626
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7950226374830583e-05,
+      "loss": 0.6367,
+      "step": 4627
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794964962366987e-05,
+      "loss": 0.5326,
+      "step": 4628
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7949072872509158e-05,
+      "loss": 0.5891,
+      "step": 4629
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7948496121348446e-05,
+      "loss": 0.5351,
+      "step": 4630
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7947919370187733e-05,
+      "loss": 0.6505,
+      "step": 4631
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794734261902702e-05,
+      "loss": 0.5334,
+      "step": 4632
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7946765867866312e-05,
+      "loss": 0.6098,
+      "step": 4633
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.79461891167056e-05,
+      "loss": 0.6379,
+      "step": 4634
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7945612365544887e-05,
+      "loss": 0.4412,
+      "step": 4635
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7945035614384175e-05,
+      "loss": 0.5188,
+      "step": 4636
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7944458863223463e-05,
+      "loss": 0.5918,
+      "step": 4637
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794388211206275e-05,
+      "loss": 0.5959,
+      "step": 4638
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794330536090204e-05,
+      "loss": 0.509,
+      "step": 4639
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794272860974133e-05,
+      "loss": 0.5525,
+      "step": 4640
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7942151858580617e-05,
+      "loss": 0.6679,
+      "step": 4641
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7941575107419905e-05,
+      "loss": 0.6118,
+      "step": 4642
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7940998356259192e-05,
+      "loss": 0.4986,
+      "step": 4643
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.794042160509848e-05,
+      "loss": 0.5225,
+      "step": 4644
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7939844853937768e-05,
+      "loss": 0.577,
+      "step": 4645
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.793926810277706e-05,
+      "loss": 0.6428,
+      "step": 4646
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7938691351616346e-05,
+      "loss": 0.7135,
+      "step": 4647
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7938114600455634e-05,
+      "loss": 0.5015,
+      "step": 4648
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7937537849294925e-05,
+      "loss": 0.5736,
+      "step": 4649
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7936961098134213e-05,
+      "loss": 0.4645,
+      "step": 4650
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7936384346973497e-05,
+      "loss": 0.6722,
+      "step": 4651
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7935807595812788e-05,
+      "loss": 0.5819,
+      "step": 4652
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7935230844652076e-05,
+      "loss": 0.6512,
+      "step": 4653
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7934654093491364e-05,
+      "loss": 0.6224,
+      "step": 4654
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7934077342330655e-05,
+      "loss": 0.5902,
+      "step": 4655
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7933500591169942e-05,
+      "loss": 0.6545,
+      "step": 4656
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.793292384000923e-05,
+      "loss": 0.5865,
+      "step": 4657
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7932347088848518e-05,
+      "loss": 0.5737,
+      "step": 4658
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7931770337687805e-05,
+      "loss": 0.4994,
+      "step": 4659
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7931193586527093e-05,
+      "loss": 0.7177,
+      "step": 4660
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7930616835366384e-05,
+      "loss": 0.6696,
+      "step": 4661
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7930040084205672e-05,
+      "loss": 0.5658,
+      "step": 4662
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792946333304496e-05,
+      "loss": 0.5991,
+      "step": 4663
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7928886581884247e-05,
+      "loss": 0.5503,
+      "step": 4664
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7928309830723535e-05,
+      "loss": 0.6326,
+      "step": 4665
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7927733079562823e-05,
+      "loss": 0.5513,
+      "step": 4666
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7927156328402114e-05,
+      "loss": 0.6043,
+      "step": 4667
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.79265795772414e-05,
+      "loss": 0.6652,
+      "step": 4668
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792600282608069e-05,
+      "loss": 0.6386,
+      "step": 4669
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7925426074919977e-05,
+      "loss": 0.768,
+      "step": 4670
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7924849323759264e-05,
+      "loss": 0.5465,
+      "step": 4671
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7924272572598552e-05,
+      "loss": 0.5878,
+      "step": 4672
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7923695821437843e-05,
+      "loss": 0.6235,
+      "step": 4673
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792311907027713e-05,
+      "loss": 0.5991,
+      "step": 4674
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792254231911642e-05,
+      "loss": 0.548,
+      "step": 4675
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792196556795571e-05,
+      "loss": 0.4502,
+      "step": 4676
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7921388816794997e-05,
+      "loss": 0.4417,
+      "step": 4677
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792081206563428e-05,
+      "loss": 0.6391,
+      "step": 4678
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.792023531447357e-05,
+      "loss": 0.571,
+      "step": 4679
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.791965856331286e-05,
+      "loss": 0.5957,
+      "step": 4680
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7919081812152148e-05,
+      "loss": 0.4743,
+      "step": 4681
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7918505060991436e-05,
+      "loss": 0.5568,
+      "step": 4682
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7917928309830727e-05,
+      "loss": 0.6345,
+      "step": 4683
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7917351558670015e-05,
+      "loss": 0.6127,
+      "step": 4684
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7916774807509302e-05,
+      "loss": 0.693,
+      "step": 4685
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.791619805634859e-05,
+      "loss": 0.5346,
+      "step": 4686
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7915621305187878e-05,
+      "loss": 0.4334,
+      "step": 4687
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7915044554027165e-05,
+      "loss": 0.5077,
+      "step": 4688
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7914467802866456e-05,
+      "loss": 0.4726,
+      "step": 4689
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7913891051705744e-05,
+      "loss": 0.5834,
+      "step": 4690
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.791331430054503e-05,
+      "loss": 0.6347,
+      "step": 4691
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.791273754938432e-05,
+      "loss": 0.5391,
+      "step": 4692
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7912160798223607e-05,
+      "loss": 0.6596,
+      "step": 4693
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7911584047062895e-05,
+      "loss": 0.8165,
+      "step": 4694
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7911007295902186e-05,
+      "loss": 0.5783,
+      "step": 4695
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7910430544741473e-05,
+      "loss": 0.5772,
+      "step": 4696
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790985379358076e-05,
+      "loss": 0.5455,
+      "step": 4697
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790927704242005e-05,
+      "loss": 0.4942,
+      "step": 4698
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7908700291259337e-05,
+      "loss": 0.5438,
+      "step": 4699
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7908123540098624e-05,
+      "loss": 0.6051,
+      "step": 4700
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7907546788937915e-05,
+      "loss": 0.6135,
+      "step": 4701
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7906970037777203e-05,
+      "loss": 0.5805,
+      "step": 4702
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790639328661649e-05,
+      "loss": 0.619,
+      "step": 4703
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790581653545578e-05,
+      "loss": 0.6893,
+      "step": 4704
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7905239784295066e-05,
+      "loss": 0.5389,
+      "step": 4705
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7904663033134354e-05,
+      "loss": 0.5433,
+      "step": 4706
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7904086281973645e-05,
+      "loss": 0.4264,
+      "step": 4707
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7903509530812932e-05,
+      "loss": 0.5445,
+      "step": 4708
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790293277965222e-05,
+      "loss": 0.5481,
+      "step": 4709
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790235602849151e-05,
+      "loss": 0.5673,
+      "step": 4710
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.79017792773308e-05,
+      "loss": 0.7157,
+      "step": 4711
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7901202526170087e-05,
+      "loss": 0.6298,
+      "step": 4712
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.790062577500937e-05,
+      "loss": 0.5915,
+      "step": 4713
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7900049023848662e-05,
+      "loss": 0.5453,
+      "step": 4714
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789947227268795e-05,
+      "loss": 0.6056,
+      "step": 4715
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7898895521527237e-05,
+      "loss": 0.711,
+      "step": 4716
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789831877036653e-05,
+      "loss": 0.5893,
+      "step": 4717
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7897742019205816e-05,
+      "loss": 0.5152,
+      "step": 4718
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7897165268045104e-05,
+      "loss": 0.6573,
+      "step": 4719
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789658851688439e-05,
+      "loss": 0.6394,
+      "step": 4720
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789601176572368e-05,
+      "loss": 0.6642,
+      "step": 4721
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7895435014562967e-05,
+      "loss": 0.5082,
+      "step": 4722
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7894858263402258e-05,
+      "loss": 0.6355,
+      "step": 4723
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7894281512241546e-05,
+      "loss": 0.5599,
+      "step": 4724
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7893704761080833e-05,
+      "loss": 0.6062,
+      "step": 4725
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789312800992012e-05,
+      "loss": 0.4269,
+      "step": 4726
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.789255125875941e-05,
+      "loss": 0.6847,
+      "step": 4727
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7891974507598696e-05,
+      "loss": 0.5375,
+      "step": 4728
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7891397756437987e-05,
+      "loss": 0.648,
+      "step": 4729
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7890821005277275e-05,
+      "loss": 0.5814,
+      "step": 4730
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7890244254116563e-05,
+      "loss": 0.6053,
+      "step": 4731
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.788966750295585e-05,
+      "loss": 0.6028,
+      "step": 4732
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7889090751795138e-05,
+      "loss": 0.7591,
+      "step": 4733
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7888514000634426e-05,
+      "loss": 0.4964,
+      "step": 4734
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7887937249473717e-05,
+      "loss": 0.5718,
+      "step": 4735
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7887360498313005e-05,
+      "loss": 0.5277,
+      "step": 4736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7886783747152292e-05,
+      "loss": 0.6187,
+      "step": 4737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7886206995991583e-05,
+      "loss": 0.546,
+      "step": 4738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.788563024483087e-05,
+      "loss": 0.5736,
+      "step": 4739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7885053493670155e-05,
+      "loss": 0.5165,
+      "step": 4740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7884476742509446e-05,
+      "loss": 0.5488,
+      "step": 4741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7883899991348734e-05,
+      "loss": 0.6304,
+      "step": 4742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7883323240188022e-05,
+      "loss": 0.5536,
+      "step": 4743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7882746489027313e-05,
+      "loss": 0.6028,
+      "step": 4744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.78821697378666e-05,
+      "loss": 0.4678,
+      "step": 4745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7881592986705888e-05,
+      "loss": 0.6899,
+      "step": 4746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7881016235545176e-05,
+      "loss": 0.5067,
+      "step": 4747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7880439484384464e-05,
+      "loss": 0.7429,
+      "step": 4748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787986273322375e-05,
+      "loss": 0.5777,
+      "step": 4749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787928598206304e-05,
+      "loss": 0.5927,
+      "step": 4750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787870923090233e-05,
+      "loss": 0.669,
+      "step": 4751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7878132479741618e-05,
+      "loss": 0.667,
+      "step": 4752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7877555728580905e-05,
+      "loss": 0.5454,
+      "step": 4753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7876978977420193e-05,
+      "loss": 0.5512,
+      "step": 4754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787640222625948e-05,
+      "loss": 0.532,
+      "step": 4755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787582547509877e-05,
+      "loss": 0.4637,
+      "step": 4756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787524872393806e-05,
+      "loss": 0.4723,
+      "step": 4757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7874671972777347e-05,
+      "loss": 0.5473,
+      "step": 4758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7874095221616635e-05,
+      "loss": 0.5667,
+      "step": 4759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7873518470455923e-05,
+      "loss": 0.7101,
+      "step": 4760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787294171929521e-05,
+      "loss": 0.4893,
+      "step": 4761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7872364968134498e-05,
+      "loss": 0.6363,
+      "step": 4762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.787178821697379e-05,
+      "loss": 0.6094,
+      "step": 4763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7871211465813077e-05,
+      "loss": 0.5768,
+      "step": 4764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7870634714652364e-05,
+      "loss": 0.5956,
+      "step": 4765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7870057963491652e-05,
+      "loss": 0.6976,
+      "step": 4766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.786948121233094e-05,
+      "loss": 0.5226,
+      "step": 4767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7868904461170227e-05,
+      "loss": 0.575,
+      "step": 4768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.786832771000952e-05,
+      "loss": 0.5912,
+      "step": 4769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7867750958848806e-05,
+      "loss": 0.5042,
+      "step": 4770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7867174207688094e-05,
+      "loss": 0.6616,
+      "step": 4771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7866597456527385e-05,
+      "loss": 0.5488,
+      "step": 4772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7866020705366673e-05,
+      "loss": 0.4063,
+      "step": 4773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.786544395420596e-05,
+      "loss": 0.6237,
+      "step": 4774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7864867203045248e-05,
+      "loss": 0.4303,
+      "step": 4775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7864290451884536e-05,
+      "loss": 0.5648,
+      "step": 4776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7863713700723823e-05,
+      "loss": 0.5012,
+      "step": 4777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7863136949563114e-05,
+      "loss": 0.6045,
+      "step": 4778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7862560198402402e-05,
+      "loss": 0.5895,
+      "step": 4779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.786198344724169e-05,
+      "loss": 0.5716,
+      "step": 4780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7861406696080978e-05,
+      "loss": 0.5469,
+      "step": 4781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7860829944920265e-05,
+      "loss": 0.5431,
+      "step": 4782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7860253193759553e-05,
+      "loss": 0.6316,
+      "step": 4783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785967644259884e-05,
+      "loss": 0.6266,
+      "step": 4784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785909969143813e-05,
+      "loss": 0.7135,
+      "step": 4785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785852294027742e-05,
+      "loss": 0.5598,
+      "step": 4786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7857946189116707e-05,
+      "loss": 0.6048,
+      "step": 4787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7857369437955995e-05,
+      "loss": 0.5596,
+      "step": 4788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7856792686795282e-05,
+      "loss": 0.5839,
+      "step": 4789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785621593563457e-05,
+      "loss": 0.5801,
+      "step": 4790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785563918447386e-05,
+      "loss": 0.4426,
+      "step": 4791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785506243331315e-05,
+      "loss": 0.5062,
+      "step": 4792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7854485682152437e-05,
+      "loss": 0.5812,
+      "step": 4793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7853908930991724e-05,
+      "loss": 0.6478,
+      "step": 4794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7853332179831012e-05,
+      "loss": 0.7261,
+      "step": 4795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.78527554286703e-05,
+      "loss": 0.6082,
+      "step": 4796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785217867750959e-05,
+      "loss": 0.5675,
+      "step": 4797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.785160192634888e-05,
+      "loss": 0.6597,
+      "step": 4798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7851025175188166e-05,
+      "loss": 0.5769,
+      "step": 4799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7850448424027457e-05,
+      "loss": 0.678,
+      "step": 4800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7849871672866745e-05,
+      "loss": 0.7091,
+      "step": 4801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.784929492170603e-05,
+      "loss": 0.6492,
+      "step": 4802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.784871817054532e-05,
+      "loss": 0.5908,
+      "step": 4803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7848141419384608e-05,
+      "loss": 0.5626,
+      "step": 4804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7847564668223896e-05,
+      "loss": 0.5117,
+      "step": 4805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7846987917063187e-05,
+      "loss": 0.6264,
+      "step": 4806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7846411165902474e-05,
+      "loss": 0.5448,
+      "step": 4807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7845834414741762e-05,
+      "loss": 0.7137,
+      "step": 4808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.784525766358105e-05,
+      "loss": 0.6111,
+      "step": 4809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7844680912420337e-05,
+      "loss": 0.5616,
+      "step": 4810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7844104161259625e-05,
+      "loss": 0.6361,
+      "step": 4811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7843527410098916e-05,
+      "loss": 0.5953,
+      "step": 4812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7842950658938204e-05,
+      "loss": 0.6083,
+      "step": 4813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.784237390777749e-05,
+      "loss": 0.5202,
+      "step": 4814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.784179715661678e-05,
+      "loss": 0.6259,
+      "step": 4815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7841220405456067e-05,
+      "loss": 0.5941,
+      "step": 4816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7840643654295355e-05,
+      "loss": 0.4955,
+      "step": 4817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7840066903134642e-05,
+      "loss": 0.5724,
+      "step": 4818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7839490151973933e-05,
+      "loss": 0.5712,
+      "step": 4819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.783891340081322e-05,
+      "loss": 0.7397,
+      "step": 4820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.783833664965251e-05,
+      "loss": 0.4637,
+      "step": 4821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7837759898491796e-05,
+      "loss": 0.6518,
+      "step": 4822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7837183147331084e-05,
+      "loss": 0.5643,
+      "step": 4823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7836606396170372e-05,
+      "loss": 0.4214,
+      "step": 4824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7836029645009663e-05,
+      "loss": 0.5358,
+      "step": 4825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.783545289384895e-05,
+      "loss": 0.4715,
+      "step": 4826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7834876142688238e-05,
+      "loss": 0.5441,
+      "step": 4827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7834299391527526e-05,
+      "loss": 0.5751,
+      "step": 4828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7833722640366814e-05,
+      "loss": 0.5133,
+      "step": 4829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.78331458892061e-05,
+      "loss": 0.642,
+      "step": 4830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7832569138045392e-05,
+      "loss": 0.5775,
+      "step": 4831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.783199238688468e-05,
+      "loss": 0.5925,
+      "step": 4832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7831415635723968e-05,
+      "loss": 0.5562,
+      "step": 4833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.783083888456326e-05,
+      "loss": 0.6026,
+      "step": 4834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7830262133402546e-05,
+      "loss": 0.7982,
+      "step": 4835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7829685382241834e-05,
+      "loss": 0.5059,
+      "step": 4836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7829108631081122e-05,
+      "loss": 0.5032,
+      "step": 4837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.782853187992041e-05,
+      "loss": 0.5975,
+      "step": 4838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7827955128759697e-05,
+      "loss": 0.5847,
+      "step": 4839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7827378377598988e-05,
+      "loss": 0.4284,
+      "step": 4840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7826801626438276e-05,
+      "loss": 0.5762,
+      "step": 4841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7826224875277564e-05,
+      "loss": 0.6249,
+      "step": 4842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.782564812411685e-05,
+      "loss": 0.4768,
+      "step": 4843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.782507137295614e-05,
+      "loss": 0.5267,
+      "step": 4844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7824494621795427e-05,
+      "loss": 0.5889,
+      "step": 4845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7823917870634718e-05,
+      "loss": 0.4402,
+      "step": 4846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7823341119474005e-05,
+      "loss": 0.5703,
+      "step": 4847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7822764368313293e-05,
+      "loss": 0.6135,
+      "step": 4848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.782218761715258e-05,
+      "loss": 0.581,
+      "step": 4849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.782161086599187e-05,
+      "loss": 0.6629,
+      "step": 4850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7821034114831156e-05,
+      "loss": 0.5339,
+      "step": 4851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7820457363670444e-05,
+      "loss": 0.6666,
+      "step": 4852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7819880612509735e-05,
+      "loss": 0.5216,
+      "step": 4853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7819303861349023e-05,
+      "loss": 0.4884,
+      "step": 4854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781872711018831e-05,
+      "loss": 0.5101,
+      "step": 4855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7818150359027598e-05,
+      "loss": 0.5659,
+      "step": 4856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7817573607866886e-05,
+      "loss": 0.6273,
+      "step": 4857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7816996856706173e-05,
+      "loss": 0.6087,
+      "step": 4858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7816420105545464e-05,
+      "loss": 0.7657,
+      "step": 4859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7815843354384752e-05,
+      "loss": 0.6019,
+      "step": 4860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781526660322404e-05,
+      "loss": 0.5161,
+      "step": 4861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781468985206333e-05,
+      "loss": 0.7835,
+      "step": 4862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781411310090262e-05,
+      "loss": 0.5778,
+      "step": 4863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7813536349741903e-05,
+      "loss": 0.5369,
+      "step": 4864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7812959598581194e-05,
+      "loss": 0.7845,
+      "step": 4865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781238284742048e-05,
+      "loss": 0.5782,
+      "step": 4866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781180609625977e-05,
+      "loss": 0.658,
+      "step": 4867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.781122934509906e-05,
+      "loss": 0.6626,
+      "step": 4868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7810652593938348e-05,
+      "loss": 0.4689,
+      "step": 4869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7810075842777636e-05,
+      "loss": 0.542,
+      "step": 4870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7809499091616923e-05,
+      "loss": 0.5999,
+      "step": 4871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.780892234045621e-05,
+      "loss": 0.4628,
+      "step": 4872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.78083455892955e-05,
+      "loss": 0.5821,
+      "step": 4873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.780776883813479e-05,
+      "loss": 0.5334,
+      "step": 4874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7807192086974078e-05,
+      "loss": 0.6121,
+      "step": 4875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7806615335813365e-05,
+      "loss": 0.6183,
+      "step": 4876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7806038584652653e-05,
+      "loss": 0.5324,
+      "step": 4877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.780546183349194e-05,
+      "loss": 0.5632,
+      "step": 4878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7804885082331228e-05,
+      "loss": 0.7227,
+      "step": 4879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.780430833117052e-05,
+      "loss": 0.6063,
+      "step": 4880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7803731580009807e-05,
+      "loss": 0.4871,
+      "step": 4881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7803154828849095e-05,
+      "loss": 0.6038,
+      "step": 4882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7802578077688382e-05,
+      "loss": 0.6352,
+      "step": 4883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.780200132652767e-05,
+      "loss": 0.5905,
+      "step": 4884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7801424575366958e-05,
+      "loss": 0.6612,
+      "step": 4885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7800847824206245e-05,
+      "loss": 0.5842,
+      "step": 4886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7800271073045537e-05,
+      "loss": 0.5854,
+      "step": 4887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7799694321884824e-05,
+      "loss": 0.5186,
+      "step": 4888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7799117570724112e-05,
+      "loss": 0.5664,
+      "step": 4889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.77985408195634e-05,
+      "loss": 0.5581,
+      "step": 4890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7797964068402687e-05,
+      "loss": 0.5341,
+      "step": 4891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7797387317241975e-05,
+      "loss": 0.6546,
+      "step": 4892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7796810566081266e-05,
+      "loss": 0.5849,
+      "step": 4893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7796233814920554e-05,
+      "loss": 0.64,
+      "step": 4894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.779565706375984e-05,
+      "loss": 0.608,
+      "step": 4895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7795080312599132e-05,
+      "loss": 0.6812,
+      "step": 4896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.779450356143842e-05,
+      "loss": 0.552,
+      "step": 4897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7793926810277708e-05,
+      "loss": 0.561,
+      "step": 4898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7793350059116996e-05,
+      "loss": 0.6023,
+      "step": 4899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7792773307956283e-05,
+      "loss": 0.4914,
+      "step": 4900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.779219655679557e-05,
+      "loss": 0.5969,
+      "step": 4901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7791619805634862e-05,
+      "loss": 0.5352,
+      "step": 4902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.779104305447415e-05,
+      "loss": 0.5978,
+      "step": 4903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7790466303313437e-05,
+      "loss": 0.6311,
+      "step": 4904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7789889552152725e-05,
+      "loss": 0.5727,
+      "step": 4905
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7789312800992013e-05,
+      "loss": 0.5499,
+      "step": 4906
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.77887360498313e-05,
+      "loss": 0.6312,
+      "step": 4907
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.778815929867059e-05,
+      "loss": 0.5869,
+      "step": 4908
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.778758254750988e-05,
+      "loss": 0.7124,
+      "step": 4909
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7787005796349167e-05,
+      "loss": 0.5598,
+      "step": 4910
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7786429045188455e-05,
+      "loss": 0.6176,
+      "step": 4911
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7785852294027742e-05,
+      "loss": 0.5287,
+      "step": 4912
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.778527554286703e-05,
+      "loss": 0.6408,
+      "step": 4913
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.778469879170632e-05,
+      "loss": 0.5027,
+      "step": 4914
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.778412204054561e-05,
+      "loss": 0.5717,
+      "step": 4915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7783545289384896e-05,
+      "loss": 0.5685,
+      "step": 4916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7782968538224184e-05,
+      "loss": 0.6069,
+      "step": 4917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7782391787063472e-05,
+      "loss": 0.5479,
+      "step": 4918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.778181503590276e-05,
+      "loss": 0.5271,
+      "step": 4919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.778123828474205e-05,
+      "loss": 0.5405,
+      "step": 4920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7780661533581338e-05,
+      "loss": 0.7288,
+      "step": 4921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7780084782420626e-05,
+      "loss": 0.5991,
+      "step": 4922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7779508031259914e-05,
+      "loss": 0.4664,
+      "step": 4923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7778931280099205e-05,
+      "loss": 0.7055,
+      "step": 4924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7778354528938492e-05,
+      "loss": 0.5247,
+      "step": 4925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7777777777777777e-05,
+      "loss": 0.6484,
+      "step": 4926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7777201026617068e-05,
+      "loss": 0.563,
+      "step": 4927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7776624275456355e-05,
+      "loss": 0.5002,
+      "step": 4928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7776047524295643e-05,
+      "loss": 0.5776,
+      "step": 4929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7775470773134934e-05,
+      "loss": 0.6547,
+      "step": 4930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7774894021974222e-05,
+      "loss": 0.4584,
+      "step": 4931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.777431727081351e-05,
+      "loss": 0.5061,
+      "step": 4932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7773740519652797e-05,
+      "loss": 0.5796,
+      "step": 4933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7773163768492085e-05,
+      "loss": 0.4979,
+      "step": 4934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7772587017331373e-05,
+      "loss": 0.494,
+      "step": 4935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7772010266170664e-05,
+      "loss": 0.7044,
+      "step": 4936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.777143351500995e-05,
+      "loss": 0.5584,
+      "step": 4937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.777085676384924e-05,
+      "loss": 0.4848,
+      "step": 4938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7770280012688527e-05,
+      "loss": 0.5524,
+      "step": 4939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7769703261527814e-05,
+      "loss": 0.6065,
+      "step": 4940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7769126510367102e-05,
+      "loss": 0.589,
+      "step": 4941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7768549759206393e-05,
+      "loss": 0.7128,
+      "step": 4942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776797300804568e-05,
+      "loss": 0.5526,
+      "step": 4943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776739625688497e-05,
+      "loss": 0.6157,
+      "step": 4944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7766819505724256e-05,
+      "loss": 0.7028,
+      "step": 4945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7766242754563544e-05,
+      "loss": 0.583,
+      "step": 4946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776566600340283e-05,
+      "loss": 0.5562,
+      "step": 4947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7765089252242123e-05,
+      "loss": 0.6791,
+      "step": 4948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776451250108141e-05,
+      "loss": 0.6412,
+      "step": 4949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7763935749920698e-05,
+      "loss": 0.6612,
+      "step": 4950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776335899875999e-05,
+      "loss": 0.4865,
+      "step": 4951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7762782247599277e-05,
+      "loss": 0.5469,
+      "step": 4952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776220549643856e-05,
+      "loss": 0.6043,
+      "step": 4953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7761628745277852e-05,
+      "loss": 0.5447,
+      "step": 4954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.776105199411714e-05,
+      "loss": 0.6698,
+      "step": 4955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7760475242956427e-05,
+      "loss": 0.652,
+      "step": 4956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7759898491795715e-05,
+      "loss": 0.4885,
+      "step": 4957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7759321740635006e-05,
+      "loss": 0.5602,
+      "step": 4958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7758744989474294e-05,
+      "loss": 0.4973,
+      "step": 4959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.775816823831358e-05,
+      "loss": 0.4969,
+      "step": 4960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.775759148715287e-05,
+      "loss": 0.6232,
+      "step": 4961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7757014735992157e-05,
+      "loss": 0.6381,
+      "step": 4962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7756437984831445e-05,
+      "loss": 0.5727,
+      "step": 4963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7755861233670736e-05,
+      "loss": 0.6205,
+      "step": 4964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7755284482510023e-05,
+      "loss": 0.4826,
+      "step": 4965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.775470773134931e-05,
+      "loss": 0.4962,
+      "step": 4966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.77541309801886e-05,
+      "loss": 0.4069,
+      "step": 4967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7753554229027886e-05,
+      "loss": 0.4923,
+      "step": 4968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7752977477867174e-05,
+      "loss": 0.4037,
+      "step": 4969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7752400726706465e-05,
+      "loss": 0.5299,
+      "step": 4970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7751823975545753e-05,
+      "loss": 0.6079,
+      "step": 4971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.775124722438504e-05,
+      "loss": 0.5917,
+      "step": 4972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7750670473224328e-05,
+      "loss": 0.5677,
+      "step": 4973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7750093722063616e-05,
+      "loss": 0.6933,
+      "step": 4974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7749516970902904e-05,
+      "loss": 0.4669,
+      "step": 4975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7748940219742195e-05,
+      "loss": 0.6417,
+      "step": 4976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7748363468581482e-05,
+      "loss": 0.6041,
+      "step": 4977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.774778671742077e-05,
+      "loss": 0.5556,
+      "step": 4978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7747209966260058e-05,
+      "loss": 0.6938,
+      "step": 4979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7746633215099345e-05,
+      "loss": 0.5762,
+      "step": 4980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7746056463938633e-05,
+      "loss": 0.5399,
+      "step": 4981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7745479712777924e-05,
+      "loss": 0.6321,
+      "step": 4982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7744902961617212e-05,
+      "loss": 0.5801,
+      "step": 4983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.77443262104565e-05,
+      "loss": 0.706,
+      "step": 4984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.774374945929579e-05,
+      "loss": 0.5354,
+      "step": 4985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.774317270813508e-05,
+      "loss": 0.5423,
+      "step": 4986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7742595956974366e-05,
+      "loss": 0.485,
+      "step": 4987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7742019205813654e-05,
+      "loss": 0.6717,
+      "step": 4988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.774144245465294e-05,
+      "loss": 0.6673,
+      "step": 4989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.774086570349223e-05,
+      "loss": 0.6165,
+      "step": 4990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7740288952331517e-05,
+      "loss": 0.483,
+      "step": 4991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7739712201170808e-05,
+      "loss": 0.5219,
+      "step": 4992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7739135450010096e-05,
+      "loss": 0.5807,
+      "step": 4993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7738558698849383e-05,
+      "loss": 0.5163,
+      "step": 4994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.773798194768867e-05,
+      "loss": 0.5366,
+      "step": 4995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.773740519652796e-05,
+      "loss": 0.5183,
+      "step": 4996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7736828445367246e-05,
+      "loss": 0.5553,
+      "step": 4997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7736251694206537e-05,
+      "loss": 0.4776,
+      "step": 4998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7735674943045825e-05,
+      "loss": 0.6039,
+      "step": 4999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7735098191885113e-05,
+      "loss": 0.708,
+      "step": 5000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.77345214407244e-05,
+      "loss": 0.5715,
+      "step": 5001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7733944689563688e-05,
+      "loss": 0.7122,
+      "step": 5002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7733367938402976e-05,
+      "loss": 0.6154,
+      "step": 5003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7732791187242267e-05,
+      "loss": 0.6975,
+      "step": 5004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7732214436081555e-05,
+      "loss": 0.6139,
+      "step": 5005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7731637684920842e-05,
+      "loss": 0.5325,
+      "step": 5006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.773106093376013e-05,
+      "loss": 0.5328,
+      "step": 5007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7730484182599418e-05,
+      "loss": 0.686,
+      "step": 5008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7729907431438705e-05,
+      "loss": 0.6173,
+      "step": 5009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7729330680277996e-05,
+      "loss": 0.6336,
+      "step": 5010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7728753929117284e-05,
+      "loss": 0.5918,
+      "step": 5011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7728177177956572e-05,
+      "loss": 0.6053,
+      "step": 5012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7727600426795863e-05,
+      "loss": 0.5636,
+      "step": 5013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.772702367563515e-05,
+      "loss": 0.5026,
+      "step": 5014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7726446924474435e-05,
+      "loss": 0.6629,
+      "step": 5015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7725870173313726e-05,
+      "loss": 0.6099,
+      "step": 5016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7725293422153014e-05,
+      "loss": 0.4693,
+      "step": 5017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.77247166709923e-05,
+      "loss": 0.4122,
+      "step": 5018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7724139919831592e-05,
+      "loss": 0.5978,
+      "step": 5019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.772356316867088e-05,
+      "loss": 0.4924,
+      "step": 5020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7722986417510168e-05,
+      "loss": 0.4648,
+      "step": 5021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7722409666349455e-05,
+      "loss": 0.6046,
+      "step": 5022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7721832915188743e-05,
+      "loss": 0.5133,
+      "step": 5023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.772125616402803e-05,
+      "loss": 0.6281,
+      "step": 5024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.772067941286732e-05,
+      "loss": 0.6722,
+      "step": 5025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.772010266170661e-05,
+      "loss": 0.5833,
+      "step": 5026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7719525910545897e-05,
+      "loss": 0.6638,
+      "step": 5027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7718949159385185e-05,
+      "loss": 0.5994,
+      "step": 5028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7718372408224473e-05,
+      "loss": 0.6441,
+      "step": 5029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771779565706376e-05,
+      "loss": 0.6238,
+      "step": 5030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7717218905903048e-05,
+      "loss": 0.4896,
+      "step": 5031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771664215474234e-05,
+      "loss": 0.5232,
+      "step": 5032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7716065403581627e-05,
+      "loss": 0.6951,
+      "step": 5033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7715488652420914e-05,
+      "loss": 0.4852,
+      "step": 5034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7714911901260202e-05,
+      "loss": 0.6446,
+      "step": 5035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771433515009949e-05,
+      "loss": 0.4822,
+      "step": 5036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7713758398938777e-05,
+      "loss": 0.7455,
+      "step": 5037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771318164777807e-05,
+      "loss": 0.6588,
+      "step": 5038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7712604896617356e-05,
+      "loss": 0.5944,
+      "step": 5039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7712028145456644e-05,
+      "loss": 0.6366,
+      "step": 5040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771145139429593e-05,
+      "loss": 0.6745,
+      "step": 5041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.771087464313522e-05,
+      "loss": 0.5495,
+      "step": 5042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7710297891974507e-05,
+      "loss": 0.3614,
+      "step": 5043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7709721140813798e-05,
+      "loss": 0.524,
+      "step": 5044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7709144389653086e-05,
+      "loss": 0.551,
+      "step": 5045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7708567638492373e-05,
+      "loss": 0.5253,
+      "step": 5046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7707990887331664e-05,
+      "loss": 0.6304,
+      "step": 5047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7707414136170952e-05,
+      "loss": 0.665,
+      "step": 5048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.770683738501024e-05,
+      "loss": 0.5434,
+      "step": 5049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7706260633849527e-05,
+      "loss": 0.4148,
+      "step": 5050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7705683882688815e-05,
+      "loss": 0.5518,
+      "step": 5051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7705107131528103e-05,
+      "loss": 0.5197,
+      "step": 5052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7704530380367394e-05,
+      "loss": 0.5532,
+      "step": 5053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.770395362920668e-05,
+      "loss": 0.5208,
+      "step": 5054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.770337687804597e-05,
+      "loss": 0.5223,
+      "step": 5055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7702800126885257e-05,
+      "loss": 0.4883,
+      "step": 5056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7702223375724545e-05,
+      "loss": 0.5274,
+      "step": 5057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7701646624563832e-05,
+      "loss": 0.7287,
+      "step": 5058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7701069873403123e-05,
+      "loss": 0.5533,
+      "step": 5059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.770049312224241e-05,
+      "loss": 0.6064,
+      "step": 5060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.76999163710817e-05,
+      "loss": 0.4656,
+      "step": 5061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7699339619920986e-05,
+      "loss": 0.6075,
+      "step": 5062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7698762868760274e-05,
+      "loss": 0.5431,
+      "step": 5063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7698186117599562e-05,
+      "loss": 0.4606,
+      "step": 5064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769760936643885e-05,
+      "loss": 0.6193,
+      "step": 5065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769703261527814e-05,
+      "loss": 0.5479,
+      "step": 5066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7696455864117428e-05,
+      "loss": 0.6137,
+      "step": 5067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7695879112956716e-05,
+      "loss": 0.5286,
+      "step": 5068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7695302361796004e-05,
+      "loss": 0.6284,
+      "step": 5069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769472561063529e-05,
+      "loss": 0.6208,
+      "step": 5070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769414885947458e-05,
+      "loss": 0.5156,
+      "step": 5071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769357210831387e-05,
+      "loss": 0.494,
+      "step": 5072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7692995357153158e-05,
+      "loss": 0.608,
+      "step": 5073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7692418605992445e-05,
+      "loss": 0.6694,
+      "step": 5074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7691841854831737e-05,
+      "loss": 0.5959,
+      "step": 5075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7691265103671024e-05,
+      "loss": 0.4454,
+      "step": 5076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.769068835251031e-05,
+      "loss": 0.6642,
+      "step": 5077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.76901116013496e-05,
+      "loss": 0.6121,
+      "step": 5078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7689534850188887e-05,
+      "loss": 0.4497,
+      "step": 5079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7688958099028175e-05,
+      "loss": 0.5364,
+      "step": 5080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7688381347867466e-05,
+      "loss": 0.6145,
+      "step": 5081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7687804596706754e-05,
+      "loss": 0.7429,
+      "step": 5082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.768722784554604e-05,
+      "loss": 0.6537,
+      "step": 5083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.768665109438533e-05,
+      "loss": 0.6027,
+      "step": 5084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7686074343224617e-05,
+      "loss": 0.6017,
+      "step": 5085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7685497592063904e-05,
+      "loss": 0.5222,
+      "step": 5086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7684920840903196e-05,
+      "loss": 0.5996,
+      "step": 5087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7684344089742483e-05,
+      "loss": 0.5388,
+      "step": 5088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.768376733858177e-05,
+      "loss": 0.5414,
+      "step": 5089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.768319058742106e-05,
+      "loss": 0.4683,
+      "step": 5090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7682613836260346e-05,
+      "loss": 0.6073,
+      "step": 5091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7682037085099634e-05,
+      "loss": 0.5913,
+      "step": 5092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7681460333938925e-05,
+      "loss": 0.6147,
+      "step": 5093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7680883582778213e-05,
+      "loss": 0.5978,
+      "step": 5094
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76803068316175e-05,
+      "loss": 0.5894,
+      "step": 5095
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7679730080456788e-05,
+      "loss": 0.6052,
+      "step": 5096
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7679153329296076e-05,
+      "loss": 0.5282,
+      "step": 5097
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7678576578135363e-05,
+      "loss": 0.4526,
+      "step": 5098
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.767799982697465e-05,
+      "loss": 0.533,
+      "step": 5099
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7677423075813942e-05,
+      "loss": 0.697,
+      "step": 5100
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.767684632465323e-05,
+      "loss": 0.6176,
+      "step": 5101
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7676269573492518e-05,
+      "loss": 0.5009,
+      "step": 5102
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7675692822331805e-05,
+      "loss": 0.6126,
+      "step": 5103
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7675116071171093e-05,
+      "loss": 0.576,
+      "step": 5104
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.767453932001038e-05,
+      "loss": 0.5441,
+      "step": 5105
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7673962568849672e-05,
+      "loss": 0.6549,
+      "step": 5106
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.767338581768896e-05,
+      "loss": 0.6951,
+      "step": 5107
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7672809066528247e-05,
+      "loss": 0.6694,
+      "step": 5108
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7672232315367538e-05,
+      "loss": 0.5235,
+      "step": 5109
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7671655564206826e-05,
+      "loss": 0.4873,
+      "step": 5110
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7671078813046114e-05,
+      "loss": 0.4691,
+      "step": 5111
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76705020618854e-05,
+      "loss": 0.7004,
+      "step": 5112
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.766992531072469e-05,
+      "loss": 0.4766,
+      "step": 5113
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7669348559563977e-05,
+      "loss": 0.4529,
+      "step": 5114
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7668771808403268e-05,
+      "loss": 0.6568,
+      "step": 5115
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7668195057242555e-05,
+      "loss": 0.5176,
+      "step": 5116
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7667618306081843e-05,
+      "loss": 0.5845,
+      "step": 5117
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.766704155492113e-05,
+      "loss": 0.4744,
+      "step": 5118
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.766646480376042e-05,
+      "loss": 0.6735,
+      "step": 5119
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7665888052599706e-05,
+      "loss": 0.6786,
+      "step": 5120
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7665311301438997e-05,
+      "loss": 0.5997,
+      "step": 5121
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7664734550278285e-05,
+      "loss": 0.4526,
+      "step": 5122
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7664157799117573e-05,
+      "loss": 0.6329,
+      "step": 5123
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.766358104795686e-05,
+      "loss": 0.5316,
+      "step": 5124
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7663004296796148e-05,
+      "loss": 0.5266,
+      "step": 5125
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7662427545635436e-05,
+      "loss": 0.5817,
+      "step": 5126
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7661850794474727e-05,
+      "loss": 0.625,
+      "step": 5127
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7661274043314014e-05,
+      "loss": 0.5673,
+      "step": 5128
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7660697292153302e-05,
+      "loss": 0.6083,
+      "step": 5129
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.766012054099259e-05,
+      "loss": 0.5173,
+      "step": 5130
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7659543789831877e-05,
+      "loss": 0.596,
+      "step": 5131
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7658967038671165e-05,
+      "loss": 0.635,
+      "step": 5132
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7658390287510453e-05,
+      "loss": 0.6223,
+      "step": 5133
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7657813536349744e-05,
+      "loss": 0.639,
+      "step": 5134
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765723678518903e-05,
+      "loss": 0.4497,
+      "step": 5135
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765666003402832e-05,
+      "loss": 0.6497,
+      "step": 5136
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765608328286761e-05,
+      "loss": 0.5687,
+      "step": 5137
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7655506531706898e-05,
+      "loss": 0.608,
+      "step": 5138
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7654929780546182e-05,
+      "loss": 0.5364,
+      "step": 5139
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7654353029385473e-05,
+      "loss": 0.5036,
+      "step": 5140
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765377627822476e-05,
+      "loss": 0.5544,
+      "step": 5141
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765319952706405e-05,
+      "loss": 0.4806,
+      "step": 5142
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765262277590334e-05,
+      "loss": 0.5245,
+      "step": 5143
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7652046024742627e-05,
+      "loss": 0.6939,
+      "step": 5144
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7651469273581915e-05,
+      "loss": 0.5686,
+      "step": 5145
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7650892522421203e-05,
+      "loss": 0.5878,
+      "step": 5146
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.765031577126049e-05,
+      "loss": 0.6286,
+      "step": 5147
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7649739020099778e-05,
+      "loss": 0.4865,
+      "step": 5148
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.764916226893907e-05,
+      "loss": 0.6109,
+      "step": 5149
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7648585517778357e-05,
+      "loss": 0.5796,
+      "step": 5150
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7648008766617645e-05,
+      "loss": 0.5591,
+      "step": 5151
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7647432015456932e-05,
+      "loss": 0.5944,
+      "step": 5152
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.764685526429622e-05,
+      "loss": 0.6108,
+      "step": 5153
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7646278513135508e-05,
+      "loss": 0.5757,
+      "step": 5154
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76457017619748e-05,
+      "loss": 0.6753,
+      "step": 5155
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7645125010814086e-05,
+      "loss": 0.5824,
+      "step": 5156
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7644548259653374e-05,
+      "loss": 0.5953,
+      "step": 5157
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7643971508492662e-05,
+      "loss": 0.6087,
+      "step": 5158
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.764339475733195e-05,
+      "loss": 0.6289,
+      "step": 5159
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7642818006171237e-05,
+      "loss": 0.5705,
+      "step": 5160
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7642241255010528e-05,
+      "loss": 0.5946,
+      "step": 5161
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7641664503849816e-05,
+      "loss": 0.6368,
+      "step": 5162
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7641087752689104e-05,
+      "loss": 0.6436,
+      "step": 5163
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.764051100152839e-05,
+      "loss": 0.542,
+      "step": 5164
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763993425036768e-05,
+      "loss": 0.4492,
+      "step": 5165
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7639357499206967e-05,
+      "loss": 0.5145,
+      "step": 5166
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7638780748046254e-05,
+      "loss": 0.5077,
+      "step": 5167
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7638203996885545e-05,
+      "loss": 0.431,
+      "step": 5168
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7637627245724833e-05,
+      "loss": 0.5806,
+      "step": 5169
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763705049456412e-05,
+      "loss": 0.5074,
+      "step": 5170
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7636473743403412e-05,
+      "loss": 0.5553,
+      "step": 5171
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76358969922427e-05,
+      "loss": 0.6544,
+      "step": 5172
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7635320241081987e-05,
+      "loss": 0.4491,
+      "step": 5173
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7634743489921275e-05,
+      "loss": 0.4829,
+      "step": 5174
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7634166738760563e-05,
+      "loss": 0.7054,
+      "step": 5175
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763358998759985e-05,
+      "loss": 0.448,
+      "step": 5176
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763301323643914e-05,
+      "loss": 0.6162,
+      "step": 5177
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763243648527843e-05,
+      "loss": 0.6255,
+      "step": 5178
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7631859734117717e-05,
+      "loss": 0.5332,
+      "step": 5179
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7631282982957004e-05,
+      "loss": 0.6166,
+      "step": 5180
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7630706231796292e-05,
+      "loss": 0.6887,
+      "step": 5181
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.763012948063558e-05,
+      "loss": 0.5589,
+      "step": 5182
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762955272947487e-05,
+      "loss": 0.486,
+      "step": 5183
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762897597831416e-05,
+      "loss": 0.5193,
+      "step": 5184
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7628399227153446e-05,
+      "loss": 0.5726,
+      "step": 5185
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7627822475992734e-05,
+      "loss": 0.5426,
+      "step": 5186
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762724572483202e-05,
+      "loss": 0.5459,
+      "step": 5187
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762666897367131e-05,
+      "loss": 0.6965,
+      "step": 5188
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76260922225106e-05,
+      "loss": 0.5275,
+      "step": 5189
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7625515471349888e-05,
+      "loss": 0.6213,
+      "step": 5190
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7624938720189176e-05,
+      "loss": 0.4695,
+      "step": 5191
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7624361969028463e-05,
+      "loss": 0.411,
+      "step": 5192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762378521786775e-05,
+      "loss": 0.487,
+      "step": 5193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762320846670704e-05,
+      "loss": 0.6156,
+      "step": 5194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.762263171554633e-05,
+      "loss": 0.5095,
+      "step": 5195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7622054964385618e-05,
+      "loss": 0.5785,
+      "step": 5196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7621478213224905e-05,
+      "loss": 0.6199,
+      "step": 5197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7620901462064196e-05,
+      "loss": 0.5916,
+      "step": 5198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7620324710903484e-05,
+      "loss": 0.5928,
+      "step": 5199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7619747959742772e-05,
+      "loss": 0.6727,
+      "step": 5200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7619171208582056e-05,
+      "loss": 0.5301,
+      "step": 5201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7618594457421347e-05,
+      "loss": 0.6184,
+      "step": 5202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7618017706260635e-05,
+      "loss": 0.6497,
+      "step": 5203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7617440955099922e-05,
+      "loss": 0.5972,
+      "step": 5204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7616864203939214e-05,
+      "loss": 0.4998,
+      "step": 5205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.76162874527785e-05,
+      "loss": 0.5564,
+      "step": 5206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.761571070161779e-05,
+      "loss": 0.6234,
+      "step": 5207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7615133950457077e-05,
+      "loss": 0.5546,
+      "step": 5208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7614557199296364e-05,
+      "loss": 0.6103,
+      "step": 5209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7613980448135652e-05,
+      "loss": 0.6811,
+      "step": 5210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7613403696974943e-05,
+      "loss": 0.6973,
+      "step": 5211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.761282694581423e-05,
+      "loss": 0.6786,
+      "step": 5212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.761225019465352e-05,
+      "loss": 0.4981,
+      "step": 5213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7611673443492806e-05,
+      "loss": 0.6405,
+      "step": 5214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7611096692332094e-05,
+      "loss": 0.6652,
+      "step": 5215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.761051994117138e-05,
+      "loss": 0.6061,
+      "step": 5216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7609943190010673e-05,
+      "loss": 0.5536,
+      "step": 5217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760936643884996e-05,
+      "loss": 0.6129,
+      "step": 5218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7608789687689248e-05,
+      "loss": 0.6391,
+      "step": 5219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7608212936528536e-05,
+      "loss": 0.5239,
+      "step": 5220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7607636185367823e-05,
+      "loss": 0.6277,
+      "step": 5221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760705943420711e-05,
+      "loss": 0.6279,
+      "step": 5222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7606482683046402e-05,
+      "loss": 0.6377,
+      "step": 5223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760590593188569e-05,
+      "loss": 0.5759,
+      "step": 5224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7605329180724977e-05,
+      "loss": 0.5521,
+      "step": 5225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760475242956427e-05,
+      "loss": 0.6015,
+      "step": 5226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7604175678403553e-05,
+      "loss": 0.6246,
+      "step": 5227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760359892724284e-05,
+      "loss": 0.622,
+      "step": 5228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760302217608213e-05,
+      "loss": 0.5548,
+      "step": 5229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.760244542492142e-05,
+      "loss": 0.5067,
+      "step": 5230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7601868673760707e-05,
+      "loss": 0.6051,
+      "step": 5231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7601291922599998e-05,
+      "loss": 0.513,
+      "step": 5232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7600715171439286e-05,
+      "loss": 0.7004,
+      "step": 5233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7600138420278573e-05,
+      "loss": 0.5001,
+      "step": 5234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.759956166911786e-05,
+      "loss": 0.4997,
+      "step": 5235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.759898491795715e-05,
+      "loss": 0.5819,
+      "step": 5236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7598408166796436e-05,
+      "loss": 0.5331,
+      "step": 5237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7597831415635724e-05,
+      "loss": 0.4756,
+      "step": 5238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7597254664475015e-05,
+      "loss": 0.5721,
+      "step": 5239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7596677913314303e-05,
+      "loss": 0.5349,
+      "step": 5240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.759610116215359e-05,
+      "loss": 0.5628,
+      "step": 5241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7595524410992878e-05,
+      "loss": 0.6081,
+      "step": 5242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7594947659832166e-05,
+      "loss": 0.6064,
+      "step": 5243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7594370908671454e-05,
+      "loss": 0.6661,
+      "step": 5244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7593794157510745e-05,
+      "loss": 0.5877,
+      "step": 5245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7593217406350032e-05,
+      "loss": 0.5566,
+      "step": 5246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.759264065518932e-05,
+      "loss": 0.658,
+      "step": 5247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7592063904028608e-05,
+      "loss": 0.5275,
+      "step": 5248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7591487152867895e-05,
+      "loss": 0.4866,
+      "step": 5249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7590910401707183e-05,
+      "loss": 0.4938,
+      "step": 5250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7590333650546474e-05,
+      "loss": 0.4869,
+      "step": 5251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7589756899385762e-05,
+      "loss": 0.5399,
+      "step": 5252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758918014822505e-05,
+      "loss": 0.4521,
+      "step": 5253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7588603397064337e-05,
+      "loss": 0.6139,
+      "step": 5254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7588026645903625e-05,
+      "loss": 0.6024,
+      "step": 5255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7587449894742913e-05,
+      "loss": 0.534,
+      "step": 5256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7586873143582204e-05,
+      "loss": 0.8513,
+      "step": 5257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758629639242149e-05,
+      "loss": 0.5808,
+      "step": 5258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758571964126078e-05,
+      "loss": 0.5338,
+      "step": 5259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758514289010007e-05,
+      "loss": 0.6065,
+      "step": 5260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7584566138939358e-05,
+      "loss": 0.5757,
+      "step": 5261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7583989387778645e-05,
+      "loss": 0.6035,
+      "step": 5262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7583412636617933e-05,
+      "loss": 0.5371,
+      "step": 5263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758283588545722e-05,
+      "loss": 0.6503,
+      "step": 5264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.758225913429651e-05,
+      "loss": 0.5764,
+      "step": 5265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.75816823831358e-05,
+      "loss": 0.5858,
+      "step": 5266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7581105631975087e-05,
+      "loss": 0.4848,
+      "step": 5267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7580528880814375e-05,
+      "loss": 0.6115,
+      "step": 5268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7579952129653663e-05,
+      "loss": 0.4741,
+      "step": 5269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.757937537849295e-05,
+      "loss": 0.713,
+      "step": 5270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7578798627332238e-05,
+      "loss": 0.5401,
+      "step": 5271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7578221876171526e-05,
+      "loss": 0.6164,
+      "step": 5272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.7577645125010817e-05,
+      "loss": 0.5829,
+      "step": 5273
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7577068373850104e-05,
+      "loss": 0.6099,
+      "step": 5274
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7576491622689392e-05,
+      "loss": 0.5225,
+      "step": 5275
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.757591487152868e-05,
+      "loss": 0.6315,
+      "step": 5276
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7575338120367968e-05,
+      "loss": 0.4437,
+      "step": 5277
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7574761369207255e-05,
+      "loss": 0.4586,
+      "step": 5278
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7574184618046546e-05,
+      "loss": 0.5305,
+      "step": 5279
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7573607866885834e-05,
+      "loss": 0.5826,
+      "step": 5280
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.757303111572512e-05,
+      "loss": 0.6099,
+      "step": 5281
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.757245436456441e-05,
+      "loss": 0.636,
+      "step": 5282
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7571877613403697e-05,
+      "loss": 0.5844,
+      "step": 5283
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7571300862242985e-05,
+      "loss": 0.6166,
+      "step": 5284
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7570724111082276e-05,
+      "loss": 0.4791,
+      "step": 5285
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7570147359921563e-05,
+      "loss": 0.5623,
+      "step": 5286
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756957060876085e-05,
+      "loss": 0.6054,
+      "step": 5287
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7568993857600142e-05,
+      "loss": 0.5784,
+      "step": 5288
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756841710643943e-05,
+      "loss": 0.4544,
+      "step": 5289
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7567840355278714e-05,
+      "loss": 0.6053,
+      "step": 5290
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7567263604118005e-05,
+      "loss": 0.544,
+      "step": 5291
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7566686852957293e-05,
+      "loss": 0.5273,
+      "step": 5292
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756611010179658e-05,
+      "loss": 0.5765,
+      "step": 5293
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7565533350635872e-05,
+      "loss": 0.5681,
+      "step": 5294
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756495659947516e-05,
+      "loss": 0.5826,
+      "step": 5295
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7564379848314447e-05,
+      "loss": 0.6198,
+      "step": 5296
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7563803097153735e-05,
+      "loss": 0.6884,
+      "step": 5297
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7563226345993022e-05,
+      "loss": 0.5322,
+      "step": 5298
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756264959483231e-05,
+      "loss": 0.4686,
+      "step": 5299
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.75620728436716e-05,
+      "loss": 0.6495,
+      "step": 5300
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.756149609251089e-05,
+      "loss": 0.4615,
+      "step": 5301
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7560919341350177e-05,
+      "loss": 0.584,
+      "step": 5302
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7560342590189464e-05,
+      "loss": 0.5895,
+      "step": 5303
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7559765839028752e-05,
+      "loss": 0.6356,
+      "step": 5304
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.755918908786804e-05,
+      "loss": 0.5106,
+      "step": 5305
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7558612336707327e-05,
+      "loss": 0.5549,
+      "step": 5306
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.755803558554662e-05,
+      "loss": 0.4928,
+      "step": 5307
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7557458834385906e-05,
+      "loss": 0.5151,
+      "step": 5308
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7556882083225194e-05,
+      "loss": 0.539,
+      "step": 5309
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.755630533206448e-05,
+      "loss": 0.6862,
+      "step": 5310
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.755572858090377e-05,
+      "loss": 0.4854,
+      "step": 5311
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7555151829743057e-05,
+      "loss": 0.5457,
+      "step": 5312
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7554575078582348e-05,
+      "loss": 0.5159,
+      "step": 5313
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7553998327421636e-05,
+      "loss": 0.5266,
+      "step": 5314
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7553421576260923e-05,
+      "loss": 0.4932,
+      "step": 5315
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.755284482510021e-05,
+      "loss": 0.6608,
+      "step": 5316
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.75522680739395e-05,
+      "loss": 0.555,
+      "step": 5317
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7551691322778786e-05,
+      "loss": 0.6007,
+      "step": 5318
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7551114571618077e-05,
+      "loss": 0.6926,
+      "step": 5319
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7550537820457365e-05,
+      "loss": 0.577,
+      "step": 5320
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7549961069296653e-05,
+      "loss": 0.5894,
+      "step": 5321
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7549384318135944e-05,
+      "loss": 0.5111,
+      "step": 5322
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.754880756697523e-05,
+      "loss": 0.6711,
+      "step": 5323
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.754823081581452e-05,
+      "loss": 0.5171,
+      "step": 5324
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7547654064653807e-05,
+      "loss": 0.7303,
+      "step": 5325
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7547077313493095e-05,
+      "loss": 0.5688,
+      "step": 5326
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7546500562332382e-05,
+      "loss": 0.7117,
+      "step": 5327
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7545923811171673e-05,
+      "loss": 0.5884,
+      "step": 5328
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.754534706001096e-05,
+      "loss": 0.6525,
+      "step": 5329
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.754477030885025e-05,
+      "loss": 0.5691,
+      "step": 5330
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7544193557689536e-05,
+      "loss": 0.5435,
+      "step": 5331
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7543616806528824e-05,
+      "loss": 0.5269,
+      "step": 5332
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7543040055368112e-05,
+      "loss": 0.4164,
+      "step": 5333
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7542463304207403e-05,
+      "loss": 0.5942,
+      "step": 5334
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.754188655304669e-05,
+      "loss": 0.5726,
+      "step": 5335
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7541309801885978e-05,
+      "loss": 0.6102,
+      "step": 5336
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7540733050725266e-05,
+      "loss": 0.5063,
+      "step": 5337
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7540156299564554e-05,
+      "loss": 0.5988,
+      "step": 5338
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753957954840384e-05,
+      "loss": 0.5435,
+      "step": 5339
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753900279724313e-05,
+      "loss": 0.5602,
+      "step": 5340
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753842604608242e-05,
+      "loss": 0.4856,
+      "step": 5341
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7537849294921708e-05,
+      "loss": 0.536,
+      "step": 5342
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7537272543760995e-05,
+      "loss": 0.6298,
+      "step": 5343
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7536695792600283e-05,
+      "loss": 0.4993,
+      "step": 5344
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753611904143957e-05,
+      "loss": 0.5522,
+      "step": 5345
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753554229027886e-05,
+      "loss": 0.6788,
+      "step": 5346
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753496553911815e-05,
+      "loss": 0.6264,
+      "step": 5347
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7534388787957437e-05,
+      "loss": 0.6281,
+      "step": 5348
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7533812036796725e-05,
+      "loss": 0.591,
+      "step": 5349
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7533235285636016e-05,
+      "loss": 0.4865,
+      "step": 5350
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7532658534475304e-05,
+      "loss": 0.6902,
+      "step": 5351
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7532081783314588e-05,
+      "loss": 0.5159,
+      "step": 5352
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.753150503215388e-05,
+      "loss": 0.5853,
+      "step": 5353
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7530928280993167e-05,
+      "loss": 0.534,
+      "step": 5354
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7530351529832454e-05,
+      "loss": 0.603,
+      "step": 5355
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7529774778671745e-05,
+      "loss": 0.6083,
+      "step": 5356
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7529198027511033e-05,
+      "loss": 0.5882,
+      "step": 5357
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.752862127635032e-05,
+      "loss": 0.6032,
+      "step": 5358
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.752804452518961e-05,
+      "loss": 0.6095,
+      "step": 5359
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7527467774028896e-05,
+      "loss": 0.7402,
+      "step": 5360
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7526891022868184e-05,
+      "loss": 0.5044,
+      "step": 5361
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7526314271707475e-05,
+      "loss": 0.488,
+      "step": 5362
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7525737520546763e-05,
+      "loss": 0.6837,
+      "step": 5363
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.752516076938605e-05,
+      "loss": 0.459,
+      "step": 5364
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7524584018225338e-05,
+      "loss": 0.6149,
+      "step": 5365
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7524007267064626e-05,
+      "loss": 0.5325,
+      "step": 5366
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7523430515903913e-05,
+      "loss": 0.5435,
+      "step": 5367
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7522853764743204e-05,
+      "loss": 0.5559,
+      "step": 5368
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7522277013582492e-05,
+      "loss": 0.6069,
+      "step": 5369
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.752170026242178e-05,
+      "loss": 0.461,
+      "step": 5370
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7521123511261068e-05,
+      "loss": 0.6551,
+      "step": 5371
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7520546760100355e-05,
+      "loss": 0.6038,
+      "step": 5372
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7519970008939643e-05,
+      "loss": 0.5855,
+      "step": 5373
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751939325777893e-05,
+      "loss": 0.5008,
+      "step": 5374
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751881650661822e-05,
+      "loss": 0.6745,
+      "step": 5375
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751823975545751e-05,
+      "loss": 0.5813,
+      "step": 5376
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7517663004296797e-05,
+      "loss": 0.5695,
+      "step": 5377
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7517086253136085e-05,
+      "loss": 0.6272,
+      "step": 5378
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7516509501975372e-05,
+      "loss": 0.601,
+      "step": 5379
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751593275081466e-05,
+      "loss": 0.5466,
+      "step": 5380
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751535599965395e-05,
+      "loss": 0.5363,
+      "step": 5381
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751477924849324e-05,
+      "loss": 0.5619,
+      "step": 5382
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7514202497332527e-05,
+      "loss": 0.5273,
+      "step": 5383
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7513625746171818e-05,
+      "loss": 0.588,
+      "step": 5384
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7513048995011105e-05,
+      "loss": 0.3733,
+      "step": 5385
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7512472243850393e-05,
+      "loss": 0.5848,
+      "step": 5386
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751189549268968e-05,
+      "loss": 0.6051,
+      "step": 5387
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.751131874152897e-05,
+      "loss": 0.5852,
+      "step": 5388
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7510741990368256e-05,
+      "loss": 0.5039,
+      "step": 5389
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7510165239207547e-05,
+      "loss": 0.5789,
+      "step": 5390
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7509588488046835e-05,
+      "loss": 0.5198,
+      "step": 5391
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7509011736886122e-05,
+      "loss": 0.5541,
+      "step": 5392
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.750843498572541e-05,
+      "loss": 0.4455,
+      "step": 5393
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7507858234564698e-05,
+      "loss": 0.5972,
+      "step": 5394
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7507281483403986e-05,
+      "loss": 0.6477,
+      "step": 5395
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7506704732243277e-05,
+      "loss": 0.6647,
+      "step": 5396
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7506127981082564e-05,
+      "loss": 0.5397,
+      "step": 5397
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7505551229921852e-05,
+      "loss": 0.5644,
+      "step": 5398
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.750497447876114e-05,
+      "loss": 0.5182,
+      "step": 5399
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7504397727600427e-05,
+      "loss": 0.7527,
+      "step": 5400
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7503820976439715e-05,
+      "loss": 0.5796,
+      "step": 5401
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7503244225279006e-05,
+      "loss": 0.6442,
+      "step": 5402
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7502667474118294e-05,
+      "loss": 0.6674,
+      "step": 5403
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.750209072295758e-05,
+      "loss": 0.5691,
+      "step": 5404
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.750151397179687e-05,
+      "loss": 0.5298,
+      "step": 5405
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7500937220636157e-05,
+      "loss": 0.5902,
+      "step": 5406
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7500360469475445e-05,
+      "loss": 0.5858,
+      "step": 5407
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7499783718314732e-05,
+      "loss": 0.4723,
+      "step": 5408
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7499206967154023e-05,
+      "loss": 0.4352,
+      "step": 5409
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749863021599331e-05,
+      "loss": 0.5998,
+      "step": 5410
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.74980534648326e-05,
+      "loss": 0.4408,
+      "step": 5411
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749747671367189e-05,
+      "loss": 0.6296,
+      "step": 5412
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7496899962511177e-05,
+      "loss": 0.5867,
+      "step": 5413
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7496323211350462e-05,
+      "loss": 0.5899,
+      "step": 5414
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7495746460189753e-05,
+      "loss": 0.6872,
+      "step": 5415
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749516970902904e-05,
+      "loss": 0.5262,
+      "step": 5416
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7494592957868328e-05,
+      "loss": 0.5225,
+      "step": 5417
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749401620670762e-05,
+      "loss": 0.5499,
+      "step": 5418
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7493439455546907e-05,
+      "loss": 0.5616,
+      "step": 5419
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7492862704386195e-05,
+      "loss": 0.5692,
+      "step": 5420
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7492285953225482e-05,
+      "loss": 0.6587,
+      "step": 5421
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749170920206477e-05,
+      "loss": 0.4674,
+      "step": 5422
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7491132450904058e-05,
+      "loss": 0.596,
+      "step": 5423
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.749055569974335e-05,
+      "loss": 0.5944,
+      "step": 5424
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7489978948582636e-05,
+      "loss": 0.4664,
+      "step": 5425
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7489402197421924e-05,
+      "loss": 0.7331,
+      "step": 5426
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7488825446261212e-05,
+      "loss": 0.5959,
+      "step": 5427
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.74882486951005e-05,
+      "loss": 0.5721,
+      "step": 5428
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7487671943939787e-05,
+      "loss": 0.5494,
+      "step": 5429
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7487095192779078e-05,
+      "loss": 0.485,
+      "step": 5430
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7486518441618366e-05,
+      "loss": 0.5345,
+      "step": 5431
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7485941690457654e-05,
+      "loss": 0.646,
+      "step": 5432
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.748536493929694e-05,
+      "loss": 0.5536,
+      "step": 5433
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.748478818813623e-05,
+      "loss": 0.5334,
+      "step": 5434
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7484211436975517e-05,
+      "loss": 0.5801,
+      "step": 5435
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7483634685814808e-05,
+      "loss": 0.628,
+      "step": 5436
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7483057934654095e-05,
+      "loss": 0.5338,
+      "step": 5437
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7482481183493383e-05,
+      "loss": 0.6172,
+      "step": 5438
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7481904432332674e-05,
+      "loss": 0.5353,
+      "step": 5439
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.748132768117196e-05,
+      "loss": 0.5215,
+      "step": 5440
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7480750930011246e-05,
+      "loss": 0.5151,
+      "step": 5441
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7480174178850537e-05,
+      "loss": 0.6346,
+      "step": 5442
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7479597427689825e-05,
+      "loss": 0.6788,
+      "step": 5443
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7479020676529113e-05,
+      "loss": 0.6696,
+      "step": 5444
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.74784439253684e-05,
+      "loss": 0.4685,
+      "step": 5445
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.747786717420769e-05,
+      "loss": 0.5748,
+      "step": 5446
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.747729042304698e-05,
+      "loss": 0.5878,
+      "step": 5447
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7476713671886267e-05,
+      "loss": 0.5683,
+      "step": 5448
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7476136920725554e-05,
+      "loss": 0.5345,
+      "step": 5449
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.7475560169564842e-05,
+      "loss": 0.5978,
+      "step": 5450
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.747498341840413e-05,
+      "loss": 0.7377,
+      "step": 5451
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.747440666724342e-05,
+      "loss": 0.5546,
+      "step": 5452
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.747382991608271e-05,
+      "loss": 0.5217,
+      "step": 5453
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7473253164921996e-05,
+      "loss": 0.6365,
+      "step": 5454
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7472676413761284e-05,
+      "loss": 0.6008,
+      "step": 5455
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.747209966260057e-05,
+      "loss": 0.6368,
+      "step": 5456
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.747152291143986e-05,
+      "loss": 0.5446,
+      "step": 5457
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.747094616027915e-05,
+      "loss": 0.7096,
+      "step": 5458
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7470369409118438e-05,
+      "loss": 0.6332,
+      "step": 5459
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7469792657957726e-05,
+      "loss": 0.5458,
+      "step": 5460
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7469215906797013e-05,
+      "loss": 0.5659,
+      "step": 5461
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.74686391556363e-05,
+      "loss": 0.471,
+      "step": 5462
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746806240447559e-05,
+      "loss": 0.5878,
+      "step": 5463
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746748565331488e-05,
+      "loss": 0.4835,
+      "step": 5464
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7466908902154168e-05,
+      "loss": 0.6863,
+      "step": 5465
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7466332150993455e-05,
+      "loss": 0.6119,
+      "step": 5466
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7465755399832743e-05,
+      "loss": 0.5453,
+      "step": 5467
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746517864867203e-05,
+      "loss": 0.5889,
+      "step": 5468
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7464601897511318e-05,
+      "loss": 0.6012,
+      "step": 5469
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746402514635061e-05,
+      "loss": 0.6356,
+      "step": 5470
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7463448395189897e-05,
+      "loss": 0.63,
+      "step": 5471
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7462871644029185e-05,
+      "loss": 0.609,
+      "step": 5472
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7462294892868476e-05,
+      "loss": 0.6718,
+      "step": 5473
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7461718141707763e-05,
+      "loss": 0.6172,
+      "step": 5474
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746114139054705e-05,
+      "loss": 0.5782,
+      "step": 5475
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.746056463938634e-05,
+      "loss": 0.5241,
+      "step": 5476
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7459987888225627e-05,
+      "loss": 0.563,
+      "step": 5477
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7459411137064914e-05,
+      "loss": 0.5851,
+      "step": 5478
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7458834385904202e-05,
+      "loss": 0.5332,
+      "step": 5479
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7458257634743493e-05,
+      "loss": 0.5464,
+      "step": 5480
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745768088358278e-05,
+      "loss": 0.6412,
+      "step": 5481
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745710413242207e-05,
+      "loss": 0.5462,
+      "step": 5482
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7456527381261356e-05,
+      "loss": 0.5402,
+      "step": 5483
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7455950630100644e-05,
+      "loss": 0.6476,
+      "step": 5484
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745537387893993e-05,
+      "loss": 0.566,
+      "step": 5485
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7454797127779222e-05,
+      "loss": 0.6704,
+      "step": 5486
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745422037661851e-05,
+      "loss": 0.5682,
+      "step": 5487
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7453643625457798e-05,
+      "loss": 0.6724,
+      "step": 5488
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7453066874297086e-05,
+      "loss": 0.4978,
+      "step": 5489
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7452490123136373e-05,
+      "loss": 0.7015,
+      "step": 5490
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745191337197566e-05,
+      "loss": 0.53,
+      "step": 5491
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7451336620814952e-05,
+      "loss": 0.4566,
+      "step": 5492
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.745075986965424e-05,
+      "loss": 0.589,
+      "step": 5493
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7450183118493527e-05,
+      "loss": 0.4941,
+      "step": 5494
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7449606367332815e-05,
+      "loss": 0.5528,
+      "step": 5495
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7449029616172103e-05,
+      "loss": 0.5879,
+      "step": 5496
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744845286501139e-05,
+      "loss": 0.586,
+      "step": 5497
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744787611385068e-05,
+      "loss": 0.6483,
+      "step": 5498
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744729936268997e-05,
+      "loss": 0.5453,
+      "step": 5499
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7446722611529257e-05,
+      "loss": 0.6375,
+      "step": 5500
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7446145860368548e-05,
+      "loss": 0.5583,
+      "step": 5501
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7445569109207832e-05,
+      "loss": 0.601,
+      "step": 5502
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744499235804712e-05,
+      "loss": 0.5617,
+      "step": 5503
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744441560688641e-05,
+      "loss": 0.5216,
+      "step": 5504
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.74438388557257e-05,
+      "loss": 0.5063,
+      "step": 5505
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7443262104564986e-05,
+      "loss": 0.6396,
+      "step": 5506
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7442685353404277e-05,
+      "loss": 0.8015,
+      "step": 5507
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7442108602243565e-05,
+      "loss": 0.5346,
+      "step": 5508
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7441531851082853e-05,
+      "loss": 0.6099,
+      "step": 5509
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.744095509992214e-05,
+      "loss": 0.5681,
+      "step": 5510
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7440378348761428e-05,
+      "loss": 0.6304,
+      "step": 5511
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7439801597600716e-05,
+      "loss": 0.6094,
+      "step": 5512
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7439224846440004e-05,
+      "loss": 0.5641,
+      "step": 5513
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7438648095279295e-05,
+      "loss": 0.5007,
+      "step": 5514
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7438071344118582e-05,
+      "loss": 0.5151,
+      "step": 5515
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.743749459295787e-05,
+      "loss": 0.4536,
+      "step": 5516
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7436917841797158e-05,
+      "loss": 0.5536,
+      "step": 5517
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7436341090636445e-05,
+      "loss": 0.4721,
+      "step": 5518
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7435764339475733e-05,
+      "loss": 0.5794,
+      "step": 5519
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7435187588315024e-05,
+      "loss": 0.562,
+      "step": 5520
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7434610837154312e-05,
+      "loss": 0.6607,
+      "step": 5521
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.74340340859936e-05,
+      "loss": 0.5299,
+      "step": 5522
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7433457334832887e-05,
+      "loss": 0.5403,
+      "step": 5523
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7432880583672175e-05,
+      "loss": 0.5123,
+      "step": 5524
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7432303832511463e-05,
+      "loss": 0.5946,
+      "step": 5525
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7431727081350754e-05,
+      "loss": 0.499,
+      "step": 5526
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.743115033019004e-05,
+      "loss": 0.485,
+      "step": 5527
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.743057357902933e-05,
+      "loss": 0.4601,
+      "step": 5528
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7429996827868617e-05,
+      "loss": 0.6045,
+      "step": 5529
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7429420076707904e-05,
+      "loss": 0.4464,
+      "step": 5530
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7428843325547192e-05,
+      "loss": 0.6545,
+      "step": 5531
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7428266574386483e-05,
+      "loss": 0.5167,
+      "step": 5532
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.742768982322577e-05,
+      "loss": 0.5707,
+      "step": 5533
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.742711307206506e-05,
+      "loss": 0.6126,
+      "step": 5534
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.742653632090435e-05,
+      "loss": 0.4659,
+      "step": 5535
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7425959569743637e-05,
+      "loss": 0.6077,
+      "step": 5536
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7425382818582925e-05,
+      "loss": 0.4806,
+      "step": 5537
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7424806067422213e-05,
+      "loss": 0.662,
+      "step": 5538
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.74242293162615e-05,
+      "loss": 0.4888,
+      "step": 5539
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7423652565100788e-05,
+      "loss": 0.5928,
+      "step": 5540
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.742307581394008e-05,
+      "loss": 0.5771,
+      "step": 5541
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7422499062779367e-05,
+      "loss": 0.5153,
+      "step": 5542
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7421922311618654e-05,
+      "loss": 0.5979,
+      "step": 5543
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7421345560457942e-05,
+      "loss": 0.5913,
+      "step": 5544
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.742076880929723e-05,
+      "loss": 0.5521,
+      "step": 5545
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7420192058136517e-05,
+      "loss": 0.4348,
+      "step": 5546
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7419615306975805e-05,
+      "loss": 0.7044,
+      "step": 5547
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7419038555815096e-05,
+      "loss": 0.4997,
+      "step": 5548
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7418461804654384e-05,
+      "loss": 0.4788,
+      "step": 5549
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.741788505349367e-05,
+      "loss": 0.5855,
+      "step": 5550
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.741730830233296e-05,
+      "loss": 0.5062,
+      "step": 5551
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7416731551172247e-05,
+      "loss": 0.6219,
+      "step": 5552
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7416154800011535e-05,
+      "loss": 0.4875,
+      "step": 5553
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7415578048850826e-05,
+      "loss": 0.6264,
+      "step": 5554
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7415001297690113e-05,
+      "loss": 0.5386,
+      "step": 5555
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.74144245465294e-05,
+      "loss": 0.6106,
+      "step": 5556
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.741384779536869e-05,
+      "loss": 0.6212,
+      "step": 5557
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7413271044207976e-05,
+      "loss": 0.5857,
+      "step": 5558
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7412694293047264e-05,
+      "loss": 0.6383,
+      "step": 5559
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7412117541886555e-05,
+      "loss": 0.5248,
+      "step": 5560
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7411540790725843e-05,
+      "loss": 0.5613,
+      "step": 5561
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.741096403956513e-05,
+      "loss": 0.4298,
+      "step": 5562
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.741038728840442e-05,
+      "loss": 0.5877,
+      "step": 5563
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7409810537243706e-05,
+      "loss": 0.4949,
+      "step": 5564
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7409233786082994e-05,
+      "loss": 0.5522,
+      "step": 5565
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7408657034922285e-05,
+      "loss": 0.6559,
+      "step": 5566
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7408080283761572e-05,
+      "loss": 0.4477,
+      "step": 5567
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740750353260086e-05,
+      "loss": 0.5587,
+      "step": 5568
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740692678144015e-05,
+      "loss": 0.5555,
+      "step": 5569
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740635003027944e-05,
+      "loss": 0.5808,
+      "step": 5570
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7405773279118727e-05,
+      "loss": 0.5791,
+      "step": 5571
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7405196527958014e-05,
+      "loss": 0.6212,
+      "step": 5572
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7404619776797302e-05,
+      "loss": 0.5001,
+      "step": 5573
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740404302563659e-05,
+      "loss": 0.6287,
+      "step": 5574
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740346627447588e-05,
+      "loss": 0.5457,
+      "step": 5575
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740288952331517e-05,
+      "loss": 0.4722,
+      "step": 5576
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7402312772154456e-05,
+      "loss": 0.5556,
+      "step": 5577
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7401736020993744e-05,
+      "loss": 0.5749,
+      "step": 5578
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740115926983303e-05,
+      "loss": 0.5734,
+      "step": 5579
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740058251867232e-05,
+      "loss": 0.5561,
+      "step": 5580
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.740000576751161e-05,
+      "loss": 0.5,
+      "step": 5581
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7399429016350898e-05,
+      "loss": 0.6161,
+      "step": 5582
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7398852265190186e-05,
+      "loss": 0.5816,
+      "step": 5583
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7398275514029473e-05,
+      "loss": 0.6355,
+      "step": 5584
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.739769876286876e-05,
+      "loss": 0.5552,
+      "step": 5585
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.739712201170805e-05,
+      "loss": 0.5481,
+      "step": 5586
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7396545260547336e-05,
+      "loss": 0.4769,
+      "step": 5587
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7395968509386627e-05,
+      "loss": 0.4401,
+      "step": 5588
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7395391758225915e-05,
+      "loss": 0.6642,
+      "step": 5589
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7394815007065203e-05,
+      "loss": 0.6104,
+      "step": 5590
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.739423825590449e-05,
+      "loss": 0.435,
+      "step": 5591
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7393661504743778e-05,
+      "loss": 0.5141,
+      "step": 5592
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7393084753583066e-05,
+      "loss": 0.5817,
+      "step": 5593
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7392508002422357e-05,
+      "loss": 0.5634,
+      "step": 5594
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7391931251261645e-05,
+      "loss": 0.6781,
+      "step": 5595
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7391354500100932e-05,
+      "loss": 0.4629,
+      "step": 5596
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7390777748940223e-05,
+      "loss": 0.6161,
+      "step": 5597
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.739020099777951e-05,
+      "loss": 0.5096,
+      "step": 5598
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.73896242466188e-05,
+      "loss": 0.556,
+      "step": 5599
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7389047495458086e-05,
+      "loss": 0.5114,
+      "step": 5600
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7388470744297374e-05,
+      "loss": 0.5429,
+      "step": 5601
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.738789399313666e-05,
+      "loss": 0.5482,
+      "step": 5602
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7387317241975953e-05,
+      "loss": 0.5921,
+      "step": 5603
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.738674049081524e-05,
+      "loss": 0.4692,
+      "step": 5604
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7386163739654528e-05,
+      "loss": 0.5825,
+      "step": 5605
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7385586988493816e-05,
+      "loss": 0.4602,
+      "step": 5606
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7385010237333104e-05,
+      "loss": 0.4125,
+      "step": 5607
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.738443348617239e-05,
+      "loss": 0.53,
+      "step": 5608
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7383856735011682e-05,
+      "loss": 0.6558,
+      "step": 5609
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.738327998385097e-05,
+      "loss": 0.5984,
+      "step": 5610
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7382703232690258e-05,
+      "loss": 0.526,
+      "step": 5611
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7382126481529545e-05,
+      "loss": 0.5119,
+      "step": 5612
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7381549730368833e-05,
+      "loss": 0.5535,
+      "step": 5613
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.738097297920812e-05,
+      "loss": 0.5023,
+      "step": 5614
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7380396228047412e-05,
+      "loss": 0.5339,
+      "step": 5615
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.73798194768867e-05,
+      "loss": 0.4648,
+      "step": 5616
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7379242725725987e-05,
+      "loss": 0.609,
+      "step": 5617
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7378665974565275e-05,
+      "loss": 0.6062,
+      "step": 5618
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7378089223404563e-05,
+      "loss": 0.6465,
+      "step": 5619
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.737751247224385e-05,
+      "loss": 0.5238,
+      "step": 5620
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7376935721083138e-05,
+      "loss": 0.4291,
+      "step": 5621
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.737635896992243e-05,
+      "loss": 0.6302,
+      "step": 5622
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7375782218761717e-05,
+      "loss": 0.5146,
+      "step": 5623
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7375205467601004e-05,
+      "loss": 0.3983,
+      "step": 5624
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7374628716440295e-05,
+      "loss": 0.5528,
+      "step": 5625
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.737405196527958e-05,
+      "loss": 0.6429,
+      "step": 5626
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7373475214118867e-05,
+      "loss": 0.6164,
+      "step": 5627
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.737289846295816e-05,
+      "loss": 0.5599,
+      "step": 5628
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7372321711797446e-05,
+      "loss": 0.456,
+      "step": 5629
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.7371744960636734e-05,
+      "loss": 0.4641,
+      "step": 5630
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7371168209476025e-05,
+      "loss": 0.5498,
+      "step": 5631
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7370591458315313e-05,
+      "loss": 0.5504,
+      "step": 5632
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.73700147071546e-05,
+      "loss": 0.5242,
+      "step": 5633
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7369437955993888e-05,
+      "loss": 0.6228,
+      "step": 5634
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7368861204833176e-05,
+      "loss": 0.5951,
+      "step": 5635
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7368284453672463e-05,
+      "loss": 0.611,
+      "step": 5636
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7367707702511754e-05,
+      "loss": 0.4117,
+      "step": 5637
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7367130951351042e-05,
+      "loss": 0.6133,
+      "step": 5638
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.736655420019033e-05,
+      "loss": 0.6725,
+      "step": 5639
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7365977449029617e-05,
+      "loss": 0.5213,
+      "step": 5640
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7365400697868905e-05,
+      "loss": 0.5938,
+      "step": 5641
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7364823946708193e-05,
+      "loss": 0.5184,
+      "step": 5642
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7364247195547484e-05,
+      "loss": 0.6037,
+      "step": 5643
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.736367044438677e-05,
+      "loss": 0.6422,
+      "step": 5644
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.736309369322606e-05,
+      "loss": 0.5865,
+      "step": 5645
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7362516942065347e-05,
+      "loss": 0.6523,
+      "step": 5646
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7361940190904635e-05,
+      "loss": 0.6107,
+      "step": 5647
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7361363439743922e-05,
+      "loss": 0.6652,
+      "step": 5648
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7360786688583213e-05,
+      "loss": 0.5305,
+      "step": 5649
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.73602099374225e-05,
+      "loss": 0.6162,
+      "step": 5650
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.735963318626179e-05,
+      "loss": 0.62,
+      "step": 5651
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7359056435101076e-05,
+      "loss": 0.5734,
+      "step": 5652
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7358479683940364e-05,
+      "loss": 0.4552,
+      "step": 5653
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7357902932779652e-05,
+      "loss": 0.5998,
+      "step": 5654
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.735732618161894e-05,
+      "loss": 0.5641,
+      "step": 5655
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.735674943045823e-05,
+      "loss": 0.53,
+      "step": 5656
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7356172679297518e-05,
+      "loss": 0.4263,
+      "step": 5657
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7355595928136806e-05,
+      "loss": 0.6053,
+      "step": 5658
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7355019176976097e-05,
+      "loss": 0.4748,
+      "step": 5659
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7354442425815385e-05,
+      "loss": 0.5341,
+      "step": 5660
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7353865674654672e-05,
+      "loss": 0.5775,
+      "step": 5661
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.735328892349396e-05,
+      "loss": 0.5053,
+      "step": 5662
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7352712172333248e-05,
+      "loss": 0.6372,
+      "step": 5663
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7352135421172535e-05,
+      "loss": 0.5028,
+      "step": 5664
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7351558670011827e-05,
+      "loss": 0.6769,
+      "step": 5665
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7350981918851114e-05,
+      "loss": 0.5249,
+      "step": 5666
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7350405167690402e-05,
+      "loss": 0.6583,
+      "step": 5667
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734982841652969e-05,
+      "loss": 0.4565,
+      "step": 5668
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7349251665368977e-05,
+      "loss": 0.6188,
+      "step": 5669
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7348674914208265e-05,
+      "loss": 0.5444,
+      "step": 5670
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7348098163047556e-05,
+      "loss": 0.5861,
+      "step": 5671
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7347521411886844e-05,
+      "loss": 0.4814,
+      "step": 5672
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734694466072613e-05,
+      "loss": 0.5974,
+      "step": 5673
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734636790956542e-05,
+      "loss": 0.5535,
+      "step": 5674
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7345791158404707e-05,
+      "loss": 0.6425,
+      "step": 5675
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7345214407243994e-05,
+      "loss": 0.5163,
+      "step": 5676
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7344637656083286e-05,
+      "loss": 0.6069,
+      "step": 5677
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7344060904922573e-05,
+      "loss": 0.5185,
+      "step": 5678
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734348415376186e-05,
+      "loss": 0.5315,
+      "step": 5679
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734290740260115e-05,
+      "loss": 0.5116,
+      "step": 5680
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7342330651440436e-05,
+      "loss": 0.6212,
+      "step": 5681
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7341753900279724e-05,
+      "loss": 0.632,
+      "step": 5682
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7341177149119015e-05,
+      "loss": 0.5371,
+      "step": 5683
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7340600397958303e-05,
+      "loss": 0.7018,
+      "step": 5684
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.734002364679759e-05,
+      "loss": 0.5775,
+      "step": 5685
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733944689563688e-05,
+      "loss": 0.564,
+      "step": 5686
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733887014447617e-05,
+      "loss": 0.6042,
+      "step": 5687
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7338293393315457e-05,
+      "loss": 0.5381,
+      "step": 5688
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733771664215474e-05,
+      "loss": 0.517,
+      "step": 5689
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7337139890994032e-05,
+      "loss": 0.5172,
+      "step": 5690
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733656313983332e-05,
+      "loss": 0.6243,
+      "step": 5691
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7335986388672608e-05,
+      "loss": 0.6279,
+      "step": 5692
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.73354096375119e-05,
+      "loss": 0.6763,
+      "step": 5693
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7334832886351186e-05,
+      "loss": 0.6968,
+      "step": 5694
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7334256135190474e-05,
+      "loss": 0.6147,
+      "step": 5695
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733367938402976e-05,
+      "loss": 0.5013,
+      "step": 5696
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733310263286905e-05,
+      "loss": 0.5704,
+      "step": 5697
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7332525881708337e-05,
+      "loss": 0.5748,
+      "step": 5698
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7331949130547628e-05,
+      "loss": 0.5943,
+      "step": 5699
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7331372379386916e-05,
+      "loss": 0.5501,
+      "step": 5700
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7330795628226204e-05,
+      "loss": 0.6951,
+      "step": 5701
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.733021887706549e-05,
+      "loss": 0.6847,
+      "step": 5702
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.732964212590478e-05,
+      "loss": 0.5586,
+      "step": 5703
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7329065374744067e-05,
+      "loss": 0.6139,
+      "step": 5704
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7328488623583358e-05,
+      "loss": 0.4788,
+      "step": 5705
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7327911872422645e-05,
+      "loss": 0.6068,
+      "step": 5706
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7327335121261933e-05,
+      "loss": 0.4569,
+      "step": 5707
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.732675837010122e-05,
+      "loss": 0.5673,
+      "step": 5708
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.732618161894051e-05,
+      "loss": 0.5639,
+      "step": 5709
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7325604867779796e-05,
+      "loss": 0.5298,
+      "step": 5710
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7325028116619087e-05,
+      "loss": 0.7138,
+      "step": 5711
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7324451365458375e-05,
+      "loss": 0.4237,
+      "step": 5712
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7323874614297662e-05,
+      "loss": 0.492,
+      "step": 5713
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7323297863136954e-05,
+      "loss": 0.426,
+      "step": 5714
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7322721111976238e-05,
+      "loss": 0.5128,
+      "step": 5715
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7322144360815526e-05,
+      "loss": 0.5127,
+      "step": 5716
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7321567609654817e-05,
+      "loss": 0.56,
+      "step": 5717
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7320990858494104e-05,
+      "loss": 0.599,
+      "step": 5718
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7320414107333392e-05,
+      "loss": 0.6741,
+      "step": 5719
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7319837356172683e-05,
+      "loss": 0.5865,
+      "step": 5720
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731926060501197e-05,
+      "loss": 0.5067,
+      "step": 5721
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731868385385126e-05,
+      "loss": 0.5846,
+      "step": 5722
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7318107102690546e-05,
+      "loss": 0.4186,
+      "step": 5723
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7317530351529834e-05,
+      "loss": 0.5518,
+      "step": 5724
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731695360036912e-05,
+      "loss": 0.5935,
+      "step": 5725
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731637684920841e-05,
+      "loss": 0.6315,
+      "step": 5726
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.73158000980477e-05,
+      "loss": 0.5899,
+      "step": 5727
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7315223346886988e-05,
+      "loss": 0.5624,
+      "step": 5728
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7314646595726276e-05,
+      "loss": 0.3642,
+      "step": 5729
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7314069844565563e-05,
+      "loss": 0.5104,
+      "step": 5730
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731349309340485e-05,
+      "loss": 0.6572,
+      "step": 5731
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731291634224414e-05,
+      "loss": 0.4812,
+      "step": 5732
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731233959108343e-05,
+      "loss": 0.5133,
+      "step": 5733
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7311762839922717e-05,
+      "loss": 0.5883,
+      "step": 5734
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7311186088762005e-05,
+      "loss": 0.5486,
+      "step": 5735
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7310609337601293e-05,
+      "loss": 0.5995,
+      "step": 5736
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.731003258644058e-05,
+      "loss": 0.569,
+      "step": 5737
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7309455835279868e-05,
+      "loss": 0.4826,
+      "step": 5738
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.730887908411916e-05,
+      "loss": 0.6475,
+      "step": 5739
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7308302332958447e-05,
+      "loss": 0.5052,
+      "step": 5740
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7307725581797735e-05,
+      "loss": 0.6009,
+      "step": 5741
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7307148830637022e-05,
+      "loss": 0.5308,
+      "step": 5742
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.730657207947631e-05,
+      "loss": 0.6051,
+      "step": 5743
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7305995328315598e-05,
+      "loss": 0.567,
+      "step": 5744
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.730541857715489e-05,
+      "loss": 0.4179,
+      "step": 5745
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7304841825994176e-05,
+      "loss": 0.5686,
+      "step": 5746
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7304265074833464e-05,
+      "loss": 0.5724,
+      "step": 5747
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7303688323672755e-05,
+      "loss": 0.6057,
+      "step": 5748
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7303111572512043e-05,
+      "loss": 0.4653,
+      "step": 5749
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.730253482135133e-05,
+      "loss": 0.6198,
+      "step": 5750
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7301958070190618e-05,
+      "loss": 0.5671,
+      "step": 5751
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7301381319029906e-05,
+      "loss": 0.5901,
+      "step": 5752
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7300804567869194e-05,
+      "loss": 0.6105,
+      "step": 5753
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7300227816708485e-05,
+      "loss": 0.6219,
+      "step": 5754
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7299651065547772e-05,
+      "loss": 0.4796,
+      "step": 5755
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729907431438706e-05,
+      "loss": 0.4693,
+      "step": 5756
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7298497563226348e-05,
+      "loss": 0.6007,
+      "step": 5757
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7297920812065635e-05,
+      "loss": 0.5435,
+      "step": 5758
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7297344060904923e-05,
+      "loss": 0.6466,
+      "step": 5759
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729676730974421e-05,
+      "loss": 0.5039,
+      "step": 5760
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7296190558583502e-05,
+      "loss": 0.4133,
+      "step": 5761
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729561380742279e-05,
+      "loss": 0.4754,
+      "step": 5762
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7295037056262077e-05,
+      "loss": 0.5478,
+      "step": 5763
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7294460305101365e-05,
+      "loss": 0.4193,
+      "step": 5764
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7293883553940653e-05,
+      "loss": 0.5818,
+      "step": 5765
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729330680277994e-05,
+      "loss": 0.6013,
+      "step": 5766
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729273005161923e-05,
+      "loss": 0.5534,
+      "step": 5767
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.729215330045852e-05,
+      "loss": 0.5235,
+      "step": 5768
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7291576549297807e-05,
+      "loss": 0.6498,
+      "step": 5769
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7290999798137094e-05,
+      "loss": 0.5689,
+      "step": 5770
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7290423046976382e-05,
+      "loss": 0.5257,
+      "step": 5771
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728984629581567e-05,
+      "loss": 0.6562,
+      "step": 5772
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728926954465496e-05,
+      "loss": 0.4812,
+      "step": 5773
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728869279349425e-05,
+      "loss": 0.4996,
+      "step": 5774
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7288116042333536e-05,
+      "loss": 0.5374,
+      "step": 5775
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7287539291172827e-05,
+      "loss": 0.5854,
+      "step": 5776
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728696254001211e-05,
+      "loss": 0.6074,
+      "step": 5777
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.72863857888514e-05,
+      "loss": 0.6337,
+      "step": 5778
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728580903769069e-05,
+      "loss": 0.5698,
+      "step": 5779
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7285232286529978e-05,
+      "loss": 0.5216,
+      "step": 5780
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7284655535369266e-05,
+      "loss": 0.6488,
+      "step": 5781
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7284078784208557e-05,
+      "loss": 0.6136,
+      "step": 5782
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7283502033047845e-05,
+      "loss": 0.7001,
+      "step": 5783
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7282925281887132e-05,
+      "loss": 0.4755,
+      "step": 5784
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.728234853072642e-05,
+      "loss": 0.5826,
+      "step": 5785
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7281771779565708e-05,
+      "loss": 0.6505,
+      "step": 5786
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7281195028404995e-05,
+      "loss": 0.5344,
+      "step": 5787
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7280618277244286e-05,
+      "loss": 0.6399,
+      "step": 5788
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7280041526083574e-05,
+      "loss": 0.6388,
+      "step": 5789
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727946477492286e-05,
+      "loss": 0.5269,
+      "step": 5790
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727888802376215e-05,
+      "loss": 0.6122,
+      "step": 5791
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7278311272601437e-05,
+      "loss": 0.557,
+      "step": 5792
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7277734521440725e-05,
+      "loss": 0.5736,
+      "step": 5793
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7277157770280012e-05,
+      "loss": 0.4437,
+      "step": 5794
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7276581019119304e-05,
+      "loss": 0.5318,
+      "step": 5795
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727600426795859e-05,
+      "loss": 0.42,
+      "step": 5796
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727542751679788e-05,
+      "loss": 0.6892,
+      "step": 5797
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7274850765637167e-05,
+      "loss": 0.5434,
+      "step": 5798
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7274274014476454e-05,
+      "loss": 0.7546,
+      "step": 5799
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7273697263315742e-05,
+      "loss": 0.5732,
+      "step": 5800
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7273120512155033e-05,
+      "loss": 0.4759,
+      "step": 5801
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727254376099432e-05,
+      "loss": 0.612,
+      "step": 5802
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727196700983361e-05,
+      "loss": 0.557,
+      "step": 5803
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7271390258672896e-05,
+      "loss": 0.585,
+      "step": 5804
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7270813507512184e-05,
+      "loss": 0.5726,
+      "step": 5805
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.727023675635147e-05,
+      "loss": 0.5488,
+      "step": 5806
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7269660005190762e-05,
+      "loss": 0.5587,
+      "step": 5807
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.726908325403005e-05,
+      "loss": 0.5057,
+      "step": 5808
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.7268506502869338e-05,
+      "loss": 0.5627,
+      "step": 5809
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.726792975170863e-05,
+      "loss": 0.5616,
+      "step": 5810
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7267353000547917e-05,
+      "loss": 0.4878,
+      "step": 5811
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7266776249387204e-05,
+      "loss": 0.4257,
+      "step": 5812
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7266199498226492e-05,
+      "loss": 0.5278,
+      "step": 5813
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.726562274706578e-05,
+      "loss": 0.6164,
+      "step": 5814
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7265045995905067e-05,
+      "loss": 0.5094,
+      "step": 5815
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.726446924474436e-05,
+      "loss": 0.5678,
+      "step": 5816
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7263892493583646e-05,
+      "loss": 0.7002,
+      "step": 5817
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7263315742422934e-05,
+      "loss": 0.5139,
+      "step": 5818
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.726273899126222e-05,
+      "loss": 0.5616,
+      "step": 5819
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.726216224010151e-05,
+      "loss": 0.5163,
+      "step": 5820
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7261585488940797e-05,
+      "loss": 0.5685,
+      "step": 5821
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7261008737780088e-05,
+      "loss": 0.6754,
+      "step": 5822
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7260431986619376e-05,
+      "loss": 0.5291,
+      "step": 5823
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7259855235458663e-05,
+      "loss": 0.5463,
+      "step": 5824
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.725927848429795e-05,
+      "loss": 0.5627,
+      "step": 5825
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.725870173313724e-05,
+      "loss": 0.5695,
+      "step": 5826
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7258124981976526e-05,
+      "loss": 0.4572,
+      "step": 5827
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7257548230815814e-05,
+      "loss": 0.6048,
+      "step": 5828
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7256971479655105e-05,
+      "loss": 0.611,
+      "step": 5829
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7256394728494393e-05,
+      "loss": 0.5309,
+      "step": 5830
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.725581797733368e-05,
+      "loss": 0.6348,
+      "step": 5831
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7255241226172968e-05,
+      "loss": 0.476,
+      "step": 5832
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7254664475012256e-05,
+      "loss": 0.5218,
+      "step": 5833
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7254087723851544e-05,
+      "loss": 0.5848,
+      "step": 5834
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7253510972690835e-05,
+      "loss": 0.5432,
+      "step": 5835
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7252934221530122e-05,
+      "loss": 0.481,
+      "step": 5836
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.725235747036941e-05,
+      "loss": 0.5754,
+      "step": 5837
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.72517807192087e-05,
+      "loss": 0.4525,
+      "step": 5838
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7251203968047985e-05,
+      "loss": 0.5926,
+      "step": 5839
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7250627216887273e-05,
+      "loss": 0.5015,
+      "step": 5840
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7250050465726564e-05,
+      "loss": 0.497,
+      "step": 5841
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7249473714565852e-05,
+      "loss": 0.4661,
+      "step": 5842
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724889696340514e-05,
+      "loss": 0.6101,
+      "step": 5843
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724832021224443e-05,
+      "loss": 0.6274,
+      "step": 5844
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7247743461083718e-05,
+      "loss": 0.6671,
+      "step": 5845
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7247166709923006e-05,
+      "loss": 0.6231,
+      "step": 5846
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7246589958762294e-05,
+      "loss": 0.5675,
+      "step": 5847
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724601320760158e-05,
+      "loss": 0.5374,
+      "step": 5848
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724543645644087e-05,
+      "loss": 0.5699,
+      "step": 5849
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724485970528016e-05,
+      "loss": 0.6495,
+      "step": 5850
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7244282954119448e-05,
+      "loss": 0.5819,
+      "step": 5851
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7243706202958735e-05,
+      "loss": 0.5734,
+      "step": 5852
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7243129451798023e-05,
+      "loss": 0.5286,
+      "step": 5853
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724255270063731e-05,
+      "loss": 0.4512,
+      "step": 5854
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.72419759494766e-05,
+      "loss": 0.6042,
+      "step": 5855
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.724139919831589e-05,
+      "loss": 0.5053,
+      "step": 5856
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7240822447155177e-05,
+      "loss": 0.5563,
+      "step": 5857
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7240245695994465e-05,
+      "loss": 0.5843,
+      "step": 5858
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7239668944833753e-05,
+      "loss": 0.5629,
+      "step": 5859
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.723909219367304e-05,
+      "loss": 0.4754,
+      "step": 5860
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7238515442512328e-05,
+      "loss": 0.4734,
+      "step": 5861
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7237938691351616e-05,
+      "loss": 0.5485,
+      "step": 5862
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7237361940190907e-05,
+      "loss": 0.5516,
+      "step": 5863
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7236785189030194e-05,
+      "loss": 0.6059,
+      "step": 5864
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7236208437869482e-05,
+      "loss": 0.4744,
+      "step": 5865
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.723563168670877e-05,
+      "loss": 0.5747,
+      "step": 5866
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7235054935548057e-05,
+      "loss": 0.4922,
+      "step": 5867
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7234478184387345e-05,
+      "loss": 0.6346,
+      "step": 5868
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7233901433226636e-05,
+      "loss": 0.6361,
+      "step": 5869
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7233324682065924e-05,
+      "loss": 0.6036,
+      "step": 5870
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.723274793090521e-05,
+      "loss": 0.5393,
+      "step": 5871
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7232171179744503e-05,
+      "loss": 0.5377,
+      "step": 5872
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.723159442858379e-05,
+      "loss": 0.4643,
+      "step": 5873
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7231017677423078e-05,
+      "loss": 0.5607,
+      "step": 5874
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7230440926262366e-05,
+      "loss": 0.4811,
+      "step": 5875
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7229864175101653e-05,
+      "loss": 0.6253,
+      "step": 5876
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722928742394094e-05,
+      "loss": 0.7193,
+      "step": 5877
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7228710672780232e-05,
+      "loss": 0.474,
+      "step": 5878
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722813392161952e-05,
+      "loss": 0.4803,
+      "step": 5879
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7227557170458808e-05,
+      "loss": 0.5818,
+      "step": 5880
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7226980419298095e-05,
+      "loss": 0.5259,
+      "step": 5881
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7226403668137383e-05,
+      "loss": 0.5934,
+      "step": 5882
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722582691697667e-05,
+      "loss": 0.6664,
+      "step": 5883
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722525016581596e-05,
+      "loss": 0.5345,
+      "step": 5884
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722467341465525e-05,
+      "loss": 0.5565,
+      "step": 5885
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7224096663494537e-05,
+      "loss": 0.6847,
+      "step": 5886
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7223519912333825e-05,
+      "loss": 0.6149,
+      "step": 5887
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7222943161173112e-05,
+      "loss": 0.5313,
+      "step": 5888
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.72223664100124e-05,
+      "loss": 0.5756,
+      "step": 5889
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722178965885169e-05,
+      "loss": 0.4543,
+      "step": 5890
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.722121290769098e-05,
+      "loss": 0.5139,
+      "step": 5891
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7220636156530267e-05,
+      "loss": 0.5906,
+      "step": 5892
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7220059405369554e-05,
+      "loss": 0.6927,
+      "step": 5893
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7219482654208842e-05,
+      "loss": 0.5654,
+      "step": 5894
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.721890590304813e-05,
+      "loss": 0.5617,
+      "step": 5895
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7218329151887417e-05,
+      "loss": 0.706,
+      "step": 5896
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.721775240072671e-05,
+      "loss": 0.6446,
+      "step": 5897
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7217175649565996e-05,
+      "loss": 0.4988,
+      "step": 5898
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7216598898405284e-05,
+      "loss": 0.4508,
+      "step": 5899
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7216022147244575e-05,
+      "loss": 0.5649,
+      "step": 5900
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.721544539608386e-05,
+      "loss": 0.6066,
+      "step": 5901
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7214868644923147e-05,
+      "loss": 0.6821,
+      "step": 5902
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7214291893762438e-05,
+      "loss": 0.5647,
+      "step": 5903
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7213715142601726e-05,
+      "loss": 0.534,
+      "step": 5904
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7213138391441013e-05,
+      "loss": 0.518,
+      "step": 5905
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7212561640280304e-05,
+      "loss": 0.6494,
+      "step": 5906
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7211984889119592e-05,
+      "loss": 0.5202,
+      "step": 5907
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.721140813795888e-05,
+      "loss": 0.6933,
+      "step": 5908
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7210831386798167e-05,
+      "loss": 0.5511,
+      "step": 5909
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7210254635637455e-05,
+      "loss": 0.4744,
+      "step": 5910
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7209677884476743e-05,
+      "loss": 0.6049,
+      "step": 5911
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7209101133316034e-05,
+      "loss": 0.5873,
+      "step": 5912
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720852438215532e-05,
+      "loss": 0.5312,
+      "step": 5913
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720794763099461e-05,
+      "loss": 0.4443,
+      "step": 5914
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7207370879833897e-05,
+      "loss": 0.577,
+      "step": 5915
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7206794128673185e-05,
+      "loss": 0.5018,
+      "step": 5916
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7206217377512472e-05,
+      "loss": 0.5132,
+      "step": 5917
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7205640626351763e-05,
+      "loss": 0.5391,
+      "step": 5918
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720506387519105e-05,
+      "loss": 0.6012,
+      "step": 5919
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720448712403034e-05,
+      "loss": 0.577,
+      "step": 5920
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7203910372869626e-05,
+      "loss": 0.5699,
+      "step": 5921
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7203333621708914e-05,
+      "loss": 0.6014,
+      "step": 5922
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7202756870548202e-05,
+      "loss": 0.6464,
+      "step": 5923
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7202180119387493e-05,
+      "loss": 0.5474,
+      "step": 5924
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720160336822678e-05,
+      "loss": 0.5739,
+      "step": 5925
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7201026617066068e-05,
+      "loss": 0.5608,
+      "step": 5926
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.720044986590536e-05,
+      "loss": 0.5644,
+      "step": 5927
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7199873114744644e-05,
+      "loss": 0.5307,
+      "step": 5928
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719929636358393e-05,
+      "loss": 0.5682,
+      "step": 5929
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719871961242322e-05,
+      "loss": 0.6038,
+      "step": 5930
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719814286126251e-05,
+      "loss": 0.4937,
+      "step": 5931
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7197566110101798e-05,
+      "loss": 0.6235,
+      "step": 5932
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7196989358941085e-05,
+      "loss": 0.5564,
+      "step": 5933
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7196412607780376e-05,
+      "loss": 0.5783,
+      "step": 5934
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7195835856619664e-05,
+      "loss": 0.6156,
+      "step": 5935
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7195259105458952e-05,
+      "loss": 0.5023,
+      "step": 5936
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719468235429824e-05,
+      "loss": 0.5281,
+      "step": 5937
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7194105603137527e-05,
+      "loss": 0.5799,
+      "step": 5938
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7193528851976815e-05,
+      "loss": 0.665,
+      "step": 5939
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7192952100816106e-05,
+      "loss": 0.5599,
+      "step": 5940
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7192375349655394e-05,
+      "loss": 0.5505,
+      "step": 5941
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719179859849468e-05,
+      "loss": 0.5047,
+      "step": 5942
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.719122184733397e-05,
+      "loss": 0.5196,
+      "step": 5943
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7190645096173257e-05,
+      "loss": 0.4486,
+      "step": 5944
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7190068345012544e-05,
+      "loss": 0.5407,
+      "step": 5945
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7189491593851835e-05,
+      "loss": 0.4835,
+      "step": 5946
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7188914842691123e-05,
+      "loss": 0.4132,
+      "step": 5947
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.718833809153041e-05,
+      "loss": 0.5303,
+      "step": 5948
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.71877613403697e-05,
+      "loss": 0.4587,
+      "step": 5949
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7187184589208986e-05,
+      "loss": 0.5209,
+      "step": 5950
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7186607838048274e-05,
+      "loss": 0.6317,
+      "step": 5951
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7186031086887565e-05,
+      "loss": 0.4673,
+      "step": 5952
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7185454335726853e-05,
+      "loss": 0.567,
+      "step": 5953
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.718487758456614e-05,
+      "loss": 0.6203,
+      "step": 5954
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7184300833405428e-05,
+      "loss": 0.5654,
+      "step": 5955
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7183724082244716e-05,
+      "loss": 0.4813,
+      "step": 5956
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7183147331084003e-05,
+      "loss": 0.4556,
+      "step": 5957
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7182570579923294e-05,
+      "loss": 0.5944,
+      "step": 5958
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7181993828762582e-05,
+      "loss": 0.5173,
+      "step": 5959
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.718141707760187e-05,
+      "loss": 0.4962,
+      "step": 5960
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.718084032644116e-05,
+      "loss": 0.5824,
+      "step": 5961
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.718026357528045e-05,
+      "loss": 0.6108,
+      "step": 5962
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7179686824119733e-05,
+      "loss": 0.5617,
+      "step": 5963
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7179110072959024e-05,
+      "loss": 0.4976,
+      "step": 5964
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.717853332179831e-05,
+      "loss": 0.6453,
+      "step": 5965
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.71779565706376e-05,
+      "loss": 0.6029,
+      "step": 5966
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7177379819476887e-05,
+      "loss": 0.5537,
+      "step": 5967
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7176803068316178e-05,
+      "loss": 0.6018,
+      "step": 5968
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7176226317155466e-05,
+      "loss": 0.5624,
+      "step": 5969
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7175649565994753e-05,
+      "loss": 0.7347,
+      "step": 5970
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.717507281483404e-05,
+      "loss": 0.5575,
+      "step": 5971
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.717449606367333e-05,
+      "loss": 0.5145,
+      "step": 5972
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7173919312512616e-05,
+      "loss": 0.584,
+      "step": 5973
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7173342561351908e-05,
+      "loss": 0.5446,
+      "step": 5974
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7172765810191195e-05,
+      "loss": 0.4474,
+      "step": 5975
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7172189059030483e-05,
+      "loss": 0.6816,
+      "step": 5976
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.717161230786977e-05,
+      "loss": 0.5653,
+      "step": 5977
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7171035556709058e-05,
+      "loss": 0.6514,
+      "step": 5978
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7170458805548346e-05,
+      "loss": 0.4879,
+      "step": 5979
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7169882054387637e-05,
+      "loss": 0.5266,
+      "step": 5980
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7169305303226925e-05,
+      "loss": 0.4928,
+      "step": 5981
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7168728552066212e-05,
+      "loss": 0.5699,
+      "step": 5982
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.71681518009055e-05,
+      "loss": 0.6384,
+      "step": 5983
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7167575049744788e-05,
+      "loss": 0.6373,
+      "step": 5984
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7166998298584075e-05,
+      "loss": 0.7008,
+      "step": 5985
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7166421547423367e-05,
+      "loss": 0.5119,
+      "step": 5986
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7165844796262654e-05,
+      "loss": 0.4652,
+      "step": 5987
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.7165268045101942e-05,
+      "loss": 0.6138,
+      "step": 5988
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7164691293941233e-05,
+      "loss": 0.5766,
+      "step": 5989
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7164114542780517e-05,
+      "loss": 0.52,
+      "step": 5990
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7163537791619805e-05,
+      "loss": 0.4518,
+      "step": 5991
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7162961040459096e-05,
+      "loss": 0.6863,
+      "step": 5992
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7162384289298384e-05,
+      "loss": 0.6024,
+      "step": 5993
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.716180753813767e-05,
+      "loss": 0.5978,
+      "step": 5994
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7161230786976962e-05,
+      "loss": 0.6463,
+      "step": 5995
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.716065403581625e-05,
+      "loss": 0.698,
+      "step": 5996
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7160077284655538e-05,
+      "loss": 0.6525,
+      "step": 5997
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7159500533494826e-05,
+      "loss": 0.5897,
+      "step": 5998
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7158923782334113e-05,
+      "loss": 0.657,
+      "step": 5999
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.71583470311734e-05,
+      "loss": 0.511,
+      "step": 6000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715777028001269e-05,
+      "loss": 0.4889,
+      "step": 6001
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715719352885198e-05,
+      "loss": 0.5799,
+      "step": 6002
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7156616777691267e-05,
+      "loss": 0.5829,
+      "step": 6003
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7156040026530555e-05,
+      "loss": 0.4795,
+      "step": 6004
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7155463275369843e-05,
+      "loss": 0.5971,
+      "step": 6005
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715488652420913e-05,
+      "loss": 0.6671,
+      "step": 6006
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7154309773048418e-05,
+      "loss": 0.6206,
+      "step": 6007
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715373302188771e-05,
+      "loss": 0.6779,
+      "step": 6008
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7153156270726997e-05,
+      "loss": 0.5347,
+      "step": 6009
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7152579519566285e-05,
+      "loss": 0.6116,
+      "step": 6010
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7152002768405572e-05,
+      "loss": 0.5128,
+      "step": 6011
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715142601724486e-05,
+      "loss": 0.599,
+      "step": 6012
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7150849266084148e-05,
+      "loss": 0.5575,
+      "step": 6013
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.715027251492344e-05,
+      "loss": 0.5535,
+      "step": 6014
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7149695763762726e-05,
+      "loss": 0.6075,
+      "step": 6015
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7149119012602014e-05,
+      "loss": 0.5113,
+      "step": 6016
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7148542261441302e-05,
+      "loss": 0.607,
+      "step": 6017
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.714796551028059e-05,
+      "loss": 0.6416,
+      "step": 6018
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7147388759119877e-05,
+      "loss": 0.5677,
+      "step": 6019
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7146812007959168e-05,
+      "loss": 0.604,
+      "step": 6020
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7146235256798456e-05,
+      "loss": 0.5746,
+      "step": 6021
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7145658505637744e-05,
+      "loss": 0.5052,
+      "step": 6022
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7145081754477035e-05,
+      "loss": 0.5919,
+      "step": 6023
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7144505003316322e-05,
+      "loss": 0.609,
+      "step": 6024
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.714392825215561e-05,
+      "loss": 0.5199,
+      "step": 6025
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7143351500994898e-05,
+      "loss": 0.6638,
+      "step": 6026
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7142774749834185e-05,
+      "loss": 0.4875,
+      "step": 6027
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7142197998673473e-05,
+      "loss": 0.4486,
+      "step": 6028
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7141621247512764e-05,
+      "loss": 0.5321,
+      "step": 6029
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7141044496352052e-05,
+      "loss": 0.5041,
+      "step": 6030
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.714046774519134e-05,
+      "loss": 0.6186,
+      "step": 6031
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7139890994030627e-05,
+      "loss": 0.6771,
+      "step": 6032
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7139314242869915e-05,
+      "loss": 0.4868,
+      "step": 6033
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7138737491709203e-05,
+      "loss": 0.4627,
+      "step": 6034
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713816074054849e-05,
+      "loss": 0.5919,
+      "step": 6035
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713758398938778e-05,
+      "loss": 0.5569,
+      "step": 6036
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713700723822707e-05,
+      "loss": 0.4583,
+      "step": 6037
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7136430487066357e-05,
+      "loss": 0.4884,
+      "step": 6038
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7135853735905644e-05,
+      "loss": 0.555,
+      "step": 6039
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7135276984744932e-05,
+      "loss": 0.5785,
+      "step": 6040
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713470023358422e-05,
+      "loss": 0.6539,
+      "step": 6041
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713412348242351e-05,
+      "loss": 0.6587,
+      "step": 6042
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.71335467312628e-05,
+      "loss": 0.5477,
+      "step": 6043
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7132969980102086e-05,
+      "loss": 0.5497,
+      "step": 6044
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7132393228941374e-05,
+      "loss": 0.447,
+      "step": 6045
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713181647778066e-05,
+      "loss": 0.5558,
+      "step": 6046
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713123972661995e-05,
+      "loss": 0.4921,
+      "step": 6047
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.713066297545924e-05,
+      "loss": 0.5092,
+      "step": 6048
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7130086224298528e-05,
+      "loss": 0.5332,
+      "step": 6049
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7129509473137816e-05,
+      "loss": 0.4926,
+      "step": 6050
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7128932721977107e-05,
+      "loss": 0.6236,
+      "step": 6051
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712835597081639e-05,
+      "loss": 0.6276,
+      "step": 6052
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712777921965568e-05,
+      "loss": 0.4828,
+      "step": 6053
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712720246849497e-05,
+      "loss": 0.5936,
+      "step": 6054
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7126625717334257e-05,
+      "loss": 0.741,
+      "step": 6055
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7126048966173545e-05,
+      "loss": 0.6006,
+      "step": 6056
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7125472215012836e-05,
+      "loss": 0.6885,
+      "step": 6057
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7124895463852124e-05,
+      "loss": 0.4561,
+      "step": 6058
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712431871269141e-05,
+      "loss": 0.5082,
+      "step": 6059
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.71237419615307e-05,
+      "loss": 0.5434,
+      "step": 6060
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7123165210369987e-05,
+      "loss": 0.656,
+      "step": 6061
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7122588459209275e-05,
+      "loss": 0.5984,
+      "step": 6062
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7122011708048566e-05,
+      "loss": 0.4961,
+      "step": 6063
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7121434956887853e-05,
+      "loss": 0.5192,
+      "step": 6064
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712085820572714e-05,
+      "loss": 0.5907,
+      "step": 6065
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.712028145456643e-05,
+      "loss": 0.6777,
+      "step": 6066
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7119704703405716e-05,
+      "loss": 0.6858,
+      "step": 6067
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7119127952245004e-05,
+      "loss": 0.7368,
+      "step": 6068
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7118551201084295e-05,
+      "loss": 0.5344,
+      "step": 6069
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7117974449923583e-05,
+      "loss": 0.552,
+      "step": 6070
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.711739769876287e-05,
+      "loss": 0.545,
+      "step": 6071
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7116820947602158e-05,
+      "loss": 0.6705,
+      "step": 6072
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7116244196441446e-05,
+      "loss": 0.7669,
+      "step": 6073
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7115667445280734e-05,
+      "loss": 0.5332,
+      "step": 6074
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.711509069412002e-05,
+      "loss": 0.6143,
+      "step": 6075
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7114513942959312e-05,
+      "loss": 0.5744,
+      "step": 6076
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.71139371917986e-05,
+      "loss": 0.6032,
+      "step": 6077
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7113360440637888e-05,
+      "loss": 0.5982,
+      "step": 6078
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7112783689477175e-05,
+      "loss": 0.5672,
+      "step": 6079
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7112206938316463e-05,
+      "loss": 0.6489,
+      "step": 6080
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.711163018715575e-05,
+      "loss": 0.6842,
+      "step": 6081
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7111053435995042e-05,
+      "loss": 0.6112,
+      "step": 6082
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.711047668483433e-05,
+      "loss": 0.5529,
+      "step": 6083
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7109899933673617e-05,
+      "loss": 0.5381,
+      "step": 6084
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.710932318251291e-05,
+      "loss": 0.4566,
+      "step": 6085
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7108746431352196e-05,
+      "loss": 0.431,
+      "step": 6086
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7108169680191484e-05,
+      "loss": 0.6224,
+      "step": 6087
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.710759292903077e-05,
+      "loss": 0.5971,
+      "step": 6088
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.710701617787006e-05,
+      "loss": 0.4361,
+      "step": 6089
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7106439426709347e-05,
+      "loss": 0.4899,
+      "step": 6090
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7105862675548638e-05,
+      "loss": 0.6736,
+      "step": 6091
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7105285924387926e-05,
+      "loss": 0.6947,
+      "step": 6092
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7104709173227213e-05,
+      "loss": 0.631,
+      "step": 6093
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.71041324220665e-05,
+      "loss": 0.5887,
+      "step": 6094
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.710355567090579e-05,
+      "loss": 0.4752,
+      "step": 6095
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7102978919745076e-05,
+      "loss": 0.6395,
+      "step": 6096
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7102402168584367e-05,
+      "loss": 0.6623,
+      "step": 6097
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7101825417423655e-05,
+      "loss": 0.6002,
+      "step": 6098
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7101248666262943e-05,
+      "loss": 0.7224,
+      "step": 6099
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.710067191510223e-05,
+      "loss": 0.5944,
+      "step": 6100
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7100095163941518e-05,
+      "loss": 0.4916,
+      "step": 6101
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7099518412780806e-05,
+      "loss": 0.4987,
+      "step": 6102
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7098941661620097e-05,
+      "loss": 0.4824,
+      "step": 6103
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7098364910459385e-05,
+      "loss": 0.6837,
+      "step": 6104
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7097788159298672e-05,
+      "loss": 0.4644,
+      "step": 6105
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.709721140813796e-05,
+      "loss": 0.6175,
+      "step": 6106
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7096634656977248e-05,
+      "loss": 0.5861,
+      "step": 6107
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7096057905816535e-05,
+      "loss": 0.6498,
+      "step": 6108
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7095481154655823e-05,
+      "loss": 0.5835,
+      "step": 6109
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7094904403495114e-05,
+      "loss": 0.5148,
+      "step": 6110
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7094327652334402e-05,
+      "loss": 0.508,
+      "step": 6111
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.709375090117369e-05,
+      "loss": 0.5376,
+      "step": 6112
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.709317415001298e-05,
+      "loss": 0.5479,
+      "step": 6113
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7092597398852265e-05,
+      "loss": 0.5675,
+      "step": 6114
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7092020647691552e-05,
+      "loss": 0.5923,
+      "step": 6115
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7091443896530844e-05,
+      "loss": 0.6555,
+      "step": 6116
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.709086714537013e-05,
+      "loss": 0.5618,
+      "step": 6117
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.709029039420942e-05,
+      "loss": 0.5441,
+      "step": 6118
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708971364304871e-05,
+      "loss": 0.6853,
+      "step": 6119
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7089136891887998e-05,
+      "loss": 0.5541,
+      "step": 6120
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7088560140727285e-05,
+      "loss": 0.7109,
+      "step": 6121
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7087983389566573e-05,
+      "loss": 0.5841,
+      "step": 6122
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708740663840586e-05,
+      "loss": 0.5544,
+      "step": 6123
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708682988724515e-05,
+      "loss": 0.707,
+      "step": 6124
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708625313608444e-05,
+      "loss": 0.5984,
+      "step": 6125
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7085676384923727e-05,
+      "loss": 0.5153,
+      "step": 6126
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7085099633763015e-05,
+      "loss": 0.5401,
+      "step": 6127
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7084522882602303e-05,
+      "loss": 0.6837,
+      "step": 6128
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708394613144159e-05,
+      "loss": 0.5438,
+      "step": 6129
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7083369380280878e-05,
+      "loss": 0.5304,
+      "step": 6130
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708279262912017e-05,
+      "loss": 0.549,
+      "step": 6131
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7082215877959457e-05,
+      "loss": 0.4739,
+      "step": 6132
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7081639126798744e-05,
+      "loss": 0.4752,
+      "step": 6133
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7081062375638032e-05,
+      "loss": 0.5922,
+      "step": 6134
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.708048562447732e-05,
+      "loss": 0.5433,
+      "step": 6135
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7079908873316607e-05,
+      "loss": 0.6012,
+      "step": 6136
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.70793321221559e-05,
+      "loss": 0.6857,
+      "step": 6137
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7078755370995186e-05,
+      "loss": 0.4572,
+      "step": 6138
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7078178619834474e-05,
+      "loss": 0.6135,
+      "step": 6139
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707760186867376e-05,
+      "loss": 0.6937,
+      "step": 6140
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707702511751305e-05,
+      "loss": 0.6594,
+      "step": 6141
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7076448366352337e-05,
+      "loss": 0.4393,
+      "step": 6142
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7075871615191625e-05,
+      "loss": 0.5789,
+      "step": 6143
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7075294864030916e-05,
+      "loss": 0.5221,
+      "step": 6144
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7074718112870203e-05,
+      "loss": 0.4927,
+      "step": 6145
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707414136170949e-05,
+      "loss": 0.5286,
+      "step": 6146
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7073564610548782e-05,
+      "loss": 0.5692,
+      "step": 6147
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707298785938807e-05,
+      "loss": 0.6662,
+      "step": 6148
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7072411108227357e-05,
+      "loss": 0.5079,
+      "step": 6149
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7071834357066645e-05,
+      "loss": 0.4806,
+      "step": 6150
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7071257605905933e-05,
+      "loss": 0.5973,
+      "step": 6151
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707068085474522e-05,
+      "loss": 0.435,
+      "step": 6152
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.707010410358451e-05,
+      "loss": 0.5842,
+      "step": 6153
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.70695273524238e-05,
+      "loss": 0.4986,
+      "step": 6154
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7068950601263087e-05,
+      "loss": 0.6085,
+      "step": 6155
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7068373850102375e-05,
+      "loss": 0.5206,
+      "step": 6156
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7067797098941662e-05,
+      "loss": 0.5833,
+      "step": 6157
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.706722034778095e-05,
+      "loss": 0.4613,
+      "step": 6158
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.706664359662024e-05,
+      "loss": 0.5,
+      "step": 6159
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.706606684545953e-05,
+      "loss": 0.5225,
+      "step": 6160
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7065490094298816e-05,
+      "loss": 0.616,
+      "step": 6161
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7064913343138104e-05,
+      "loss": 0.5156,
+      "step": 6162
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7064336591977392e-05,
+      "loss": 0.5836,
+      "step": 6163
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.706375984081668e-05,
+      "loss": 0.5773,
+      "step": 6164
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.706318308965597e-05,
+      "loss": 0.4109,
+      "step": 6165
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.7062606338495258e-05,
+      "loss": 0.4155,
+      "step": 6166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7062029587334546e-05,
+      "loss": 0.5918,
+      "step": 6167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7061452836173834e-05,
+      "loss": 0.5371,
+      "step": 6168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.706087608501312e-05,
+      "loss": 0.4249,
+      "step": 6169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.706029933385241e-05,
+      "loss": 0.4495,
+      "step": 6170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.70597225826917e-05,
+      "loss": 0.4561,
+      "step": 6171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7059145831530988e-05,
+      "loss": 0.5584,
+      "step": 6172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7058569080370275e-05,
+      "loss": 0.4294,
+      "step": 6173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7057992329209563e-05,
+      "loss": 0.5436,
+      "step": 6174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7057415578048854e-05,
+      "loss": 0.4584,
+      "step": 6175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.705683882688814e-05,
+      "loss": 0.5736,
+      "step": 6176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7056262075727426e-05,
+      "loss": 0.4076,
+      "step": 6177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7055685324566717e-05,
+      "loss": 0.5365,
+      "step": 6178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7055108573406005e-05,
+      "loss": 0.5899,
+      "step": 6179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7054531822245293e-05,
+      "loss": 0.5646,
+      "step": 6180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7053955071084584e-05,
+      "loss": 0.5366,
+      "step": 6181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.705337831992387e-05,
+      "loss": 0.4752,
+      "step": 6182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.705280156876316e-05,
+      "loss": 0.5746,
+      "step": 6183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7052224817602447e-05,
+      "loss": 0.5189,
+      "step": 6184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7051648066441734e-05,
+      "loss": 0.494,
+      "step": 6185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7051071315281022e-05,
+      "loss": 0.5709,
+      "step": 6186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7050494564120313e-05,
+      "loss": 0.4413,
+      "step": 6187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.70499178129596e-05,
+      "loss": 0.4963,
+      "step": 6188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704934106179889e-05,
+      "loss": 0.4853,
+      "step": 6189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7048764310638176e-05,
+      "loss": 0.5928,
+      "step": 6190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7048187559477464e-05,
+      "loss": 0.6023,
+      "step": 6191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704761080831675e-05,
+      "loss": 0.5689,
+      "step": 6192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7047034057156043e-05,
+      "loss": 0.6614,
+      "step": 6193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704645730599533e-05,
+      "loss": 0.6617,
+      "step": 6194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7045880554834618e-05,
+      "loss": 0.6431,
+      "step": 6195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7045303803673906e-05,
+      "loss": 0.4859,
+      "step": 6196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7044727052513193e-05,
+      "loss": 0.5467,
+      "step": 6197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704415030135248e-05,
+      "loss": 0.4987,
+      "step": 6198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7043573550191772e-05,
+      "loss": 0.5025,
+      "step": 6199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704299679903106e-05,
+      "loss": 0.5358,
+      "step": 6200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7042420047870348e-05,
+      "loss": 0.5158,
+      "step": 6201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704184329670964e-05,
+      "loss": 0.6399,
+      "step": 6202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7041266545548923e-05,
+      "loss": 0.5652,
+      "step": 6203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.704068979438821e-05,
+      "loss": 0.4734,
+      "step": 6204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7040113043227502e-05,
+      "loss": 0.5936,
+      "step": 6205
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.703953629206679e-05,
+      "loss": 0.4287,
+      "step": 6206
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7038959540906077e-05,
+      "loss": 0.538,
+      "step": 6207
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7038382789745368e-05,
+      "loss": 0.5932,
+      "step": 6208
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7037806038584656e-05,
+      "loss": 0.5248,
+      "step": 6209
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7037229287423944e-05,
+      "loss": 0.5626,
+      "step": 6210
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.703665253626323e-05,
+      "loss": 0.5638,
+      "step": 6211
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.703607578510252e-05,
+      "loss": 0.6665,
+      "step": 6212
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7035499033941807e-05,
+      "loss": 0.6217,
+      "step": 6213
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7034922282781094e-05,
+      "loss": 0.4587,
+      "step": 6214
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7034345531620385e-05,
+      "loss": 0.4475,
+      "step": 6215
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7033768780459673e-05,
+      "loss": 0.6265,
+      "step": 6216
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.703319202929896e-05,
+      "loss": 0.6299,
+      "step": 6217
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.703261527813825e-05,
+      "loss": 0.7338,
+      "step": 6218
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7032038526977536e-05,
+      "loss": 0.6179,
+      "step": 6219
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7031461775816824e-05,
+      "loss": 0.4238,
+      "step": 6220
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7030885024656115e-05,
+      "loss": 0.5782,
+      "step": 6221
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7030308273495403e-05,
+      "loss": 0.5627,
+      "step": 6222
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.702973152233469e-05,
+      "loss": 0.5422,
+      "step": 6223
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7029154771173978e-05,
+      "loss": 0.591,
+      "step": 6224
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7028578020013266e-05,
+      "loss": 0.6045,
+      "step": 6225
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7028001268852553e-05,
+      "loss": 0.4805,
+      "step": 6226
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7027424517691844e-05,
+      "loss": 0.5474,
+      "step": 6227
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7026847766531132e-05,
+      "loss": 0.4311,
+      "step": 6228
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.702627101537042e-05,
+      "loss": 0.4238,
+      "step": 6229
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7025694264209707e-05,
+      "loss": 0.5979,
+      "step": 6230
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7025117513048995e-05,
+      "loss": 0.6688,
+      "step": 6231
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7024540761888283e-05,
+      "loss": 0.6062,
+      "step": 6232
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7023964010727574e-05,
+      "loss": 0.5374,
+      "step": 6233
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.702338725956686e-05,
+      "loss": 0.4484,
+      "step": 6234
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.702281050840615e-05,
+      "loss": 0.5758,
+      "step": 6235
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.702223375724544e-05,
+      "loss": 0.6482,
+      "step": 6236
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7021657006084728e-05,
+      "loss": 0.5727,
+      "step": 6237
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7021080254924012e-05,
+      "loss": 0.6982,
+      "step": 6238
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7020503503763303e-05,
+      "loss": 0.5701,
+      "step": 6239
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.701992675260259e-05,
+      "loss": 0.6701,
+      "step": 6240
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.701935000144188e-05,
+      "loss": 0.4757,
+      "step": 6241
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.701877325028117e-05,
+      "loss": 0.5575,
+      "step": 6242
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7018196499120457e-05,
+      "loss": 0.5106,
+      "step": 6243
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7017619747959745e-05,
+      "loss": 0.4285,
+      "step": 6244
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7017042996799033e-05,
+      "loss": 0.596,
+      "step": 6245
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.701646624563832e-05,
+      "loss": 0.5394,
+      "step": 6246
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7015889494477608e-05,
+      "loss": 0.5199,
+      "step": 6247
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7015312743316896e-05,
+      "loss": 0.6556,
+      "step": 6248
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7014735992156187e-05,
+      "loss": 0.5635,
+      "step": 6249
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7014159240995475e-05,
+      "loss": 0.6253,
+      "step": 6250
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7013582489834762e-05,
+      "loss": 0.5396,
+      "step": 6251
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.701300573867405e-05,
+      "loss": 0.6835,
+      "step": 6252
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7012428987513338e-05,
+      "loss": 0.5395,
+      "step": 6253
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7011852236352625e-05,
+      "loss": 0.7035,
+      "step": 6254
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7011275485191916e-05,
+      "loss": 0.6429,
+      "step": 6255
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7010698734031204e-05,
+      "loss": 0.4232,
+      "step": 6256
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7010121982870492e-05,
+      "loss": 0.3604,
+      "step": 6257
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.700954523170978e-05,
+      "loss": 0.6233,
+      "step": 6258
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7008968480549067e-05,
+      "loss": 0.5533,
+      "step": 6259
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7008391729388355e-05,
+      "loss": 0.5831,
+      "step": 6260
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7007814978227646e-05,
+      "loss": 0.6571,
+      "step": 6261
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7007238227066934e-05,
+      "loss": 0.5733,
+      "step": 6262
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.700666147590622e-05,
+      "loss": 0.5114,
+      "step": 6263
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7006084724745512e-05,
+      "loss": 0.6387,
+      "step": 6264
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7005507973584797e-05,
+      "loss": 0.5343,
+      "step": 6265
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7004931222424084e-05,
+      "loss": 0.4856,
+      "step": 6266
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7004354471263375e-05,
+      "loss": 0.6519,
+      "step": 6267
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7003777720102663e-05,
+      "loss": 0.6878,
+      "step": 6268
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.700320096894195e-05,
+      "loss": 0.6118,
+      "step": 6269
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7002624217781242e-05,
+      "loss": 0.534,
+      "step": 6270
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.700204746662053e-05,
+      "loss": 0.6105,
+      "step": 6271
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7001470715459817e-05,
+      "loss": 0.4862,
+      "step": 6272
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7000893964299105e-05,
+      "loss": 0.589,
+      "step": 6273
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.7000317213138393e-05,
+      "loss": 0.5771,
+      "step": 6274
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699974046197768e-05,
+      "loss": 0.5219,
+      "step": 6275
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699916371081697e-05,
+      "loss": 0.645,
+      "step": 6276
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699858695965626e-05,
+      "loss": 0.484,
+      "step": 6277
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6998010208495547e-05,
+      "loss": 0.5461,
+      "step": 6278
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6997433457334834e-05,
+      "loss": 0.567,
+      "step": 6279
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6996856706174122e-05,
+      "loss": 0.6726,
+      "step": 6280
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699627995501341e-05,
+      "loss": 0.5046,
+      "step": 6281
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6995703203852698e-05,
+      "loss": 0.5537,
+      "step": 6282
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699512645269199e-05,
+      "loss": 0.4482,
+      "step": 6283
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6994549701531276e-05,
+      "loss": 0.507,
+      "step": 6284
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6993972950370564e-05,
+      "loss": 0.569,
+      "step": 6285
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699339619920985e-05,
+      "loss": 0.5773,
+      "step": 6286
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.699281944804914e-05,
+      "loss": 0.6138,
+      "step": 6287
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6992242696888427e-05,
+      "loss": 0.5148,
+      "step": 6288
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6991665945727718e-05,
+      "loss": 0.5584,
+      "step": 6289
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6991089194567006e-05,
+      "loss": 0.5857,
+      "step": 6290
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6990512443406293e-05,
+      "loss": 0.4704,
+      "step": 6291
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.698993569224558e-05,
+      "loss": 0.5933,
+      "step": 6292
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.698935894108487e-05,
+      "loss": 0.6669,
+      "step": 6293
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6988782189924157e-05,
+      "loss": 0.4049,
+      "step": 6294
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6988205438763448e-05,
+      "loss": 0.6619,
+      "step": 6295
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6987628687602735e-05,
+      "loss": 0.4754,
+      "step": 6296
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6987051936442023e-05,
+      "loss": 0.421,
+      "step": 6297
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6986475185281314e-05,
+      "loss": 0.5091,
+      "step": 6298
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6985898434120602e-05,
+      "loss": 0.5812,
+      "step": 6299
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6985321682959886e-05,
+      "loss": 0.5314,
+      "step": 6300
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6984744931799177e-05,
+      "loss": 0.5105,
+      "step": 6301
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6984168180638465e-05,
+      "loss": 0.6582,
+      "step": 6302
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6983591429477752e-05,
+      "loss": 0.579,
+      "step": 6303
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6983014678317044e-05,
+      "loss": 0.5519,
+      "step": 6304
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.698243792715633e-05,
+      "loss": 0.6642,
+      "step": 6305
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.698186117599562e-05,
+      "loss": 0.572,
+      "step": 6306
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6981284424834907e-05,
+      "loss": 0.5714,
+      "step": 6307
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6980707673674194e-05,
+      "loss": 0.416,
+      "step": 6308
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6980130922513482e-05,
+      "loss": 0.64,
+      "step": 6309
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6979554171352773e-05,
+      "loss": 0.6898,
+      "step": 6310
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697897742019206e-05,
+      "loss": 0.5798,
+      "step": 6311
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697840066903135e-05,
+      "loss": 0.5287,
+      "step": 6312
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6977823917870636e-05,
+      "loss": 0.6133,
+      "step": 6313
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6977247166709924e-05,
+      "loss": 0.5951,
+      "step": 6314
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697667041554921e-05,
+      "loss": 0.5972,
+      "step": 6315
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.69760936643885e-05,
+      "loss": 0.5961,
+      "step": 6316
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697551691322779e-05,
+      "loss": 0.554,
+      "step": 6317
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6974940162067078e-05,
+      "loss": 0.5537,
+      "step": 6318
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6974363410906366e-05,
+      "loss": 0.5878,
+      "step": 6319
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6973786659745653e-05,
+      "loss": 0.5381,
+      "step": 6320
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697320990858494e-05,
+      "loss": 0.5224,
+      "step": 6321
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697263315742423e-05,
+      "loss": 0.5673,
+      "step": 6322
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.697205640626352e-05,
+      "loss": 0.5571,
+      "step": 6323
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6971479655102807e-05,
+      "loss": 0.5537,
+      "step": 6324
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6970902903942095e-05,
+      "loss": 0.5391,
+      "step": 6325
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6970326152781386e-05,
+      "loss": 0.5927,
+      "step": 6326
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.696974940162067e-05,
+      "loss": 0.5828,
+      "step": 6327
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6969172650459958e-05,
+      "loss": 0.6172,
+      "step": 6328
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.696859589929925e-05,
+      "loss": 0.4829,
+      "step": 6329
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6968019148138537e-05,
+      "loss": 0.666,
+      "step": 6330
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6967442396977825e-05,
+      "loss": 0.6627,
+      "step": 6331
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6966865645817116e-05,
+      "loss": 0.5679,
+      "step": 6332
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6966288894656403e-05,
+      "loss": 0.556,
+      "step": 6333
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.696571214349569e-05,
+      "loss": 0.4626,
+      "step": 6334
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.696513539233498e-05,
+      "loss": 0.519,
+      "step": 6335
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6964558641174266e-05,
+      "loss": 0.5869,
+      "step": 6336
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6963981890013554e-05,
+      "loss": 0.5188,
+      "step": 6337
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6963405138852845e-05,
+      "loss": 0.601,
+      "step": 6338
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6962828387692133e-05,
+      "loss": 0.4742,
+      "step": 6339
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.696225163653142e-05,
+      "loss": 0.5378,
+      "step": 6340
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6961674885370708e-05,
+      "loss": 0.5731,
+      "step": 6341
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6961098134209996e-05,
+      "loss": 0.5966,
+      "step": 6342
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6960521383049284e-05,
+      "loss": 0.6349,
+      "step": 6343
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6959944631888575e-05,
+      "loss": 0.7164,
+      "step": 6344
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.6959367880727862e-05,
+      "loss": 0.5121,
+      "step": 6345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.695879112956715e-05,
+      "loss": 0.5556,
+      "step": 6346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6958214378406438e-05,
+      "loss": 0.6447,
+      "step": 6347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6957637627245725e-05,
+      "loss": 0.5534,
+      "step": 6348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6957060876085013e-05,
+      "loss": 0.5295,
+      "step": 6349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.69564841249243e-05,
+      "loss": 0.4846,
+      "step": 6350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6955907373763592e-05,
+      "loss": 0.7604,
+      "step": 6351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.695533062260288e-05,
+      "loss": 0.5968,
+      "step": 6352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6954753871442167e-05,
+      "loss": 0.7031,
+      "step": 6353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6954177120281455e-05,
+      "loss": 0.5662,
+      "step": 6354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6953600369120743e-05,
+      "loss": 0.4974,
+      "step": 6355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.695302361796003e-05,
+      "loss": 0.5679,
+      "step": 6356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.695244686679932e-05,
+      "loss": 0.5907,
+      "step": 6357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.695187011563861e-05,
+      "loss": 0.5953,
+      "step": 6358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6951293364477897e-05,
+      "loss": 0.4818,
+      "step": 6359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6950716613317188e-05,
+      "loss": 0.5437,
+      "step": 6360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6950139862156475e-05,
+      "loss": 0.5156,
+      "step": 6361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.694956311099576e-05,
+      "loss": 0.5739,
+      "step": 6362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.694898635983505e-05,
+      "loss": 0.5818,
+      "step": 6363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.694840960867434e-05,
+      "loss": 0.5838,
+      "step": 6364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6947832857513626e-05,
+      "loss": 0.4229,
+      "step": 6365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6947256106352917e-05,
+      "loss": 0.5631,
+      "step": 6366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6946679355192205e-05,
+      "loss": 0.5438,
+      "step": 6367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6946102604031493e-05,
+      "loss": 0.5392,
+      "step": 6368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.694552585287078e-05,
+      "loss": 0.5853,
+      "step": 6369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6944949101710068e-05,
+      "loss": 0.5857,
+      "step": 6370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6944372350549356e-05,
+      "loss": 0.6386,
+      "step": 6371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6943795599388647e-05,
+      "loss": 0.5708,
+      "step": 6372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6943218848227934e-05,
+      "loss": 0.6307,
+      "step": 6373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6942642097067222e-05,
+      "loss": 0.6027,
+      "step": 6374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.694206534590651e-05,
+      "loss": 0.6337,
+      "step": 6375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6941488594745798e-05,
+      "loss": 0.4857,
+      "step": 6376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6940911843585085e-05,
+      "loss": 0.4498,
+      "step": 6377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6940335092424376e-05,
+      "loss": 0.461,
+      "step": 6378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6939758341263664e-05,
+      "loss": 0.6127,
+      "step": 6379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693918159010295e-05,
+      "loss": 0.5203,
+      "step": 6380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693860483894224e-05,
+      "loss": 0.7077,
+      "step": 6381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6938028087781527e-05,
+      "loss": 0.561,
+      "step": 6382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6937451336620815e-05,
+      "loss": 0.5901,
+      "step": 6383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6936874585460102e-05,
+      "loss": 0.6427,
+      "step": 6384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6936297834299393e-05,
+      "loss": 0.5882,
+      "step": 6385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693572108313868e-05,
+      "loss": 0.5316,
+      "step": 6386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693514433197797e-05,
+      "loss": 0.5813,
+      "step": 6387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693456758081726e-05,
+      "loss": 0.6057,
+      "step": 6388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6933990829656544e-05,
+      "loss": 0.5919,
+      "step": 6389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6933414078495832e-05,
+      "loss": 0.4994,
+      "step": 6390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6932837327335123e-05,
+      "loss": 0.6032,
+      "step": 6391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693226057617441e-05,
+      "loss": 0.5521,
+      "step": 6392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.69316838250137e-05,
+      "loss": 0.5547,
+      "step": 6393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.693110707385299e-05,
+      "loss": 0.4783,
+      "step": 6394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6930530322692277e-05,
+      "loss": 0.5674,
+      "step": 6395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6929953571531565e-05,
+      "loss": 0.5894,
+      "step": 6396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6929376820370852e-05,
+      "loss": 0.5032,
+      "step": 6397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.692880006921014e-05,
+      "loss": 0.5247,
+      "step": 6398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6928223318049428e-05,
+      "loss": 0.5159,
+      "step": 6399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.692764656688872e-05,
+      "loss": 0.5071,
+      "step": 6400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6927069815728007e-05,
+      "loss": 0.4366,
+      "step": 6401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6926493064567294e-05,
+      "loss": 0.5487,
+      "step": 6402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6925916313406582e-05,
+      "loss": 0.4793,
+      "step": 6403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.692533956224587e-05,
+      "loss": 0.4876,
+      "step": 6404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6924762811085157e-05,
+      "loss": 0.6405,
+      "step": 6405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.692418605992445e-05,
+      "loss": 0.576,
+      "step": 6406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6923609308763736e-05,
+      "loss": 0.5951,
+      "step": 6407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6923032557603024e-05,
+      "loss": 0.5291,
+      "step": 6408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.692245580644231e-05,
+      "loss": 0.4865,
+      "step": 6409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.69218790552816e-05,
+      "loss": 0.5854,
+      "step": 6410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6921302304120887e-05,
+      "loss": 0.4863,
+      "step": 6411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6920725552960178e-05,
+      "loss": 0.5504,
+      "step": 6412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6920148801799466e-05,
+      "loss": 0.558,
+      "step": 6413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6919572050638753e-05,
+      "loss": 0.5167,
+      "step": 6414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691899529947804e-05,
+      "loss": 0.5376,
+      "step": 6415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691841854831733e-05,
+      "loss": 0.7101,
+      "step": 6416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6917841797156616e-05,
+      "loss": 0.4923,
+      "step": 6417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6917265045995904e-05,
+      "loss": 0.5782,
+      "step": 6418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6916688294835195e-05,
+      "loss": 0.5167,
+      "step": 6419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6916111543674483e-05,
+      "loss": 0.5397,
+      "step": 6420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691553479251377e-05,
+      "loss": 0.6841,
+      "step": 6421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691495804135306e-05,
+      "loss": 0.5552,
+      "step": 6422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691438129019235e-05,
+      "loss": 0.5666,
+      "step": 6423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6913804539031637e-05,
+      "loss": 0.5606,
+      "step": 6424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6913227787870925e-05,
+      "loss": 0.515,
+      "step": 6425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6912651036710212e-05,
+      "loss": 0.4731,
+      "step": 6426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.69120742855495e-05,
+      "loss": 0.4989,
+      "step": 6427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691149753438879e-05,
+      "loss": 0.5193,
+      "step": 6428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.691092078322808e-05,
+      "loss": 0.6014,
+      "step": 6429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6910344032067366e-05,
+      "loss": 0.7239,
+      "step": 6430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6909767280906654e-05,
+      "loss": 0.5781,
+      "step": 6431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6909190529745942e-05,
+      "loss": 0.635,
+      "step": 6432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690861377858523e-05,
+      "loss": 0.5034,
+      "step": 6433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690803702742452e-05,
+      "loss": 0.6921,
+      "step": 6434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6907460276263808e-05,
+      "loss": 0.6129,
+      "step": 6435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6906883525103096e-05,
+      "loss": 0.5393,
+      "step": 6436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6906306773942384e-05,
+      "loss": 0.5066,
+      "step": 6437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690573002278167e-05,
+      "loss": 0.546,
+      "step": 6438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690515327162096e-05,
+      "loss": 0.4434,
+      "step": 6439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690457652046025e-05,
+      "loss": 0.5216,
+      "step": 6440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6903999769299538e-05,
+      "loss": 0.5111,
+      "step": 6441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6903423018138825e-05,
+      "loss": 0.5267,
+      "step": 6442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6902846266978113e-05,
+      "loss": 0.6652,
+      "step": 6443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.69022695158174e-05,
+      "loss": 0.51,
+      "step": 6444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690169276465669e-05,
+      "loss": 0.6109,
+      "step": 6445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.690111601349598e-05,
+      "loss": 0.505,
+      "step": 6446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6900539262335267e-05,
+      "loss": 0.5681,
+      "step": 6447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6899962511174555e-05,
+      "loss": 0.5663,
+      "step": 6448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6899385760013846e-05,
+      "loss": 0.6063,
+      "step": 6449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6898809008853134e-05,
+      "loss": 0.6431,
+      "step": 6450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6898232257692418e-05,
+      "loss": 0.6504,
+      "step": 6451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.689765550653171e-05,
+      "loss": 0.5765,
+      "step": 6452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6897078755370997e-05,
+      "loss": 0.5663,
+      "step": 6453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6896502004210284e-05,
+      "loss": 0.5965,
+      "step": 6454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6895925253049572e-05,
+      "loss": 0.4761,
+      "step": 6455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6895348501888863e-05,
+      "loss": 0.5978,
+      "step": 6456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.689477175072815e-05,
+      "loss": 0.5503,
+      "step": 6457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.689419499956744e-05,
+      "loss": 0.5191,
+      "step": 6458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6893618248406726e-05,
+      "loss": 0.553,
+      "step": 6459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6893041497246014e-05,
+      "loss": 0.5567,
+      "step": 6460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.68924647460853e-05,
+      "loss": 0.5565,
+      "step": 6461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6891887994924593e-05,
+      "loss": 0.6072,
+      "step": 6462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.689131124376388e-05,
+      "loss": 0.4406,
+      "step": 6463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6890734492603168e-05,
+      "loss": 0.5847,
+      "step": 6464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6890157741442456e-05,
+      "loss": 0.4826,
+      "step": 6465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6889580990281743e-05,
+      "loss": 0.5612,
+      "step": 6466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688900423912103e-05,
+      "loss": 0.5226,
+      "step": 6467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6888427487960322e-05,
+      "loss": 0.5037,
+      "step": 6468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688785073679961e-05,
+      "loss": 0.4593,
+      "step": 6469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6887273985638898e-05,
+      "loss": 0.5476,
+      "step": 6470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6886697234478185e-05,
+      "loss": 0.5437,
+      "step": 6471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6886120483317473e-05,
+      "loss": 0.4557,
+      "step": 6472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688554373215676e-05,
+      "loss": 0.5266,
+      "step": 6473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688496698099605e-05,
+      "loss": 0.6326,
+      "step": 6474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688439022983534e-05,
+      "loss": 0.6081,
+      "step": 6475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6883813478674627e-05,
+      "loss": 0.6207,
+      "step": 6476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6883236727513918e-05,
+      "loss": 0.6259,
+      "step": 6477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6882659976353202e-05,
+      "loss": 0.5938,
+      "step": 6478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688208322519249e-05,
+      "loss": 0.5178,
+      "step": 6479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688150647403178e-05,
+      "loss": 0.4053,
+      "step": 6480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.688092972287107e-05,
+      "loss": 0.449,
+      "step": 6481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6880352971710357e-05,
+      "loss": 0.5617,
+      "step": 6482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6879776220549648e-05,
+      "loss": 0.5327,
+      "step": 6483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6879199469388935e-05,
+      "loss": 0.4334,
+      "step": 6484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6878622718228223e-05,
+      "loss": 0.5623,
+      "step": 6485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.687804596706751e-05,
+      "loss": 0.6495,
+      "step": 6486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.68774692159068e-05,
+      "loss": 0.6326,
+      "step": 6487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6876892464746086e-05,
+      "loss": 0.5866,
+      "step": 6488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6876315713585374e-05,
+      "loss": 0.515,
+      "step": 6489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6875738962424665e-05,
+      "loss": 0.6456,
+      "step": 6490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6875162211263952e-05,
+      "loss": 0.4908,
+      "step": 6491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.687458546010324e-05,
+      "loss": 0.6012,
+      "step": 6492
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6874008708942528e-05,
+      "loss": 0.5012,
+      "step": 6493
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6873431957781816e-05,
+      "loss": 0.5118,
+      "step": 6494
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6872855206621103e-05,
+      "loss": 0.5925,
+      "step": 6495
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6872278455460394e-05,
+      "loss": 0.4714,
+      "step": 6496
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6871701704299682e-05,
+      "loss": 0.5543,
+      "step": 6497
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.687112495313897e-05,
+      "loss": 0.5523,
+      "step": 6498
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6870548201978257e-05,
+      "loss": 0.5408,
+      "step": 6499
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6869971450817545e-05,
+      "loss": 0.5091,
+      "step": 6500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6869394699656833e-05,
+      "loss": 0.5249,
+      "step": 6501
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6868817948496124e-05,
+      "loss": 0.5358,
+      "step": 6502
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686824119733541e-05,
+      "loss": 0.6311,
+      "step": 6503
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.68676644461747e-05,
+      "loss": 0.6007,
+      "step": 6504
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6867087695013987e-05,
+      "loss": 0.6539,
+      "step": 6505
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6866510943853275e-05,
+      "loss": 0.4992,
+      "step": 6506
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6865934192692562e-05,
+      "loss": 0.4747,
+      "step": 6507
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6865357441531853e-05,
+      "loss": 0.6211,
+      "step": 6508
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686478069037114e-05,
+      "loss": 0.565,
+      "step": 6509
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686420393921043e-05,
+      "loss": 0.6064,
+      "step": 6510
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686362718804972e-05,
+      "loss": 0.5737,
+      "step": 6511
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6863050436889007e-05,
+      "loss": 0.4819,
+      "step": 6512
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6862473685728292e-05,
+      "loss": 0.6379,
+      "step": 6513
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6861896934567583e-05,
+      "loss": 0.6524,
+      "step": 6514
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686132018340687e-05,
+      "loss": 0.6154,
+      "step": 6515
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6860743432246158e-05,
+      "loss": 0.5209,
+      "step": 6516
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.686016668108545e-05,
+      "loss": 0.5992,
+      "step": 6517
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6859589929924737e-05,
+      "loss": 0.4396,
+      "step": 6518
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6859013178764025e-05,
+      "loss": 0.6157,
+      "step": 6519
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6858436427603312e-05,
+      "loss": 0.6427,
+      "step": 6520
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.68578596764426e-05,
+      "loss": 0.4877,
+      "step": 6521
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6857282925281888e-05,
+      "loss": 0.6338,
+      "step": 6522
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6856706174121175e-05,
+      "loss": 0.5533,
+      "step": 6523
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.6856129422960466e-05,
+      "loss": 0.4759,
+      "step": 6524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6855552671799754e-05,
+      "loss": 0.483,
+      "step": 6525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6854975920639042e-05,
+      "loss": 0.6255,
+      "step": 6526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.685439916947833e-05,
+      "loss": 0.3913,
+      "step": 6527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6853822418317617e-05,
+      "loss": 0.5977,
+      "step": 6528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6853245667156905e-05,
+      "loss": 0.5706,
+      "step": 6529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6852668915996196e-05,
+      "loss": 0.6456,
+      "step": 6530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6852092164835484e-05,
+      "loss": 0.4852,
+      "step": 6531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.685151541367477e-05,
+      "loss": 0.4495,
+      "step": 6532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.685093866251406e-05,
+      "loss": 0.5577,
+      "step": 6533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6850361911353347e-05,
+      "loss": 0.5492,
+      "step": 6534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6849785160192634e-05,
+      "loss": 0.5029,
+      "step": 6535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6849208409031925e-05,
+      "loss": 0.6615,
+      "step": 6536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6848631657871213e-05,
+      "loss": 0.6169,
+      "step": 6537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.68480549067105e-05,
+      "loss": 0.4086,
+      "step": 6538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6847478155549792e-05,
+      "loss": 0.5645,
+      "step": 6539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6846901404389076e-05,
+      "loss": 0.6037,
+      "step": 6540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6846324653228364e-05,
+      "loss": 0.4981,
+      "step": 6541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6845747902067655e-05,
+      "loss": 0.6031,
+      "step": 6542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6845171150906943e-05,
+      "loss": 0.5334,
+      "step": 6543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.684459439974623e-05,
+      "loss": 0.478,
+      "step": 6544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.684401764858552e-05,
+      "loss": 0.6267,
+      "step": 6545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.684344089742481e-05,
+      "loss": 0.5682,
+      "step": 6546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6842864146264097e-05,
+      "loss": 0.6388,
+      "step": 6547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6842287395103384e-05,
+      "loss": 0.5601,
+      "step": 6548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6841710643942672e-05,
+      "loss": 0.5168,
+      "step": 6549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.684113389278196e-05,
+      "loss": 0.5244,
+      "step": 6550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.684055714162125e-05,
+      "loss": 0.6508,
+      "step": 6551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.683998039046054e-05,
+      "loss": 0.5149,
+      "step": 6552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6839403639299826e-05,
+      "loss": 0.5852,
+      "step": 6553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6838826888139114e-05,
+      "loss": 0.4804,
+      "step": 6554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.68382501369784e-05,
+      "loss": 0.4688,
+      "step": 6555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.683767338581769e-05,
+      "loss": 0.5296,
+      "step": 6556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6837096634656977e-05,
+      "loss": 0.5584,
+      "step": 6557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6836519883496268e-05,
+      "loss": 0.4393,
+      "step": 6558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6835943132335556e-05,
+      "loss": 0.4917,
+      "step": 6559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6835366381174843e-05,
+      "loss": 0.5732,
+      "step": 6560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.683478963001413e-05,
+      "loss": 0.6438,
+      "step": 6561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.683421287885342e-05,
+      "loss": 0.5846,
+      "step": 6562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6833636127692706e-05,
+      "loss": 0.5823,
+      "step": 6563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6833059376531998e-05,
+      "loss": 0.4884,
+      "step": 6564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6832482625371285e-05,
+      "loss": 0.6157,
+      "step": 6565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6831905874210573e-05,
+      "loss": 0.5246,
+      "step": 6566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.683132912304986e-05,
+      "loss": 0.6068,
+      "step": 6567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6830752371889148e-05,
+      "loss": 0.5062,
+      "step": 6568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6830175620728436e-05,
+      "loss": 0.487,
+      "step": 6569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6829598869567727e-05,
+      "loss": 0.4898,
+      "step": 6570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6829022118407015e-05,
+      "loss": 0.479,
+      "step": 6571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6828445367246302e-05,
+      "loss": 0.4934,
+      "step": 6572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6827868616085593e-05,
+      "loss": 0.6042,
+      "step": 6573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.682729186492488e-05,
+      "loss": 0.5729,
+      "step": 6574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6826715113764165e-05,
+      "loss": 0.6333,
+      "step": 6575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6826138362603457e-05,
+      "loss": 0.5475,
+      "step": 6576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6825561611442744e-05,
+      "loss": 0.492,
+      "step": 6577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6824984860282032e-05,
+      "loss": 0.629,
+      "step": 6578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6824408109121323e-05,
+      "loss": 0.6449,
+      "step": 6579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.682383135796061e-05,
+      "loss": 0.5438,
+      "step": 6580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.68232546067999e-05,
+      "loss": 0.4762,
+      "step": 6581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6822677855639186e-05,
+      "loss": 0.5675,
+      "step": 6582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6822101104478474e-05,
+      "loss": 0.492,
+      "step": 6583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.682152435331776e-05,
+      "loss": 0.656,
+      "step": 6584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6820947602157052e-05,
+      "loss": 0.6467,
+      "step": 6585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.682037085099634e-05,
+      "loss": 0.6905,
+      "step": 6586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6819794099835628e-05,
+      "loss": 0.6049,
+      "step": 6587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6819217348674916e-05,
+      "loss": 0.5214,
+      "step": 6588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6818640597514203e-05,
+      "loss": 0.5309,
+      "step": 6589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.681806384635349e-05,
+      "loss": 0.5055,
+      "step": 6590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6817487095192782e-05,
+      "loss": 0.5107,
+      "step": 6591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.681691034403207e-05,
+      "loss": 0.5168,
+      "step": 6592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6816333592871357e-05,
+      "loss": 0.5428,
+      "step": 6593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6815756841710645e-05,
+      "loss": 0.474,
+      "step": 6594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6815180090549933e-05,
+      "loss": 0.5781,
+      "step": 6595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.681460333938922e-05,
+      "loss": 0.5601,
+      "step": 6596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6814026588228508e-05,
+      "loss": 0.5199,
+      "step": 6597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.68134498370678e-05,
+      "loss": 0.6772,
+      "step": 6598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6812873085907087e-05,
+      "loss": 0.4381,
+      "step": 6599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6812296334746375e-05,
+      "loss": 0.5367,
+      "step": 6600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6811719583585666e-05,
+      "loss": 0.4825,
+      "step": 6601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.681114283242495e-05,
+      "loss": 0.5181,
+      "step": 6602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6810566081264238e-05,
+      "loss": 0.7399,
+      "step": 6603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.680998933010353e-05,
+      "loss": 0.5615,
+      "step": 6604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6809412578942816e-05,
+      "loss": 0.5862,
+      "step": 6605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6808835827782104e-05,
+      "loss": 0.5189,
+      "step": 6606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6808259076621395e-05,
+      "loss": 0.6062,
+      "step": 6607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6807682325460683e-05,
+      "loss": 0.4821,
+      "step": 6608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.680710557429997e-05,
+      "loss": 0.5196,
+      "step": 6609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6806528823139258e-05,
+      "loss": 0.6253,
+      "step": 6610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6805952071978546e-05,
+      "loss": 0.4847,
+      "step": 6611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6805375320817834e-05,
+      "loss": 0.5567,
+      "step": 6612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6804798569657125e-05,
+      "loss": 0.4997,
+      "step": 6613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6804221818496412e-05,
+      "loss": 0.4747,
+      "step": 6614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.68036450673357e-05,
+      "loss": 0.4373,
+      "step": 6615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6803068316174988e-05,
+      "loss": 0.6249,
+      "step": 6616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6802491565014275e-05,
+      "loss": 0.5743,
+      "step": 6617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6801914813853563e-05,
+      "loss": 0.6469,
+      "step": 6618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6801338062692854e-05,
+      "loss": 0.4848,
+      "step": 6619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6800761311532142e-05,
+      "loss": 0.7287,
+      "step": 6620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.680018456037143e-05,
+      "loss": 0.5894,
+      "step": 6621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6799607809210717e-05,
+      "loss": 0.5811,
+      "step": 6622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6799031058050005e-05,
+      "loss": 0.5922,
+      "step": 6623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6798454306889293e-05,
+      "loss": 0.6576,
+      "step": 6624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6797877555728584e-05,
+      "loss": 0.4802,
+      "step": 6625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679730080456787e-05,
+      "loss": 0.5157,
+      "step": 6626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679672405340716e-05,
+      "loss": 0.4827,
+      "step": 6627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6796147302246447e-05,
+      "loss": 0.6051,
+      "step": 6628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6795570551085734e-05,
+      "loss": 0.6082,
+      "step": 6629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6794993799925022e-05,
+      "loss": 0.5382,
+      "step": 6630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679441704876431e-05,
+      "loss": 0.6485,
+      "step": 6631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.67938402976036e-05,
+      "loss": 0.657,
+      "step": 6632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679326354644289e-05,
+      "loss": 0.5492,
+      "step": 6633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6792686795282176e-05,
+      "loss": 0.5064,
+      "step": 6634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6792110044121467e-05,
+      "loss": 0.5201,
+      "step": 6635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6791533292960755e-05,
+      "loss": 0.6692,
+      "step": 6636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679095654180004e-05,
+      "loss": 0.647,
+      "step": 6637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.679037979063933e-05,
+      "loss": 0.4741,
+      "step": 6638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6789803039478618e-05,
+      "loss": 0.5232,
+      "step": 6639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6789226288317906e-05,
+      "loss": 0.5733,
+      "step": 6640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6788649537157197e-05,
+      "loss": 0.6126,
+      "step": 6641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6788072785996484e-05,
+      "loss": 0.6011,
+      "step": 6642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6787496034835772e-05,
+      "loss": 0.5763,
+      "step": 6643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.678691928367506e-05,
+      "loss": 0.6879,
+      "step": 6644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6786342532514347e-05,
+      "loss": 0.6331,
+      "step": 6645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6785765781353635e-05,
+      "loss": 0.6401,
+      "step": 6646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6785189030192926e-05,
+      "loss": 0.5729,
+      "step": 6647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6784612279032214e-05,
+      "loss": 0.5524,
+      "step": 6648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.67840355278715e-05,
+      "loss": 0.4986,
+      "step": 6649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.678345877671079e-05,
+      "loss": 0.582,
+      "step": 6650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6782882025550077e-05,
+      "loss": 0.5836,
+      "step": 6651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6782305274389365e-05,
+      "loss": 0.5463,
+      "step": 6652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6781728523228656e-05,
+      "loss": 0.6448,
+      "step": 6653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6781151772067943e-05,
+      "loss": 0.5595,
+      "step": 6654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.678057502090723e-05,
+      "loss": 0.3916,
+      "step": 6655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677999826974652e-05,
+      "loss": 0.4475,
+      "step": 6656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6779421518585806e-05,
+      "loss": 0.5741,
+      "step": 6657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6778844767425094e-05,
+      "loss": 0.526,
+      "step": 6658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6778268016264385e-05,
+      "loss": 0.6036,
+      "step": 6659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6777691265103673e-05,
+      "loss": 0.6814,
+      "step": 6660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677711451394296e-05,
+      "loss": 0.6909,
+      "step": 6661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6776537762782248e-05,
+      "loss": 0.4826,
+      "step": 6662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677596101162154e-05,
+      "loss": 0.4586,
+      "step": 6663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6775384260460824e-05,
+      "loss": 0.5801,
+      "step": 6664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677480750930011e-05,
+      "loss": 0.6181,
+      "step": 6665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6774230758139402e-05,
+      "loss": 0.4923,
+      "step": 6666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677365400697869e-05,
+      "loss": 0.6202,
+      "step": 6667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6773077255817978e-05,
+      "loss": 0.5999,
+      "step": 6668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677250050465727e-05,
+      "loss": 0.5535,
+      "step": 6669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6771923753496557e-05,
+      "loss": 0.5097,
+      "step": 6670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6771347002335844e-05,
+      "loss": 0.5557,
+      "step": 6671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6770770251175132e-05,
+      "loss": 0.5702,
+      "step": 6672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.677019350001442e-05,
+      "loss": 0.5424,
+      "step": 6673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6769616748853707e-05,
+      "loss": 0.6119,
+      "step": 6674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6769039997693e-05,
+      "loss": 0.5346,
+      "step": 6675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6768463246532286e-05,
+      "loss": 0.4822,
+      "step": 6676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6767886495371574e-05,
+      "loss": 0.549,
+      "step": 6677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.676730974421086e-05,
+      "loss": 0.4266,
+      "step": 6678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.676673299305015e-05,
+      "loss": 0.4046,
+      "step": 6679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6766156241889437e-05,
+      "loss": 0.6925,
+      "step": 6680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6765579490728728e-05,
+      "loss": 0.5828,
+      "step": 6681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6765002739568016e-05,
+      "loss": 0.5536,
+      "step": 6682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6764425988407303e-05,
+      "loss": 0.5181,
+      "step": 6683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.676384923724659e-05,
+      "loss": 0.5184,
+      "step": 6684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.676327248608588e-05,
+      "loss": 0.7006,
+      "step": 6685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6762695734925166e-05,
+      "loss": 0.4769,
+      "step": 6686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6762118983764457e-05,
+      "loss": 0.5907,
+      "step": 6687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6761542232603745e-05,
+      "loss": 0.6998,
+      "step": 6688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6760965481443033e-05,
+      "loss": 0.4836,
+      "step": 6689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.676038873028232e-05,
+      "loss": 0.5516,
+      "step": 6690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6759811979121608e-05,
+      "loss": 0.5631,
+      "step": 6691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6759235227960896e-05,
+      "loss": 0.5488,
+      "step": 6692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6758658476800187e-05,
+      "loss": 0.5331,
+      "step": 6693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6758081725639475e-05,
+      "loss": 0.5986,
+      "step": 6694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6757504974478762e-05,
+      "loss": 0.5919,
+      "step": 6695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675692822331805e-05,
+      "loss": 0.5882,
+      "step": 6696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675635147215734e-05,
+      "loss": 0.6102,
+      "step": 6697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675577472099663e-05,
+      "loss": 0.7407,
+      "step": 6698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6755197969835913e-05,
+      "loss": 0.5372,
+      "step": 6699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.6754621218675204e-05,
+      "loss": 0.4834,
+      "step": 6700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675404446751449e-05,
+      "loss": 0.5561,
+      "step": 6701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675346771635378e-05,
+      "loss": 0.5953,
+      "step": 6702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.675289096519307e-05,
+      "loss": 0.5261,
+      "step": 6703
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6752314214032358e-05,
+      "loss": 0.5421,
+      "step": 6704
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6751737462871646e-05,
+      "loss": 0.4987,
+      "step": 6705
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6751160711710934e-05,
+      "loss": 0.4639,
+      "step": 6706
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.675058396055022e-05,
+      "loss": 0.5506,
+      "step": 6707
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.675000720938951e-05,
+      "loss": 0.5377,
+      "step": 6708
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.67494304582288e-05,
+      "loss": 0.5365,
+      "step": 6709
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6748853707068088e-05,
+      "loss": 0.5447,
+      "step": 6710
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6748276955907375e-05,
+      "loss": 0.6142,
+      "step": 6711
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6747700204746663e-05,
+      "loss": 0.5455,
+      "step": 6712
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.674712345358595e-05,
+      "loss": 0.6953,
+      "step": 6713
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.674654670242524e-05,
+      "loss": 0.6163,
+      "step": 6714
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.674596995126453e-05,
+      "loss": 0.7017,
+      "step": 6715
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6745393200103817e-05,
+      "loss": 0.7948,
+      "step": 6716
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6744816448943105e-05,
+      "loss": 0.4618,
+      "step": 6717
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6744239697782393e-05,
+      "loss": 0.4427,
+      "step": 6718
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.674366294662168e-05,
+      "loss": 0.7158,
+      "step": 6719
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6743086195460968e-05,
+      "loss": 0.5341,
+      "step": 6720
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.674250944430026e-05,
+      "loss": 0.5721,
+      "step": 6721
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6741932693139547e-05,
+      "loss": 0.4689,
+      "step": 6722
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6741355941978834e-05,
+      "loss": 0.5808,
+      "step": 6723
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6740779190818125e-05,
+      "loss": 0.5199,
+      "step": 6724
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6740202439657413e-05,
+      "loss": 0.5863,
+      "step": 6725
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6739625688496697e-05,
+      "loss": 0.4011,
+      "step": 6726
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.673904893733599e-05,
+      "loss": 0.4418,
+      "step": 6727
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6738472186175276e-05,
+      "loss": 0.5777,
+      "step": 6728
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6737895435014564e-05,
+      "loss": 0.5673,
+      "step": 6729
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6737318683853855e-05,
+      "loss": 0.6517,
+      "step": 6730
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6736741932693143e-05,
+      "loss": 0.699,
+      "step": 6731
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.673616518153243e-05,
+      "loss": 0.4934,
+      "step": 6732
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6735588430371718e-05,
+      "loss": 0.5437,
+      "step": 6733
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6735011679211006e-05,
+      "loss": 0.4518,
+      "step": 6734
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6734434928050293e-05,
+      "loss": 0.5379,
+      "step": 6735
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.673385817688958e-05,
+      "loss": 0.6672,
+      "step": 6736
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6733281425728872e-05,
+      "loss": 0.4055,
+      "step": 6737
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.673270467456816e-05,
+      "loss": 0.5189,
+      "step": 6738
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6732127923407447e-05,
+      "loss": 0.5592,
+      "step": 6739
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6731551172246735e-05,
+      "loss": 0.4875,
+      "step": 6740
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6730974421086023e-05,
+      "loss": 0.4964,
+      "step": 6741
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.673039766992531e-05,
+      "loss": 0.5848,
+      "step": 6742
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.67298209187646e-05,
+      "loss": 0.4269,
+      "step": 6743
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672924416760389e-05,
+      "loss": 0.5403,
+      "step": 6744
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6728667416443177e-05,
+      "loss": 0.5493,
+      "step": 6745
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6728090665282465e-05,
+      "loss": 0.5621,
+      "step": 6746
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6727513914121752e-05,
+      "loss": 0.7097,
+      "step": 6747
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672693716296104e-05,
+      "loss": 0.5626,
+      "step": 6748
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672636041180033e-05,
+      "loss": 0.5501,
+      "step": 6749
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672578366063962e-05,
+      "loss": 0.5722,
+      "step": 6750
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6725206909478906e-05,
+      "loss": 0.4613,
+      "step": 6751
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6724630158318194e-05,
+      "loss": 0.702,
+      "step": 6752
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6724053407157482e-05,
+      "loss": 0.572,
+      "step": 6753
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672347665599677e-05,
+      "loss": 0.5007,
+      "step": 6754
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.672289990483606e-05,
+      "loss": 0.5417,
+      "step": 6755
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6722323153675348e-05,
+      "loss": 0.4073,
+      "step": 6756
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6721746402514636e-05,
+      "loss": 0.6904,
+      "step": 6757
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6721169651353927e-05,
+      "loss": 0.5638,
+      "step": 6758
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6720592900193215e-05,
+      "loss": 0.6635,
+      "step": 6759
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6720016149032502e-05,
+      "loss": 0.6087,
+      "step": 6760
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.671943939787179e-05,
+      "loss": 0.4986,
+      "step": 6761
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6718862646711078e-05,
+      "loss": 0.6764,
+      "step": 6762
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6718285895550365e-05,
+      "loss": 0.6101,
+      "step": 6763
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6717709144389657e-05,
+      "loss": 0.6067,
+      "step": 6764
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6717132393228944e-05,
+      "loss": 0.5001,
+      "step": 6765
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6716555642068232e-05,
+      "loss": 0.5692,
+      "step": 6766
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.671597889090752e-05,
+      "loss": 0.5949,
+      "step": 6767
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6715402139746807e-05,
+      "loss": 0.5394,
+      "step": 6768
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6714825388586095e-05,
+      "loss": 0.4691,
+      "step": 6769
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6714248637425383e-05,
+      "loss": 0.7126,
+      "step": 6770
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6713671886264674e-05,
+      "loss": 0.5478,
+      "step": 6771
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.671309513510396e-05,
+      "loss": 0.5747,
+      "step": 6772
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.671251838394325e-05,
+      "loss": 0.644,
+      "step": 6773
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6711941632782537e-05,
+      "loss": 0.5598,
+      "step": 6774
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6711364881621824e-05,
+      "loss": 0.5811,
+      "step": 6775
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6710788130461112e-05,
+      "loss": 0.5151,
+      "step": 6776
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6710211379300403e-05,
+      "loss": 0.6051,
+      "step": 6777
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670963462813969e-05,
+      "loss": 0.5249,
+      "step": 6778
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670905787697898e-05,
+      "loss": 0.5029,
+      "step": 6779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6708481125818266e-05,
+      "loss": 0.4528,
+      "step": 6780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6707904374657554e-05,
+      "loss": 0.5851,
+      "step": 6781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670732762349684e-05,
+      "loss": 0.371,
+      "step": 6782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6706750872336133e-05,
+      "loss": 0.4898,
+      "step": 6783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670617412117542e-05,
+      "loss": 0.492,
+      "step": 6784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6705597370014708e-05,
+      "loss": 0.6062,
+      "step": 6785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6705020618854e-05,
+      "loss": 0.5578,
+      "step": 6786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6704443867693287e-05,
+      "loss": 0.6055,
+      "step": 6787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670386711653257e-05,
+      "loss": 0.5998,
+      "step": 6788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6703290365371862e-05,
+      "loss": 0.5605,
+      "step": 6789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670271361421115e-05,
+      "loss": 0.5846,
+      "step": 6790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6702136863050438e-05,
+      "loss": 0.6404,
+      "step": 6791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.670156011188973e-05,
+      "loss": 0.4507,
+      "step": 6792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6700983360729016e-05,
+      "loss": 0.5359,
+      "step": 6793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6700406609568304e-05,
+      "loss": 0.5081,
+      "step": 6794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669982985840759e-05,
+      "loss": 0.403,
+      "step": 6795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669925310724688e-05,
+      "loss": 0.5847,
+      "step": 6796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6698676356086167e-05,
+      "loss": 0.5382,
+      "step": 6797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6698099604925458e-05,
+      "loss": 0.535,
+      "step": 6798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6697522853764746e-05,
+      "loss": 0.5612,
+      "step": 6799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6696946102604034e-05,
+      "loss": 0.4569,
+      "step": 6800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669636935144332e-05,
+      "loss": 0.4828,
+      "step": 6801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669579260028261e-05,
+      "loss": 0.6547,
+      "step": 6802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6695215849121897e-05,
+      "loss": 0.5876,
+      "step": 6803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6694639097961184e-05,
+      "loss": 0.5818,
+      "step": 6804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6694062346800475e-05,
+      "loss": 0.6169,
+      "step": 6805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6693485595639763e-05,
+      "loss": 0.4718,
+      "step": 6806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669290884447905e-05,
+      "loss": 0.6042,
+      "step": 6807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.669233209331834e-05,
+      "loss": 0.5712,
+      "step": 6808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6691755342157626e-05,
+      "loss": 0.5807,
+      "step": 6809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6691178590996914e-05,
+      "loss": 0.5936,
+      "step": 6810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6690601839836205e-05,
+      "loss": 0.5746,
+      "step": 6811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6690025088675493e-05,
+      "loss": 0.5781,
+      "step": 6812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668944833751478e-05,
+      "loss": 0.5854,
+      "step": 6813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6688871586354068e-05,
+      "loss": 0.624,
+      "step": 6814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6688294835193356e-05,
+      "loss": 0.5827,
+      "step": 6815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6687718084032643e-05,
+      "loss": 0.6132,
+      "step": 6816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6687141332871934e-05,
+      "loss": 0.5289,
+      "step": 6817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6686564581711222e-05,
+      "loss": 0.4997,
+      "step": 6818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668598783055051e-05,
+      "loss": 0.6111,
+      "step": 6819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.66854110793898e-05,
+      "loss": 0.6426,
+      "step": 6820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668483432822909e-05,
+      "loss": 0.4266,
+      "step": 6821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6684257577068376e-05,
+      "loss": 0.5072,
+      "step": 6822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6683680825907664e-05,
+      "loss": 0.4925,
+      "step": 6823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668310407474695e-05,
+      "loss": 0.5432,
+      "step": 6824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668252732358624e-05,
+      "loss": 0.6578,
+      "step": 6825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.668195057242553e-05,
+      "loss": 0.431,
+      "step": 6826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6681373821264818e-05,
+      "loss": 0.7151,
+      "step": 6827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6680797070104106e-05,
+      "loss": 0.6405,
+      "step": 6828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6680220318943393e-05,
+      "loss": 0.5709,
+      "step": 6829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.667964356778268e-05,
+      "loss": 0.6331,
+      "step": 6830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.667906681662197e-05,
+      "loss": 0.5528,
+      "step": 6831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.667849006546126e-05,
+      "loss": 0.5624,
+      "step": 6832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6677913314300547e-05,
+      "loss": 0.5451,
+      "step": 6833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6677336563139835e-05,
+      "loss": 0.4713,
+      "step": 6834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6676759811979123e-05,
+      "loss": 0.5872,
+      "step": 6835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.667618306081841e-05,
+      "loss": 0.6653,
+      "step": 6836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6675606309657698e-05,
+      "loss": 0.5879,
+      "step": 6837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6675029558496986e-05,
+      "loss": 0.6292,
+      "step": 6838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6674452807336277e-05,
+      "loss": 0.6081,
+      "step": 6839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6673876056175565e-05,
+      "loss": 0.5737,
+      "step": 6840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6673299305014852e-05,
+      "loss": 0.6024,
+      "step": 6841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.667272255385414e-05,
+      "loss": 0.5171,
+      "step": 6842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6672145802693428e-05,
+      "loss": 0.5541,
+      "step": 6843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6671569051532715e-05,
+      "loss": 0.5374,
+      "step": 6844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6670992300372006e-05,
+      "loss": 0.648,
+      "step": 6845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6670415549211294e-05,
+      "loss": 0.4909,
+      "step": 6846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6669838798050582e-05,
+      "loss": 0.6085,
+      "step": 6847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6669262046889873e-05,
+      "loss": 0.5024,
+      "step": 6848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.666868529572916e-05,
+      "loss": 0.5995,
+      "step": 6849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6668108544568445e-05,
+      "loss": 0.6228,
+      "step": 6850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6667531793407736e-05,
+      "loss": 0.5222,
+      "step": 6851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6666955042247024e-05,
+      "loss": 0.5933,
+      "step": 6852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.666637829108631e-05,
+      "loss": 0.5172,
+      "step": 6853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6665801539925602e-05,
+      "loss": 0.461,
+      "step": 6854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.666522478876489e-05,
+      "loss": 0.534,
+      "step": 6855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6664648037604178e-05,
+      "loss": 0.6186,
+      "step": 6856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6664071286443465e-05,
+      "loss": 0.6162,
+      "step": 6857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6663494535282753e-05,
+      "loss": 0.5255,
+      "step": 6858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.666291778412204e-05,
+      "loss": 0.5782,
+      "step": 6859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6662341032961332e-05,
+      "loss": 0.5784,
+      "step": 6860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.666176428180062e-05,
+      "loss": 0.4856,
+      "step": 6861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6661187530639907e-05,
+      "loss": 0.4093,
+      "step": 6862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6660610779479195e-05,
+      "loss": 0.4486,
+      "step": 6863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6660034028318483e-05,
+      "loss": 0.6852,
+      "step": 6864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.665945727715777e-05,
+      "loss": 0.6034,
+      "step": 6865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.665888052599706e-05,
+      "loss": 0.4495,
+      "step": 6866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.665830377483635e-05,
+      "loss": 0.5069,
+      "step": 6867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6657727023675637e-05,
+      "loss": 0.5728,
+      "step": 6868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6657150272514924e-05,
+      "loss": 0.5106,
+      "step": 6869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6656573521354212e-05,
+      "loss": 0.5527,
+      "step": 6870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.66559967701935e-05,
+      "loss": 0.5793,
+      "step": 6871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6655420019032787e-05,
+      "loss": 0.5283,
+      "step": 6872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.665484326787208e-05,
+      "loss": 0.5445,
+      "step": 6873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6654266516711366e-05,
+      "loss": 0.6122,
+      "step": 6874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6653689765550654e-05,
+      "loss": 0.4593,
+      "step": 6875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6653113014389945e-05,
+      "loss": 0.5616,
+      "step": 6876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.665253626322923e-05,
+      "loss": 0.509,
+      "step": 6877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6651959512068517e-05,
+      "loss": 0.5297,
+      "step": 6878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6651382760907808e-05,
+      "loss": 0.5399,
+      "step": 6879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6650806009747096e-05,
+      "loss": 0.5396,
+      "step": 6880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.6650229258586383e-05,
+      "loss": 0.6029,
+      "step": 6881
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6649652507425675e-05,
+      "loss": 0.6282,
+      "step": 6882
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6649075756264962e-05,
+      "loss": 0.5151,
+      "step": 6883
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.664849900510425e-05,
+      "loss": 0.575,
+      "step": 6884
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6647922253943538e-05,
+      "loss": 0.4758,
+      "step": 6885
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6647345502782825e-05,
+      "loss": 0.4595,
+      "step": 6886
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6646768751622113e-05,
+      "loss": 0.4936,
+      "step": 6887
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6646192000461404e-05,
+      "loss": 0.5536,
+      "step": 6888
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.664561524930069e-05,
+      "loss": 0.5088,
+      "step": 6889
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.664503849813998e-05,
+      "loss": 0.5258,
+      "step": 6890
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6644461746979267e-05,
+      "loss": 0.5838,
+      "step": 6891
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6643884995818555e-05,
+      "loss": 0.5095,
+      "step": 6892
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6643308244657842e-05,
+      "loss": 0.5443,
+      "step": 6893
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6642731493497134e-05,
+      "loss": 0.5313,
+      "step": 6894
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.664215474233642e-05,
+      "loss": 0.5413,
+      "step": 6895
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.664157799117571e-05,
+      "loss": 0.6006,
+      "step": 6896
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6641001240014997e-05,
+      "loss": 0.5585,
+      "step": 6897
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6640424488854284e-05,
+      "loss": 0.7007,
+      "step": 6898
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6639847737693572e-05,
+      "loss": 0.4929,
+      "step": 6899
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6639270986532863e-05,
+      "loss": 0.5678,
+      "step": 6900
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663869423537215e-05,
+      "loss": 0.5136,
+      "step": 6901
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663811748421144e-05,
+      "loss": 0.5605,
+      "step": 6902
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6637540733050726e-05,
+      "loss": 0.6038,
+      "step": 6903
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6636963981890014e-05,
+      "loss": 0.4325,
+      "step": 6904
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.66363872307293e-05,
+      "loss": 0.5104,
+      "step": 6905
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663581047956859e-05,
+      "loss": 0.487,
+      "step": 6906
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663523372840788e-05,
+      "loss": 0.6989,
+      "step": 6907
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6634656977247168e-05,
+      "loss": 0.4534,
+      "step": 6908
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6634080226086456e-05,
+      "loss": 0.5215,
+      "step": 6909
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6633503474925747e-05,
+      "loss": 0.4884,
+      "step": 6910
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6632926723765034e-05,
+      "loss": 0.5058,
+      "step": 6911
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663234997260432e-05,
+      "loss": 0.6641,
+      "step": 6912
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.663177322144361e-05,
+      "loss": 0.5789,
+      "step": 6913
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6631196470282897e-05,
+      "loss": 0.5911,
+      "step": 6914
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6630619719122185e-05,
+      "loss": 0.391,
+      "step": 6915
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6630042967961476e-05,
+      "loss": 0.6495,
+      "step": 6916
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6629466216800764e-05,
+      "loss": 0.5841,
+      "step": 6917
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.662888946564005e-05,
+      "loss": 0.6266,
+      "step": 6918
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.662831271447934e-05,
+      "loss": 0.5526,
+      "step": 6919
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6627735963318627e-05,
+      "loss": 0.6013,
+      "step": 6920
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6627159212157915e-05,
+      "loss": 0.6376,
+      "step": 6921
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6626582460997206e-05,
+      "loss": 0.5168,
+      "step": 6922
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6626005709836493e-05,
+      "loss": 0.5497,
+      "step": 6923
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.662542895867578e-05,
+      "loss": 0.5723,
+      "step": 6924
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.662485220751507e-05,
+      "loss": 0.4655,
+      "step": 6925
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6624275456354356e-05,
+      "loss": 0.5644,
+      "step": 6926
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6623698705193644e-05,
+      "loss": 0.5143,
+      "step": 6927
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6623121954032935e-05,
+      "loss": 0.5863,
+      "step": 6928
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6622545202872223e-05,
+      "loss": 0.5652,
+      "step": 6929
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.662196845171151e-05,
+      "loss": 0.5833,
+      "step": 6930
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6621391700550798e-05,
+      "loss": 0.5405,
+      "step": 6931
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6620814949390086e-05,
+      "loss": 0.5638,
+      "step": 6932
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6620238198229374e-05,
+      "loss": 0.4429,
+      "step": 6933
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6619661447068665e-05,
+      "loss": 0.4836,
+      "step": 6934
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6619084695907952e-05,
+      "loss": 0.4278,
+      "step": 6935
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661850794474724e-05,
+      "loss": 0.6287,
+      "step": 6936
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661793119358653e-05,
+      "loss": 0.5515,
+      "step": 6937
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661735444242582e-05,
+      "loss": 0.5508,
+      "step": 6938
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6616777691265103e-05,
+      "loss": 0.5264,
+      "step": 6939
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661620094010439e-05,
+      "loss": 0.6111,
+      "step": 6940
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6615624188943682e-05,
+      "loss": 0.5036,
+      "step": 6941
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661504743778297e-05,
+      "loss": 0.703,
+      "step": 6942
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6614470686622257e-05,
+      "loss": 0.4709,
+      "step": 6943
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6613893935461548e-05,
+      "loss": 0.5047,
+      "step": 6944
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6613317184300836e-05,
+      "loss": 0.6295,
+      "step": 6945
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6612740433140124e-05,
+      "loss": 0.5545,
+      "step": 6946
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.661216368197941e-05,
+      "loss": 0.5391,
+      "step": 6947
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.66115869308187e-05,
+      "loss": 0.4527,
+      "step": 6948
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6611010179657987e-05,
+      "loss": 0.5742,
+      "step": 6949
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6610433428497278e-05,
+      "loss": 0.5078,
+      "step": 6950
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6609856677336565e-05,
+      "loss": 0.5255,
+      "step": 6951
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6609279926175853e-05,
+      "loss": 0.6712,
+      "step": 6952
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.660870317501514e-05,
+      "loss": 0.5722,
+      "step": 6953
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.660812642385443e-05,
+      "loss": 0.6065,
+      "step": 6954
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6607549672693716e-05,
+      "loss": 0.5177,
+      "step": 6955
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6606972921533007e-05,
+      "loss": 0.5418,
+      "step": 6956
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6606396170372295e-05,
+      "loss": 0.5725,
+      "step": 6957
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6605819419211583e-05,
+      "loss": 0.5823,
+      "step": 6958
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.660524266805087e-05,
+      "loss": 0.5803,
+      "step": 6959
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6604665916890158e-05,
+      "loss": 0.5554,
+      "step": 6960
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6604089165729446e-05,
+      "loss": 0.5619,
+      "step": 6961
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6603512414568737e-05,
+      "loss": 0.5456,
+      "step": 6962
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6602935663408024e-05,
+      "loss": 0.606,
+      "step": 6963
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6602358912247312e-05,
+      "loss": 0.6129,
+      "step": 6964
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.66017821610866e-05,
+      "loss": 0.5566,
+      "step": 6965
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6601205409925887e-05,
+      "loss": 0.4829,
+      "step": 6966
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6600628658765175e-05,
+      "loss": 0.593,
+      "step": 6967
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6600051907604466e-05,
+      "loss": 0.567,
+      "step": 6968
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6599475156443754e-05,
+      "loss": 0.5828,
+      "step": 6969
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659889840528304e-05,
+      "loss": 0.5447,
+      "step": 6970
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6598321654122333e-05,
+      "loss": 0.5957,
+      "step": 6971
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659774490296162e-05,
+      "loss": 0.5651,
+      "step": 6972
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6597168151800908e-05,
+      "loss": 0.5317,
+      "step": 6973
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6596591400640196e-05,
+      "loss": 0.5815,
+      "step": 6974
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6596014649479483e-05,
+      "loss": 0.4211,
+      "step": 6975
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659543789831877e-05,
+      "loss": 0.5976,
+      "step": 6976
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659486114715806e-05,
+      "loss": 0.518,
+      "step": 6977
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659428439599735e-05,
+      "loss": 0.4975,
+      "step": 6978
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6593707644836638e-05,
+      "loss": 0.5986,
+      "step": 6979
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6593130893675925e-05,
+      "loss": 0.5891,
+      "step": 6980
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6592554142515213e-05,
+      "loss": 0.5078,
+      "step": 6981
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.65919773913545e-05,
+      "loss": 0.6271,
+      "step": 6982
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6591400640193788e-05,
+      "loss": 0.538,
+      "step": 6983
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.659082388903308e-05,
+      "loss": 0.5121,
+      "step": 6984
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6590247137872367e-05,
+      "loss": 0.403,
+      "step": 6985
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6589670386711655e-05,
+      "loss": 0.5235,
+      "step": 6986
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6589093635550942e-05,
+      "loss": 0.5149,
+      "step": 6987
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.658851688439023e-05,
+      "loss": 0.5689,
+      "step": 6988
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6587940133229518e-05,
+      "loss": 0.5884,
+      "step": 6989
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.658736338206881e-05,
+      "loss": 0.6012,
+      "step": 6990
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6586786630908097e-05,
+      "loss": 0.5309,
+      "step": 6991
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6586209879747384e-05,
+      "loss": 0.4518,
+      "step": 6992
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6585633128586672e-05,
+      "loss": 0.5394,
+      "step": 6993
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.658505637742596e-05,
+      "loss": 0.6649,
+      "step": 6994
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6584479626265247e-05,
+      "loss": 0.4786,
+      "step": 6995
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.658390287510454e-05,
+      "loss": 0.5193,
+      "step": 6996
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6583326123943826e-05,
+      "loss": 0.5888,
+      "step": 6997
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6582749372783114e-05,
+      "loss": 0.52,
+      "step": 6998
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6582172621622405e-05,
+      "loss": 0.6527,
+      "step": 6999
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6581595870461692e-05,
+      "loss": 0.613,
+      "step": 7000
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6581019119300977e-05,
+      "loss": 0.5537,
+      "step": 7001
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6580442368140268e-05,
+      "loss": 0.5527,
+      "step": 7002
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6579865616979556e-05,
+      "loss": 0.5447,
+      "step": 7003
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6579288865818843e-05,
+      "loss": 0.5457,
+      "step": 7004
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6578712114658134e-05,
+      "loss": 0.5451,
+      "step": 7005
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6578135363497422e-05,
+      "loss": 0.6323,
+      "step": 7006
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657755861233671e-05,
+      "loss": 0.5308,
+      "step": 7007
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6576981861175997e-05,
+      "loss": 0.492,
+      "step": 7008
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6576405110015285e-05,
+      "loss": 0.5172,
+      "step": 7009
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6575828358854573e-05,
+      "loss": 0.5588,
+      "step": 7010
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657525160769386e-05,
+      "loss": 0.571,
+      "step": 7011
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657467485653315e-05,
+      "loss": 0.5945,
+      "step": 7012
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657409810537244e-05,
+      "loss": 0.5626,
+      "step": 7013
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6573521354211727e-05,
+      "loss": 0.6046,
+      "step": 7014
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6572944603051015e-05,
+      "loss": 0.5228,
+      "step": 7015
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6572367851890302e-05,
+      "loss": 0.5067,
+      "step": 7016
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657179110072959e-05,
+      "loss": 0.5896,
+      "step": 7017
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657121434956888e-05,
+      "loss": 0.445,
+      "step": 7018
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.657063759840817e-05,
+      "loss": 0.6099,
+      "step": 7019
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6570060847247456e-05,
+      "loss": 0.6174,
+      "step": 7020
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6569484096086744e-05,
+      "loss": 0.5654,
+      "step": 7021
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6568907344926032e-05,
+      "loss": 0.4941,
+      "step": 7022
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656833059376532e-05,
+      "loss": 0.5293,
+      "step": 7023
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656775384260461e-05,
+      "loss": 0.5267,
+      "step": 7024
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6567177091443898e-05,
+      "loss": 0.5288,
+      "step": 7025
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6566600340283186e-05,
+      "loss": 0.4872,
+      "step": 7026
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6566023589122474e-05,
+      "loss": 0.5356,
+      "step": 7027
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656544683796176e-05,
+      "loss": 0.5551,
+      "step": 7028
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656487008680105e-05,
+      "loss": 0.5246,
+      "step": 7029
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656429333564034e-05,
+      "loss": 0.5636,
+      "step": 7030
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6563716584479628e-05,
+      "loss": 0.575,
+      "step": 7031
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6563139833318915e-05,
+      "loss": 0.5449,
+      "step": 7032
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6562563082158206e-05,
+      "loss": 0.602,
+      "step": 7033
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6561986330997494e-05,
+      "loss": 0.5733,
+      "step": 7034
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6561409579836782e-05,
+      "loss": 0.4825,
+      "step": 7035
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.656083282867607e-05,
+      "loss": 0.5691,
+      "step": 7036
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6560256077515357e-05,
+      "loss": 0.6568,
+      "step": 7037
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6559679326354645e-05,
+      "loss": 0.6901,
+      "step": 7038
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6559102575193936e-05,
+      "loss": 0.4121,
+      "step": 7039
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6558525824033224e-05,
+      "loss": 0.6506,
+      "step": 7040
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.655794907287251e-05,
+      "loss": 0.5297,
+      "step": 7041
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.65573723217118e-05,
+      "loss": 0.4156,
+      "step": 7042
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6556795570551087e-05,
+      "loss": 0.6166,
+      "step": 7043
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6556218819390374e-05,
+      "loss": 0.6254,
+      "step": 7044
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6555642068229662e-05,
+      "loss": 0.6208,
+      "step": 7045
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6555065317068953e-05,
+      "loss": 0.5889,
+      "step": 7046
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.655448856590824e-05,
+      "loss": 0.6137,
+      "step": 7047
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.655391181474753e-05,
+      "loss": 0.6689,
+      "step": 7048
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6553335063586816e-05,
+      "loss": 0.6431,
+      "step": 7049
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6552758312426104e-05,
+      "loss": 0.4841,
+      "step": 7050
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.655218156126539e-05,
+      "loss": 0.5662,
+      "step": 7051
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6551604810104683e-05,
+      "loss": 0.5132,
+      "step": 7052
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.655102805894397e-05,
+      "loss": 0.5372,
+      "step": 7053
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6550451307783258e-05,
+      "loss": 0.48,
+      "step": 7054
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6549874556622546e-05,
+      "loss": 0.6562,
+      "step": 7055
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6549297805461833e-05,
+      "loss": 0.5885,
+      "step": 7056
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.654872105430112e-05,
+      "loss": 0.5044,
+      "step": 7057
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6548144303140412e-05,
+      "loss": 0.536,
+      "step": 7058
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.65475675519797e-05,
+      "loss": 0.484,
+      "step": 7059
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.6546990800818987e-05,
+      "loss": 0.51,
+      "step": 7060
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654641404965828e-05,
+      "loss": 0.4916,
+      "step": 7061
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6545837298497566e-05,
+      "loss": 0.6834,
+      "step": 7062
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654526054733685e-05,
+      "loss": 0.5725,
+      "step": 7063
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654468379617614e-05,
+      "loss": 0.5764,
+      "step": 7064
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654410704501543e-05,
+      "loss": 0.5964,
+      "step": 7065
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6543530293854717e-05,
+      "loss": 0.6043,
+      "step": 7066
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6542953542694008e-05,
+      "loss": 0.5442,
+      "step": 7067
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6542376791533296e-05,
+      "loss": 0.5502,
+      "step": 7068
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6541800040372583e-05,
+      "loss": 0.6468,
+      "step": 7069
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654122328921187e-05,
+      "loss": 0.675,
+      "step": 7070
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.654064653805116e-05,
+      "loss": 0.6298,
+      "step": 7071
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6540069786890446e-05,
+      "loss": 0.4587,
+      "step": 7072
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6539493035729738e-05,
+      "loss": 0.4454,
+      "step": 7073
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6538916284569025e-05,
+      "loss": 0.6132,
+      "step": 7074
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6538339533408313e-05,
+      "loss": 0.6061,
+      "step": 7075
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.65377627822476e-05,
+      "loss": 0.483,
+      "step": 7076
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6537186031086888e-05,
+      "loss": 0.5574,
+      "step": 7077
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6536609279926176e-05,
+      "loss": 0.5116,
+      "step": 7078
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6536032528765464e-05,
+      "loss": 0.5941,
+      "step": 7079
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6535455777604755e-05,
+      "loss": 0.4961,
+      "step": 7080
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6534879026444042e-05,
+      "loss": 0.5138,
+      "step": 7081
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.653430227528333e-05,
+      "loss": 0.705,
+      "step": 7082
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6533725524122618e-05,
+      "loss": 0.5295,
+      "step": 7083
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6533148772961905e-05,
+      "loss": 0.7475,
+      "step": 7084
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6532572021801193e-05,
+      "loss": 0.5847,
+      "step": 7085
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6531995270640484e-05,
+      "loss": 0.5597,
+      "step": 7086
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6531418519479772e-05,
+      "loss": 0.5988,
+      "step": 7087
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.653084176831906e-05,
+      "loss": 0.6091,
+      "step": 7088
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6530265017158347e-05,
+      "loss": 0.5302,
+      "step": 7089
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6529688265997635e-05,
+      "loss": 0.5288,
+      "step": 7090
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6529111514836923e-05,
+      "loss": 0.5533,
+      "step": 7091
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6528534763676214e-05,
+      "loss": 0.5295,
+      "step": 7092
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.65279580125155e-05,
+      "loss": 0.5353,
+      "step": 7093
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652738126135479e-05,
+      "loss": 0.4712,
+      "step": 7094
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652680451019408e-05,
+      "loss": 0.5253,
+      "step": 7095
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6526227759033368e-05,
+      "loss": 0.4883,
+      "step": 7096
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6525651007872656e-05,
+      "loss": 0.6622,
+      "step": 7097
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6525074256711943e-05,
+      "loss": 0.5248,
+      "step": 7098
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652449750555123e-05,
+      "loss": 0.544,
+      "step": 7099
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652392075439052e-05,
+      "loss": 0.5445,
+      "step": 7100
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652334400322981e-05,
+      "loss": 0.5822,
+      "step": 7101
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6522767252069097e-05,
+      "loss": 0.5457,
+      "step": 7102
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6522190500908385e-05,
+      "loss": 0.5588,
+      "step": 7103
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6521613749747673e-05,
+      "loss": 0.5295,
+      "step": 7104
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.652103699858696e-05,
+      "loss": 0.5603,
+      "step": 7105
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6520460247426248e-05,
+      "loss": 0.5731,
+      "step": 7106
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651988349626554e-05,
+      "loss": 0.5733,
+      "step": 7107
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6519306745104827e-05,
+      "loss": 0.64,
+      "step": 7108
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6518729993944115e-05,
+      "loss": 0.7343,
+      "step": 7109
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6518153242783402e-05,
+      "loss": 0.591,
+      "step": 7110
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651757649162269e-05,
+      "loss": 0.4732,
+      "step": 7111
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6516999740461978e-05,
+      "loss": 0.5075,
+      "step": 7112
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651642298930127e-05,
+      "loss": 0.5071,
+      "step": 7113
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6515846238140556e-05,
+      "loss": 0.6217,
+      "step": 7114
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6515269486979844e-05,
+      "loss": 0.5632,
+      "step": 7115
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6514692735819132e-05,
+      "loss": 0.6746,
+      "step": 7116
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651411598465842e-05,
+      "loss": 0.5547,
+      "step": 7117
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6513539233497707e-05,
+      "loss": 0.5993,
+      "step": 7118
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6512962482336995e-05,
+      "loss": 0.6366,
+      "step": 7119
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6512385731176286e-05,
+      "loss": 0.5779,
+      "step": 7120
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6511808980015574e-05,
+      "loss": 0.664,
+      "step": 7121
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651123222885486e-05,
+      "loss": 0.5615,
+      "step": 7122
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6510655477694152e-05,
+      "loss": 0.587,
+      "step": 7123
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.651007872653344e-05,
+      "loss": 0.4803,
+      "step": 7124
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6509501975372724e-05,
+      "loss": 0.4901,
+      "step": 7125
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6508925224212015e-05,
+      "loss": 0.6027,
+      "step": 7126
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6508348473051303e-05,
+      "loss": 0.5434,
+      "step": 7127
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650777172189059e-05,
+      "loss": 0.5398,
+      "step": 7128
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6507194970729882e-05,
+      "loss": 0.4687,
+      "step": 7129
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650661821956917e-05,
+      "loss": 0.4471,
+      "step": 7130
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6506041468408457e-05,
+      "loss": 0.468,
+      "step": 7131
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6505464717247745e-05,
+      "loss": 0.5366,
+      "step": 7132
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6504887966087033e-05,
+      "loss": 0.6248,
+      "step": 7133
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650431121492632e-05,
+      "loss": 0.5293,
+      "step": 7134
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650373446376561e-05,
+      "loss": 0.4614,
+      "step": 7135
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.65031577126049e-05,
+      "loss": 0.5703,
+      "step": 7136
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6502580961444187e-05,
+      "loss": 0.5387,
+      "step": 7137
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6502004210283474e-05,
+      "loss": 0.5147,
+      "step": 7138
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6501427459122762e-05,
+      "loss": 0.5641,
+      "step": 7139
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650085070796205e-05,
+      "loss": 0.6881,
+      "step": 7140
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.650027395680134e-05,
+      "loss": 0.5908,
+      "step": 7141
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649969720564063e-05,
+      "loss": 0.6616,
+      "step": 7142
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6499120454479916e-05,
+      "loss": 0.5931,
+      "step": 7143
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6498543703319204e-05,
+      "loss": 0.5769,
+      "step": 7144
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649796695215849e-05,
+      "loss": 0.4913,
+      "step": 7145
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649739020099778e-05,
+      "loss": 0.5799,
+      "step": 7146
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649681344983707e-05,
+      "loss": 0.6025,
+      "step": 7147
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6496236698676358e-05,
+      "loss": 0.5738,
+      "step": 7148
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6495659947515646e-05,
+      "loss": 0.5679,
+      "step": 7149
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6495083196354933e-05,
+      "loss": 0.6598,
+      "step": 7150
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649450644519422e-05,
+      "loss": 0.4805,
+      "step": 7151
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649392969403351e-05,
+      "loss": 0.5434,
+      "step": 7152
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6493352942872796e-05,
+      "loss": 0.623,
+      "step": 7153
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6492776191712087e-05,
+      "loss": 0.5217,
+      "step": 7154
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6492199440551375e-05,
+      "loss": 0.6142,
+      "step": 7155
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6491622689390663e-05,
+      "loss": 0.5405,
+      "step": 7156
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6491045938229954e-05,
+      "loss": 0.6715,
+      "step": 7157
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.649046918706924e-05,
+      "loss": 0.6017,
+      "step": 7158
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.648989243590853e-05,
+      "loss": 0.5395,
+      "step": 7159
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6489315684747817e-05,
+      "loss": 0.5159,
+      "step": 7160
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6488738933587105e-05,
+      "loss": 0.5805,
+      "step": 7161
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6488162182426392e-05,
+      "loss": 0.5452,
+      "step": 7162
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6487585431265683e-05,
+      "loss": 0.5027,
+      "step": 7163
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.648700868010497e-05,
+      "loss": 0.519,
+      "step": 7164
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.648643192894426e-05,
+      "loss": 0.5616,
+      "step": 7165
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6485855177783546e-05,
+      "loss": 0.5798,
+      "step": 7166
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6485278426622834e-05,
+      "loss": 0.4722,
+      "step": 7167
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6484701675462122e-05,
+      "loss": 0.6176,
+      "step": 7168
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6484124924301413e-05,
+      "loss": 0.5353,
+      "step": 7169
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.64835481731407e-05,
+      "loss": 0.6118,
+      "step": 7170
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6482971421979988e-05,
+      "loss": 0.5275,
+      "step": 7171
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6482394670819276e-05,
+      "loss": 0.4743,
+      "step": 7172
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6481817919658564e-05,
+      "loss": 0.4209,
+      "step": 7173
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.648124116849785e-05,
+      "loss": 0.55,
+      "step": 7174
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6480664417337142e-05,
+      "loss": 0.541,
+      "step": 7175
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.648008766617643e-05,
+      "loss": 0.5525,
+      "step": 7176
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6479510915015718e-05,
+      "loss": 0.5545,
+      "step": 7177
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6478934163855005e-05,
+      "loss": 0.6367,
+      "step": 7178
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6478357412694293e-05,
+      "loss": 0.5564,
+      "step": 7179
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.647778066153358e-05,
+      "loss": 0.5756,
+      "step": 7180
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6477203910372872e-05,
+      "loss": 0.4652,
+      "step": 7181
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.647662715921216e-05,
+      "loss": 0.4606,
+      "step": 7182
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6476050408051447e-05,
+      "loss": 0.5063,
+      "step": 7183
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6475473656890735e-05,
+      "loss": 0.5845,
+      "step": 7184
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6474896905730026e-05,
+      "loss": 0.685,
+      "step": 7185
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6474320154569314e-05,
+      "loss": 0.6639,
+      "step": 7186
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6473743403408598e-05,
+      "loss": 0.493,
+      "step": 7187
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.647316665224789e-05,
+      "loss": 0.542,
+      "step": 7188
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6472589901087177e-05,
+      "loss": 0.5825,
+      "step": 7189
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6472013149926464e-05,
+      "loss": 0.4906,
+      "step": 7190
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6471436398765756e-05,
+      "loss": 0.454,
+      "step": 7191
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6470859647605043e-05,
+      "loss": 0.5713,
+      "step": 7192
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.647028289644433e-05,
+      "loss": 0.4966,
+      "step": 7193
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.646970614528362e-05,
+      "loss": 0.6324,
+      "step": 7194
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6469129394122906e-05,
+      "loss": 0.624,
+      "step": 7195
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6468552642962194e-05,
+      "loss": 0.4336,
+      "step": 7196
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6467975891801485e-05,
+      "loss": 0.615,
+      "step": 7197
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6467399140640773e-05,
+      "loss": 0.6509,
+      "step": 7198
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.646682238948006e-05,
+      "loss": 0.5089,
+      "step": 7199
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6466245638319348e-05,
+      "loss": 0.6102,
+      "step": 7200
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6465668887158636e-05,
+      "loss": 0.5331,
+      "step": 7201
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6465092135997923e-05,
+      "loss": 0.5827,
+      "step": 7202
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6464515384837215e-05,
+      "loss": 0.5574,
+      "step": 7203
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6463938633676502e-05,
+      "loss": 0.5386,
+      "step": 7204
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.646336188251579e-05,
+      "loss": 0.4517,
+      "step": 7205
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6462785131355078e-05,
+      "loss": 0.5631,
+      "step": 7206
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6462208380194365e-05,
+      "loss": 0.4803,
+      "step": 7207
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6461631629033653e-05,
+      "loss": 0.5925,
+      "step": 7208
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6461054877872944e-05,
+      "loss": 0.5638,
+      "step": 7209
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6460478126712232e-05,
+      "loss": 0.4793,
+      "step": 7210
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645990137555152e-05,
+      "loss": 0.5805,
+      "step": 7211
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645932462439081e-05,
+      "loss": 0.6679,
+      "step": 7212
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6458747873230098e-05,
+      "loss": 0.4583,
+      "step": 7213
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6458171122069382e-05,
+      "loss": 0.5109,
+      "step": 7214
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6457594370908674e-05,
+      "loss": 0.4173,
+      "step": 7215
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645701761974796e-05,
+      "loss": 0.4972,
+      "step": 7216
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645644086858725e-05,
+      "loss": 0.6521,
+      "step": 7217
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645586411742654e-05,
+      "loss": 0.486,
+      "step": 7218
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6455287366265828e-05,
+      "loss": 0.4609,
+      "step": 7219
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6454710615105115e-05,
+      "loss": 0.5491,
+      "step": 7220
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6454133863944403e-05,
+      "loss": 0.4811,
+      "step": 7221
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645355711278369e-05,
+      "loss": 0.5649,
+      "step": 7222
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645298036162298e-05,
+      "loss": 0.4626,
+      "step": 7223
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6452403610462266e-05,
+      "loss": 0.5054,
+      "step": 7224
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6451826859301557e-05,
+      "loss": 0.6119,
+      "step": 7225
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6451250108140845e-05,
+      "loss": 0.5819,
+      "step": 7226
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6450673356980133e-05,
+      "loss": 0.6163,
+      "step": 7227
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.645009660581942e-05,
+      "loss": 0.564,
+      "step": 7228
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6449519854658708e-05,
+      "loss": 0.6411,
+      "step": 7229
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6448943103497996e-05,
+      "loss": 0.4809,
+      "step": 7230
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6448366352337287e-05,
+      "loss": 0.574,
+      "step": 7231
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6447789601176574e-05,
+      "loss": 0.4554,
+      "step": 7232
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6447212850015862e-05,
+      "loss": 0.3394,
+      "step": 7233
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.644663609885515e-05,
+      "loss": 0.5834,
+      "step": 7234
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6446059347694437e-05,
+      "loss": 0.581,
+      "step": 7235
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6445482596533725e-05,
+      "loss": 0.6129,
+      "step": 7236
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6444905845373016e-05,
+      "loss": 0.439,
+      "step": 7237
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.6444329094212304e-05,
+      "loss": 0.6017,
+      "step": 7238
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.644375234305159e-05,
+      "loss": 0.4631,
+      "step": 7239
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.644317559189088e-05,
+      "loss": 0.5275,
+      "step": 7240
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6442598840730167e-05,
+      "loss": 0.6107,
+      "step": 7241
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6442022089569455e-05,
+      "loss": 0.6251,
+      "step": 7242
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6441445338408746e-05,
+      "loss": 0.5366,
+      "step": 7243
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6440868587248033e-05,
+      "loss": 0.5247,
+      "step": 7244
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.644029183608732e-05,
+      "loss": 0.4755,
+      "step": 7245
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6439715084926612e-05,
+      "loss": 0.5996,
+      "step": 7246
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.64391383337659e-05,
+      "loss": 0.4862,
+      "step": 7247
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6438561582605187e-05,
+      "loss": 0.633,
+      "step": 7248
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6437984831444475e-05,
+      "loss": 0.3995,
+      "step": 7249
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6437408080283763e-05,
+      "loss": 0.5182,
+      "step": 7250
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.643683132912305e-05,
+      "loss": 0.5708,
+      "step": 7251
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.643625457796234e-05,
+      "loss": 0.4683,
+      "step": 7252
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.643567782680163e-05,
+      "loss": 0.5663,
+      "step": 7253
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6435101075640917e-05,
+      "loss": 0.4934,
+      "step": 7254
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6434524324480205e-05,
+      "loss": 0.5988,
+      "step": 7255
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6433947573319492e-05,
+      "loss": 0.4519,
+      "step": 7256
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.643337082215878e-05,
+      "loss": 0.5041,
+      "step": 7257
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6432794070998068e-05,
+      "loss": 0.6059,
+      "step": 7258
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.643221731983736e-05,
+      "loss": 0.5336,
+      "step": 7259
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6431640568676646e-05,
+      "loss": 0.5669,
+      "step": 7260
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6431063817515934e-05,
+      "loss": 0.589,
+      "step": 7261
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6430487066355222e-05,
+      "loss": 0.3572,
+      "step": 7262
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.642991031519451e-05,
+      "loss": 0.5559,
+      "step": 7263
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6429333564033797e-05,
+      "loss": 0.5207,
+      "step": 7264
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6428756812873088e-05,
+      "loss": 0.5875,
+      "step": 7265
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6428180061712376e-05,
+      "loss": 0.4772,
+      "step": 7266
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6427603310551664e-05,
+      "loss": 0.633,
+      "step": 7267
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.642702655939095e-05,
+      "loss": 0.5771,
+      "step": 7268
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.642644980823024e-05,
+      "loss": 0.5815,
+      "step": 7269
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6425873057069527e-05,
+      "loss": 0.4065,
+      "step": 7270
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6425296305908818e-05,
+      "loss": 0.5727,
+      "step": 7271
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6424719554748105e-05,
+      "loss": 0.5344,
+      "step": 7272
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6424142803587393e-05,
+      "loss": 0.4843,
+      "step": 7273
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6423566052426684e-05,
+      "loss": 0.6111,
+      "step": 7274
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6422989301265972e-05,
+      "loss": 0.5796,
+      "step": 7275
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6422412550105256e-05,
+      "loss": 0.581,
+      "step": 7276
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6421835798944547e-05,
+      "loss": 0.5124,
+      "step": 7277
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6421259047783835e-05,
+      "loss": 0.6119,
+      "step": 7278
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6420682296623123e-05,
+      "loss": 0.5746,
+      "step": 7279
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6420105545462414e-05,
+      "loss": 0.5025,
+      "step": 7280
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.64195287943017e-05,
+      "loss": 0.4596,
+      "step": 7281
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641895204314099e-05,
+      "loss": 0.4527,
+      "step": 7282
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6418375291980277e-05,
+      "loss": 0.5425,
+      "step": 7283
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6417798540819564e-05,
+      "loss": 0.4927,
+      "step": 7284
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6417221789658852e-05,
+      "loss": 0.627,
+      "step": 7285
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6416645038498143e-05,
+      "loss": 0.4944,
+      "step": 7286
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641606828733743e-05,
+      "loss": 0.596,
+      "step": 7287
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641549153617672e-05,
+      "loss": 0.61,
+      "step": 7288
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6414914785016006e-05,
+      "loss": 0.6066,
+      "step": 7289
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6414338033855294e-05,
+      "loss": 0.6293,
+      "step": 7290
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641376128269458e-05,
+      "loss": 0.4368,
+      "step": 7291
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641318453153387e-05,
+      "loss": 0.5729,
+      "step": 7292
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641260778037316e-05,
+      "loss": 0.6525,
+      "step": 7293
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6412031029212448e-05,
+      "loss": 0.602,
+      "step": 7294
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6411454278051736e-05,
+      "loss": 0.4063,
+      "step": 7295
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6410877526891023e-05,
+      "loss": 0.6163,
+      "step": 7296
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.641030077573031e-05,
+      "loss": 0.5703,
+      "step": 7297
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.64097240245696e-05,
+      "loss": 0.5165,
+      "step": 7298
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640914727340889e-05,
+      "loss": 0.5134,
+      "step": 7299
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6408570522248178e-05,
+      "loss": 0.4521,
+      "step": 7300
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6407993771087465e-05,
+      "loss": 0.6918,
+      "step": 7301
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6407417019926753e-05,
+      "loss": 0.5273,
+      "step": 7302
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640684026876604e-05,
+      "loss": 0.5275,
+      "step": 7303
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640626351760533e-05,
+      "loss": 0.5659,
+      "step": 7304
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640568676644462e-05,
+      "loss": 0.596,
+      "step": 7305
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6405110015283907e-05,
+      "loss": 0.5645,
+      "step": 7306
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6404533264123195e-05,
+      "loss": 0.4997,
+      "step": 7307
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6403956512962486e-05,
+      "loss": 0.5104,
+      "step": 7308
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6403379761801774e-05,
+      "loss": 0.52,
+      "step": 7309
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640280301064106e-05,
+      "loss": 0.6379,
+      "step": 7310
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.640222625948035e-05,
+      "loss": 0.4755,
+      "step": 7311
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6401649508319637e-05,
+      "loss": 0.5554,
+      "step": 7312
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6401072757158924e-05,
+      "loss": 0.5341,
+      "step": 7313
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6400496005998215e-05,
+      "loss": 0.5696,
+      "step": 7314
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6399919254837503e-05,
+      "loss": 0.6098,
+      "step": 7315
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.639934250367679e-05,
+      "loss": 0.5732,
+      "step": 7316
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.639876575251608e-05,
+      "loss": 0.4776,
+      "step": 7317
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6398189001355366e-05,
+      "loss": 0.4461,
+      "step": 7318
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6397612250194654e-05,
+      "loss": 0.5899,
+      "step": 7319
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6397035499033945e-05,
+      "loss": 0.5845,
+      "step": 7320
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6396458747873233e-05,
+      "loss": 0.6183,
+      "step": 7321
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.639588199671252e-05,
+      "loss": 0.4946,
+      "step": 7322
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6395305245551808e-05,
+      "loss": 0.6303,
+      "step": 7323
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6394728494391096e-05,
+      "loss": 0.5313,
+      "step": 7324
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6394151743230383e-05,
+      "loss": 0.4995,
+      "step": 7325
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.639357499206967e-05,
+      "loss": 0.4742,
+      "step": 7326
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6392998240908962e-05,
+      "loss": 0.5258,
+      "step": 7327
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.639242148974825e-05,
+      "loss": 0.5684,
+      "step": 7328
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6391844738587537e-05,
+      "loss": 0.5858,
+      "step": 7329
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6391267987426825e-05,
+      "loss": 0.4084,
+      "step": 7330
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6390691236266113e-05,
+      "loss": 0.425,
+      "step": 7331
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.63901144851054e-05,
+      "loss": 0.5528,
+      "step": 7332
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638953773394469e-05,
+      "loss": 0.5732,
+      "step": 7333
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638896098278398e-05,
+      "loss": 0.6769,
+      "step": 7334
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6388384231623267e-05,
+      "loss": 0.4978,
+      "step": 7335
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6387807480462558e-05,
+      "loss": 0.6855,
+      "step": 7336
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6387230729301846e-05,
+      "loss": 0.504,
+      "step": 7337
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638665397814113e-05,
+      "loss": 0.5526,
+      "step": 7338
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638607722698042e-05,
+      "loss": 0.4723,
+      "step": 7339
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638550047581971e-05,
+      "loss": 0.4354,
+      "step": 7340
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6384923724658996e-05,
+      "loss": 0.5803,
+      "step": 7341
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6384346973498287e-05,
+      "loss": 0.5062,
+      "step": 7342
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6383770222337575e-05,
+      "loss": 0.6927,
+      "step": 7343
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6383193471176863e-05,
+      "loss": 0.5493,
+      "step": 7344
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.638261672001615e-05,
+      "loss": 0.558,
+      "step": 7345
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6382039968855438e-05,
+      "loss": 0.5921,
+      "step": 7346
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6381463217694726e-05,
+      "loss": 0.551,
+      "step": 7347
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6380886466534017e-05,
+      "loss": 0.5103,
+      "step": 7348
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6380309715373305e-05,
+      "loss": 0.4326,
+      "step": 7349
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6379732964212592e-05,
+      "loss": 0.5918,
+      "step": 7350
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.637915621305188e-05,
+      "loss": 0.5025,
+      "step": 7351
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6378579461891168e-05,
+      "loss": 0.5252,
+      "step": 7352
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6378002710730455e-05,
+      "loss": 0.5801,
+      "step": 7353
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6377425959569746e-05,
+      "loss": 0.4122,
+      "step": 7354
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6376849208409034e-05,
+      "loss": 0.5746,
+      "step": 7355
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6376272457248322e-05,
+      "loss": 0.6121,
+      "step": 7356
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.637569570608761e-05,
+      "loss": 0.5898,
+      "step": 7357
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6375118954926897e-05,
+      "loss": 0.7057,
+      "step": 7358
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6374542203766185e-05,
+      "loss": 0.59,
+      "step": 7359
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6373965452605473e-05,
+      "loss": 0.5326,
+      "step": 7360
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6373388701444764e-05,
+      "loss": 0.7656,
+      "step": 7361
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.637281195028405e-05,
+      "loss": 0.4333,
+      "step": 7362
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.637223519912334e-05,
+      "loss": 0.4337,
+      "step": 7363
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6371658447962627e-05,
+      "loss": 0.5167,
+      "step": 7364
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6371081696801914e-05,
+      "loss": 0.3848,
+      "step": 7365
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6370504945641202e-05,
+      "loss": 0.5753,
+      "step": 7366
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6369928194480493e-05,
+      "loss": 0.5909,
+      "step": 7367
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636935144331978e-05,
+      "loss": 0.5186,
+      "step": 7368
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636877469215907e-05,
+      "loss": 0.7277,
+      "step": 7369
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636819794099836e-05,
+      "loss": 0.5239,
+      "step": 7370
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6367621189837647e-05,
+      "loss": 0.5988,
+      "step": 7371
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6367044438676935e-05,
+      "loss": 0.5666,
+      "step": 7372
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6366467687516223e-05,
+      "loss": 0.66,
+      "step": 7373
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636589093635551e-05,
+      "loss": 0.6277,
+      "step": 7374
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6365314185194798e-05,
+      "loss": 0.5867,
+      "step": 7375
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636473743403409e-05,
+      "loss": 0.6401,
+      "step": 7376
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6364160682873377e-05,
+      "loss": 0.4647,
+      "step": 7377
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6363583931712664e-05,
+      "loss": 0.6538,
+      "step": 7378
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6363007180551952e-05,
+      "loss": 0.5306,
+      "step": 7379
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636243042939124e-05,
+      "loss": 0.5707,
+      "step": 7380
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6361853678230528e-05,
+      "loss": 0.5525,
+      "step": 7381
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.636127692706982e-05,
+      "loss": 0.5035,
+      "step": 7382
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6360700175909106e-05,
+      "loss": 0.4705,
+      "step": 7383
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6360123424748394e-05,
+      "loss": 0.617,
+      "step": 7384
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.635954667358768e-05,
+      "loss": 0.4522,
+      "step": 7385
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.635896992242697e-05,
+      "loss": 0.4583,
+      "step": 7386
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6358393171266257e-05,
+      "loss": 0.5221,
+      "step": 7387
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6357816420105548e-05,
+      "loss": 0.5004,
+      "step": 7388
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6357239668944836e-05,
+      "loss": 0.5842,
+      "step": 7389
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6356662917784123e-05,
+      "loss": 0.552,
+      "step": 7390
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.635608616662341e-05,
+      "loss": 0.5134,
+      "step": 7391
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.63555094154627e-05,
+      "loss": 0.5856,
+      "step": 7392
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6354932664301987e-05,
+      "loss": 0.4957,
+      "step": 7393
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6354355913141274e-05,
+      "loss": 0.6062,
+      "step": 7394
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6353779161980565e-05,
+      "loss": 0.54,
+      "step": 7395
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6353202410819853e-05,
+      "loss": 0.6538,
+      "step": 7396
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.635262565965914e-05,
+      "loss": 0.6085,
+      "step": 7397
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6352048908498432e-05,
+      "loss": 0.6577,
+      "step": 7398
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.635147215733772e-05,
+      "loss": 0.4933,
+      "step": 7399
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6350895406177004e-05,
+      "loss": 0.6364,
+      "step": 7400
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6350318655016295e-05,
+      "loss": 0.6228,
+      "step": 7401
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6349741903855582e-05,
+      "loss": 0.6979,
+      "step": 7402
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634916515269487e-05,
+      "loss": 0.3984,
+      "step": 7403
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634858840153416e-05,
+      "loss": 0.5456,
+      "step": 7404
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634801165037345e-05,
+      "loss": 0.5529,
+      "step": 7405
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6347434899212737e-05,
+      "loss": 0.5845,
+      "step": 7406
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6346858148052024e-05,
+      "loss": 0.6491,
+      "step": 7407
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6346281396891312e-05,
+      "loss": 0.71,
+      "step": 7408
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.63457046457306e-05,
+      "loss": 0.5574,
+      "step": 7409
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634512789456989e-05,
+      "loss": 0.6131,
+      "step": 7410
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634455114340918e-05,
+      "loss": 0.5869,
+      "step": 7411
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6343974392248466e-05,
+      "loss": 0.6008,
+      "step": 7412
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6343397641087754e-05,
+      "loss": 0.5926,
+      "step": 7413
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634282088992704e-05,
+      "loss": 0.5841,
+      "step": 7414
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634224413876633e-05,
+      "loss": 0.5401,
+      "step": 7415
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.634166738760562e-05,
+      "loss": 0.6557,
+      "step": 7416
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6341090636444908e-05,
+      "loss": 0.5716,
+      "step": 7417
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.6340513885284196e-05,
+      "loss": 0.4878,
+      "step": 7418
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6339937134123483e-05,
+      "loss": 0.5126,
+      "step": 7419
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.633936038296277e-05,
+      "loss": 0.3601,
+      "step": 7420
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.633878363180206e-05,
+      "loss": 0.5602,
+      "step": 7421
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.633820688064135e-05,
+      "loss": 0.619,
+      "step": 7422
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6337630129480637e-05,
+      "loss": 0.4724,
+      "step": 7423
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6337053378319925e-05,
+      "loss": 0.4746,
+      "step": 7424
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6336476627159216e-05,
+      "loss": 0.5865,
+      "step": 7425
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.63358998759985e-05,
+      "loss": 0.4802,
+      "step": 7426
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6335323124837788e-05,
+      "loss": 0.5,
+      "step": 7427
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6334746373677076e-05,
+      "loss": 0.5282,
+      "step": 7428
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6334169622516367e-05,
+      "loss": 0.5911,
+      "step": 7429
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6333592871355655e-05,
+      "loss": 0.5793,
+      "step": 7430
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6333016120194942e-05,
+      "loss": 0.5539,
+      "step": 7431
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6332439369034233e-05,
+      "loss": 0.77,
+      "step": 7432
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.633186261787352e-05,
+      "loss": 0.5969,
+      "step": 7433
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.633128586671281e-05,
+      "loss": 0.5127,
+      "step": 7434
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6330709115552096e-05,
+      "loss": 0.4094,
+      "step": 7435
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6330132364391384e-05,
+      "loss": 0.622,
+      "step": 7436
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6329555613230672e-05,
+      "loss": 0.5379,
+      "step": 7437
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6328978862069963e-05,
+      "loss": 0.4779,
+      "step": 7438
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.632840211090925e-05,
+      "loss": 0.5241,
+      "step": 7439
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6327825359748538e-05,
+      "loss": 0.4807,
+      "step": 7440
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6327248608587826e-05,
+      "loss": 0.5483,
+      "step": 7441
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6326671857427114e-05,
+      "loss": 0.637,
+      "step": 7442
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.63260951062664e-05,
+      "loss": 0.501,
+      "step": 7443
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6325518355105692e-05,
+      "loss": 0.5614,
+      "step": 7444
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.632494160394498e-05,
+      "loss": 0.4986,
+      "step": 7445
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6324364852784268e-05,
+      "loss": 0.4882,
+      "step": 7446
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6323788101623555e-05,
+      "loss": 0.4606,
+      "step": 7447
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6323211350462843e-05,
+      "loss": 0.6559,
+      "step": 7448
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.632263459930213e-05,
+      "loss": 0.4712,
+      "step": 7449
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6322057848141422e-05,
+      "loss": 0.5049,
+      "step": 7450
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.632148109698071e-05,
+      "loss": 0.6599,
+      "step": 7451
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6320904345819997e-05,
+      "loss": 0.6126,
+      "step": 7452
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6320327594659285e-05,
+      "loss": 0.58,
+      "step": 7453
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6319750843498573e-05,
+      "loss": 0.6062,
+      "step": 7454
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.631917409233786e-05,
+      "loss": 0.5541,
+      "step": 7455
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.631859734117715e-05,
+      "loss": 0.4505,
+      "step": 7456
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.631802059001644e-05,
+      "loss": 0.4325,
+      "step": 7457
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6317443838855727e-05,
+      "loss": 0.598,
+      "step": 7458
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6316867087695018e-05,
+      "loss": 0.6108,
+      "step": 7459
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6316290336534305e-05,
+      "loss": 0.5568,
+      "step": 7460
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6315713585373593e-05,
+      "loss": 0.6299,
+      "step": 7461
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6315136834212877e-05,
+      "loss": 0.5443,
+      "step": 7462
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.631456008305217e-05,
+      "loss": 0.4973,
+      "step": 7463
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6313983331891456e-05,
+      "loss": 0.6984,
+      "step": 7464
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6313406580730744e-05,
+      "loss": 0.5418,
+      "step": 7465
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6312829829570035e-05,
+      "loss": 0.5898,
+      "step": 7466
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6312253078409323e-05,
+      "loss": 0.4924,
+      "step": 7467
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.631167632724861e-05,
+      "loss": 0.4998,
+      "step": 7468
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6311099576087898e-05,
+      "loss": 0.5609,
+      "step": 7469
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6310522824927186e-05,
+      "loss": 0.6597,
+      "step": 7470
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6309946073766473e-05,
+      "loss": 0.5976,
+      "step": 7471
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6309369322605764e-05,
+      "loss": 0.4523,
+      "step": 7472
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6308792571445052e-05,
+      "loss": 0.4997,
+      "step": 7473
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.630821582028434e-05,
+      "loss": 0.5029,
+      "step": 7474
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6307639069123628e-05,
+      "loss": 0.4877,
+      "step": 7475
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6307062317962915e-05,
+      "loss": 0.5685,
+      "step": 7476
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6306485566802203e-05,
+      "loss": 0.5323,
+      "step": 7477
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6305908815641494e-05,
+      "loss": 0.5401,
+      "step": 7478
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.630533206448078e-05,
+      "loss": 0.5473,
+      "step": 7479
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.630475531332007e-05,
+      "loss": 0.5235,
+      "step": 7480
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6304178562159357e-05,
+      "loss": 0.5945,
+      "step": 7481
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6303601810998645e-05,
+      "loss": 0.5129,
+      "step": 7482
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6303025059837932e-05,
+      "loss": 0.4966,
+      "step": 7483
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6302448308677223e-05,
+      "loss": 0.5082,
+      "step": 7484
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.630187155751651e-05,
+      "loss": 0.5459,
+      "step": 7485
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.63012948063558e-05,
+      "loss": 0.5512,
+      "step": 7486
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.630071805519509e-05,
+      "loss": 0.5426,
+      "step": 7487
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6300141304034374e-05,
+      "loss": 0.5863,
+      "step": 7488
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6299564552873662e-05,
+      "loss": 0.5446,
+      "step": 7489
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6298987801712953e-05,
+      "loss": 0.5009,
+      "step": 7490
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.629841105055224e-05,
+      "loss": 0.5056,
+      "step": 7491
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.629783429939153e-05,
+      "loss": 0.6275,
+      "step": 7492
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.629725754823082e-05,
+      "loss": 0.5708,
+      "step": 7493
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6296680797070107e-05,
+      "loss": 0.4756,
+      "step": 7494
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6296104045909395e-05,
+      "loss": 0.589,
+      "step": 7495
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6295527294748682e-05,
+      "loss": 0.4216,
+      "step": 7496
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.629495054358797e-05,
+      "loss": 0.5032,
+      "step": 7497
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6294373792427258e-05,
+      "loss": 0.4564,
+      "step": 7498
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6293797041266546e-05,
+      "loss": 0.5533,
+      "step": 7499
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6293220290105837e-05,
+      "loss": 0.4865,
+      "step": 7500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6292643538945124e-05,
+      "loss": 0.5723,
+      "step": 7501
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6292066787784412e-05,
+      "loss": 0.7035,
+      "step": 7502
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.62914900366237e-05,
+      "loss": 0.532,
+      "step": 7503
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6290913285462987e-05,
+      "loss": 0.6358,
+      "step": 7504
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6290336534302275e-05,
+      "loss": 0.5833,
+      "step": 7505
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6289759783141566e-05,
+      "loss": 0.5346,
+      "step": 7506
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6289183031980854e-05,
+      "loss": 0.459,
+      "step": 7507
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628860628082014e-05,
+      "loss": 0.5892,
+      "step": 7508
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628802952965943e-05,
+      "loss": 0.4704,
+      "step": 7509
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6287452778498717e-05,
+      "loss": 0.5326,
+      "step": 7510
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6286876027338005e-05,
+      "loss": 0.4977,
+      "step": 7511
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6286299276177296e-05,
+      "loss": 0.5479,
+      "step": 7512
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6285722525016583e-05,
+      "loss": 0.6485,
+      "step": 7513
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628514577385587e-05,
+      "loss": 0.4883,
+      "step": 7514
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628456902269516e-05,
+      "loss": 0.6313,
+      "step": 7515
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6283992271534446e-05,
+      "loss": 0.4952,
+      "step": 7516
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6283415520373734e-05,
+      "loss": 0.5422,
+      "step": 7517
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6282838769213025e-05,
+      "loss": 0.635,
+      "step": 7518
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6282262018052313e-05,
+      "loss": 0.4165,
+      "step": 7519
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.62816852668916e-05,
+      "loss": 0.652,
+      "step": 7520
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628110851573089e-05,
+      "loss": 0.5014,
+      "step": 7521
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.628053176457018e-05,
+      "loss": 0.5208,
+      "step": 7522
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6279955013409467e-05,
+      "loss": 0.5129,
+      "step": 7523
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6279378262248755e-05,
+      "loss": 0.6383,
+      "step": 7524
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6278801511088042e-05,
+      "loss": 0.569,
+      "step": 7525
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.627822475992733e-05,
+      "loss": 0.5895,
+      "step": 7526
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.627764800876662e-05,
+      "loss": 0.5036,
+      "step": 7527
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.627707125760591e-05,
+      "loss": 0.5858,
+      "step": 7528
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6276494506445196e-05,
+      "loss": 0.4778,
+      "step": 7529
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6275917755284484e-05,
+      "loss": 0.5724,
+      "step": 7530
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6275341004123772e-05,
+      "loss": 0.4996,
+      "step": 7531
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.627476425296306e-05,
+      "loss": 0.6344,
+      "step": 7532
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6274187501802347e-05,
+      "loss": 0.5378,
+      "step": 7533
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6273610750641638e-05,
+      "loss": 0.4688,
+      "step": 7534
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6273033999480926e-05,
+      "loss": 0.5502,
+      "step": 7535
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6272457248320214e-05,
+      "loss": 0.4419,
+      "step": 7536
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.62718804971595e-05,
+      "loss": 0.5004,
+      "step": 7537
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.627130374599879e-05,
+      "loss": 0.5377,
+      "step": 7538
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6270726994838077e-05,
+      "loss": 0.5181,
+      "step": 7539
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6270150243677368e-05,
+      "loss": 0.5885,
+      "step": 7540
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6269573492516655e-05,
+      "loss": 0.568,
+      "step": 7541
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6268996741355943e-05,
+      "loss": 0.6241,
+      "step": 7542
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.626841999019523e-05,
+      "loss": 0.5993,
+      "step": 7543
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.626784323903452e-05,
+      "loss": 0.6145,
+      "step": 7544
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6267266487873806e-05,
+      "loss": 0.4684,
+      "step": 7545
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6266689736713097e-05,
+      "loss": 0.6086,
+      "step": 7546
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6266112985552385e-05,
+      "loss": 0.6013,
+      "step": 7547
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6265536234391673e-05,
+      "loss": 0.61,
+      "step": 7548
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6264959483230964e-05,
+      "loss": 0.5948,
+      "step": 7549
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6264382732070248e-05,
+      "loss": 0.6055,
+      "step": 7550
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6263805980909536e-05,
+      "loss": 0.5844,
+      "step": 7551
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6263229229748827e-05,
+      "loss": 0.673,
+      "step": 7552
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6262652478588114e-05,
+      "loss": 0.382,
+      "step": 7553
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6262075727427402e-05,
+      "loss": 0.6755,
+      "step": 7554
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6261498976266693e-05,
+      "loss": 0.6484,
+      "step": 7555
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.626092222510598e-05,
+      "loss": 0.5705,
+      "step": 7556
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.626034547394527e-05,
+      "loss": 0.5195,
+      "step": 7557
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6259768722784556e-05,
+      "loss": 0.6532,
+      "step": 7558
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6259191971623844e-05,
+      "loss": 0.5916,
+      "step": 7559
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625861522046313e-05,
+      "loss": 0.6054,
+      "step": 7560
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6258038469302423e-05,
+      "loss": 0.5215,
+      "step": 7561
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625746171814171e-05,
+      "loss": 0.4412,
+      "step": 7562
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6256884966980998e-05,
+      "loss": 0.5585,
+      "step": 7563
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6256308215820286e-05,
+      "loss": 0.4633,
+      "step": 7564
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6255731464659573e-05,
+      "loss": 0.5496,
+      "step": 7565
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625515471349886e-05,
+      "loss": 0.5427,
+      "step": 7566
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625457796233815e-05,
+      "loss": 0.6958,
+      "step": 7567
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625400121117744e-05,
+      "loss": 0.4504,
+      "step": 7568
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6253424460016728e-05,
+      "loss": 0.5026,
+      "step": 7569
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6252847708856015e-05,
+      "loss": 0.4873,
+      "step": 7570
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6252270957695303e-05,
+      "loss": 0.4636,
+      "step": 7571
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625169420653459e-05,
+      "loss": 0.4762,
+      "step": 7572
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6251117455373878e-05,
+      "loss": 0.5958,
+      "step": 7573
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.625054070421317e-05,
+      "loss": 0.6066,
+      "step": 7574
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6249963953052457e-05,
+      "loss": 0.4884,
+      "step": 7575
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6249387201891745e-05,
+      "loss": 0.6643,
+      "step": 7576
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6248810450731032e-05,
+      "loss": 0.55,
+      "step": 7577
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.624823369957032e-05,
+      "loss": 0.6185,
+      "step": 7578
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6247656948409608e-05,
+      "loss": 0.6183,
+      "step": 7579
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.62470801972489e-05,
+      "loss": 0.5249,
+      "step": 7580
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6246503446088187e-05,
+      "loss": 0.5453,
+      "step": 7581
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6245926694927474e-05,
+      "loss": 0.5252,
+      "step": 7582
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6245349943766765e-05,
+      "loss": 0.6103,
+      "step": 7583
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6244773192606053e-05,
+      "loss": 0.5948,
+      "step": 7584
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.624419644144534e-05,
+      "loss": 0.625,
+      "step": 7585
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.624361969028463e-05,
+      "loss": 0.6385,
+      "step": 7586
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6243042939123916e-05,
+      "loss": 0.4871,
+      "step": 7587
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6242466187963204e-05,
+      "loss": 0.4405,
+      "step": 7588
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6241889436802495e-05,
+      "loss": 0.4881,
+      "step": 7589
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6241312685641782e-05,
+      "loss": 0.4985,
+      "step": 7590
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.624073593448107e-05,
+      "loss": 0.6625,
+      "step": 7591
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6240159183320358e-05,
+      "loss": 0.5727,
+      "step": 7592
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6239582432159646e-05,
+      "loss": 0.5234,
+      "step": 7593
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6239005680998933e-05,
+      "loss": 0.6506,
+      "step": 7594
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6238428929838224e-05,
+      "loss": 0.4593,
+      "step": 7595
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.6237852178677512e-05,
+      "loss": 0.5432,
+      "step": 7596
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.62372754275168e-05,
+      "loss": 0.469,
+      "step": 7597
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6236698676356087e-05,
+      "loss": 0.5839,
+      "step": 7598
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6236121925195375e-05,
+      "loss": 0.4931,
+      "step": 7599
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6235545174034663e-05,
+      "loss": 0.5128,
+      "step": 7600
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6234968422873954e-05,
+      "loss": 0.4518,
+      "step": 7601
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.623439167171324e-05,
+      "loss": 0.5205,
+      "step": 7602
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.623381492055253e-05,
+      "loss": 0.5865,
+      "step": 7603
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6233238169391817e-05,
+      "loss": 0.5141,
+      "step": 7604
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6232661418231105e-05,
+      "loss": 0.603,
+      "step": 7605
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6232084667070392e-05,
+      "loss": 0.6306,
+      "step": 7606
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.623150791590968e-05,
+      "loss": 0.5401,
+      "step": 7607
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.623093116474897e-05,
+      "loss": 0.5898,
+      "step": 7608
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.623035441358826e-05,
+      "loss": 0.5493,
+      "step": 7609
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6229777662427546e-05,
+      "loss": 0.5599,
+      "step": 7610
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6229200911266837e-05,
+      "loss": 0.5864,
+      "step": 7611
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6228624160106125e-05,
+      "loss": 0.5004,
+      "step": 7612
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.622804740894541e-05,
+      "loss": 0.5463,
+      "step": 7613
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.62274706577847e-05,
+      "loss": 0.5731,
+      "step": 7614
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6226893906623988e-05,
+      "loss": 0.5236,
+      "step": 7615
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6226317155463276e-05,
+      "loss": 0.5157,
+      "step": 7616
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6225740404302567e-05,
+      "loss": 0.4185,
+      "step": 7617
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6225163653141855e-05,
+      "loss": 0.4184,
+      "step": 7618
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6224586901981142e-05,
+      "loss": 0.6614,
+      "step": 7619
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.622401015082043e-05,
+      "loss": 0.5216,
+      "step": 7620
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6223433399659718e-05,
+      "loss": 0.5971,
+      "step": 7621
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6222856648499005e-05,
+      "loss": 0.4193,
+      "step": 7622
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6222279897338296e-05,
+      "loss": 0.4833,
+      "step": 7623
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6221703146177584e-05,
+      "loss": 0.3993,
+      "step": 7624
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6221126395016872e-05,
+      "loss": 0.4535,
+      "step": 7625
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.622054964385616e-05,
+      "loss": 0.5509,
+      "step": 7626
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6219972892695447e-05,
+      "loss": 0.5234,
+      "step": 7627
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6219396141534735e-05,
+      "loss": 0.4385,
+      "step": 7628
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6218819390374026e-05,
+      "loss": 0.5002,
+      "step": 7629
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6218242639213314e-05,
+      "loss": 0.6279,
+      "step": 7630
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.62176658880526e-05,
+      "loss": 0.5009,
+      "step": 7631
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.621708913689189e-05,
+      "loss": 0.4828,
+      "step": 7632
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6216512385731177e-05,
+      "loss": 0.5019,
+      "step": 7633
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6215935634570464e-05,
+      "loss": 0.5103,
+      "step": 7634
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6215358883409755e-05,
+      "loss": 0.4808,
+      "step": 7635
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6214782132249043e-05,
+      "loss": 0.5311,
+      "step": 7636
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.621420538108833e-05,
+      "loss": 0.503,
+      "step": 7637
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.621362862992762e-05,
+      "loss": 0.5293,
+      "step": 7638
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6213051878766906e-05,
+      "loss": 0.473,
+      "step": 7639
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6212475127606194e-05,
+      "loss": 0.4448,
+      "step": 7640
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.621189837644548e-05,
+      "loss": 0.5035,
+      "step": 7641
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6211321625284773e-05,
+      "loss": 0.6303,
+      "step": 7642
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.621074487412406e-05,
+      "loss": 0.5368,
+      "step": 7643
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6210168122963348e-05,
+      "loss": 0.7058,
+      "step": 7644
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620959137180264e-05,
+      "loss": 0.5821,
+      "step": 7645
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6209014620641927e-05,
+      "loss": 0.4327,
+      "step": 7646
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6208437869481214e-05,
+      "loss": 0.6247,
+      "step": 7647
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6207861118320502e-05,
+      "loss": 0.3686,
+      "step": 7648
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620728436715979e-05,
+      "loss": 0.6578,
+      "step": 7649
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6206707615999077e-05,
+      "loss": 0.517,
+      "step": 7650
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620613086483837e-05,
+      "loss": 0.4925,
+      "step": 7651
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6205554113677656e-05,
+      "loss": 0.5778,
+      "step": 7652
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6204977362516944e-05,
+      "loss": 0.487,
+      "step": 7653
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620440061135623e-05,
+      "loss": 0.5654,
+      "step": 7654
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620382386019552e-05,
+      "loss": 0.5891,
+      "step": 7655
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6203247109034807e-05,
+      "loss": 0.5325,
+      "step": 7656
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6202670357874098e-05,
+      "loss": 0.5447,
+      "step": 7657
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6202093606713386e-05,
+      "loss": 0.5641,
+      "step": 7658
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6201516855552673e-05,
+      "loss": 0.5773,
+      "step": 7659
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620094010439196e-05,
+      "loss": 0.4101,
+      "step": 7660
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.620036335323125e-05,
+      "loss": 0.6692,
+      "step": 7661
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6199786602070536e-05,
+      "loss": 0.4774,
+      "step": 7662
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6199209850909828e-05,
+      "loss": 0.5355,
+      "step": 7663
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6198633099749115e-05,
+      "loss": 0.5959,
+      "step": 7664
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6198056348588403e-05,
+      "loss": 0.4101,
+      "step": 7665
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.619747959742769e-05,
+      "loss": 0.6128,
+      "step": 7666
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6196902846266978e-05,
+      "loss": 0.5846,
+      "step": 7667
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6196326095106266e-05,
+      "loss": 0.4734,
+      "step": 7668
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6195749343945557e-05,
+      "loss": 0.5721,
+      "step": 7669
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6195172592784845e-05,
+      "loss": 0.6183,
+      "step": 7670
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6194595841624132e-05,
+      "loss": 0.4979,
+      "step": 7671
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.619401909046342e-05,
+      "loss": 0.5474,
+      "step": 7672
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.619344233930271e-05,
+      "loss": 0.6569,
+      "step": 7673
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6192865588142e-05,
+      "loss": 0.5098,
+      "step": 7674
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6192288836981283e-05,
+      "loss": 0.6255,
+      "step": 7675
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6191712085820574e-05,
+      "loss": 0.4622,
+      "step": 7676
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6191135334659862e-05,
+      "loss": 0.5662,
+      "step": 7677
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.619055858349915e-05,
+      "loss": 0.5447,
+      "step": 7678
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618998183233844e-05,
+      "loss": 0.4801,
+      "step": 7679
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618940508117773e-05,
+      "loss": 0.5456,
+      "step": 7680
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6188828330017016e-05,
+      "loss": 0.6626,
+      "step": 7681
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6188251578856304e-05,
+      "loss": 0.5245,
+      "step": 7682
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618767482769559e-05,
+      "loss": 0.6155,
+      "step": 7683
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618709807653488e-05,
+      "loss": 0.4196,
+      "step": 7684
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618652132537417e-05,
+      "loss": 0.4808,
+      "step": 7685
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6185944574213458e-05,
+      "loss": 0.4698,
+      "step": 7686
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6185367823052746e-05,
+      "loss": 0.5182,
+      "step": 7687
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6184791071892033e-05,
+      "loss": 0.5159,
+      "step": 7688
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618421432073132e-05,
+      "loss": 0.5971,
+      "step": 7689
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618363756957061e-05,
+      "loss": 0.6647,
+      "step": 7690
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.61830608184099e-05,
+      "loss": 0.4503,
+      "step": 7691
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6182484067249187e-05,
+      "loss": 0.4878,
+      "step": 7692
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6181907316088475e-05,
+      "loss": 0.5762,
+      "step": 7693
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6181330564927763e-05,
+      "loss": 0.519,
+      "step": 7694
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.618075381376705e-05,
+      "loss": 0.5317,
+      "step": 7695
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6180177062606338e-05,
+      "loss": 0.4069,
+      "step": 7696
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.617960031144563e-05,
+      "loss": 0.4767,
+      "step": 7697
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6179023560284917e-05,
+      "loss": 0.501,
+      "step": 7698
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6178446809124205e-05,
+      "loss": 0.4457,
+      "step": 7699
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6177870057963496e-05,
+      "loss": 0.4547,
+      "step": 7700
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.617729330680278e-05,
+      "loss": 0.5798,
+      "step": 7701
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6176716555642068e-05,
+      "loss": 0.5566,
+      "step": 7702
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.617613980448136e-05,
+      "loss": 0.5086,
+      "step": 7703
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6175563053320646e-05,
+      "loss": 0.5229,
+      "step": 7704
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6174986302159934e-05,
+      "loss": 0.6164,
+      "step": 7705
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6174409550999222e-05,
+      "loss": 0.5007,
+      "step": 7706
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6173832799838513e-05,
+      "loss": 0.5859,
+      "step": 7707
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.61732560486778e-05,
+      "loss": 0.6582,
+      "step": 7708
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6172679297517088e-05,
+      "loss": 0.7156,
+      "step": 7709
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6172102546356376e-05,
+      "loss": 0.5224,
+      "step": 7710
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6171525795195664e-05,
+      "loss": 0.5341,
+      "step": 7711
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.617094904403495e-05,
+      "loss": 0.5756,
+      "step": 7712
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6170372292874242e-05,
+      "loss": 0.7302,
+      "step": 7713
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616979554171353e-05,
+      "loss": 0.5994,
+      "step": 7714
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6169218790552818e-05,
+      "loss": 0.6967,
+      "step": 7715
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6168642039392105e-05,
+      "loss": 0.6171,
+      "step": 7716
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6168065288231393e-05,
+      "loss": 0.5846,
+      "step": 7717
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616748853707068e-05,
+      "loss": 0.5498,
+      "step": 7718
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6166911785909972e-05,
+      "loss": 0.4823,
+      "step": 7719
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616633503474926e-05,
+      "loss": 0.5063,
+      "step": 7720
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6165758283588547e-05,
+      "loss": 0.4993,
+      "step": 7721
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6165181532427835e-05,
+      "loss": 0.4837,
+      "step": 7722
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6164604781267123e-05,
+      "loss": 0.464,
+      "step": 7723
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616402803010641e-05,
+      "loss": 0.5084,
+      "step": 7724
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.61634512789457e-05,
+      "loss": 0.5206,
+      "step": 7725
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616287452778499e-05,
+      "loss": 0.4217,
+      "step": 7726
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6162297776624277e-05,
+      "loss": 0.6062,
+      "step": 7727
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6161721025463564e-05,
+      "loss": 0.4881,
+      "step": 7728
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6161144274302852e-05,
+      "loss": 0.7145,
+      "step": 7729
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.616056752314214e-05,
+      "loss": 0.5429,
+      "step": 7730
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.615999077198143e-05,
+      "loss": 0.5036,
+      "step": 7731
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.615941402082072e-05,
+      "loss": 0.5736,
+      "step": 7732
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6158837269660006e-05,
+      "loss": 0.4918,
+      "step": 7733
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6158260518499297e-05,
+      "loss": 0.4499,
+      "step": 7734
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6157683767338585e-05,
+      "loss": 0.6327,
+      "step": 7735
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6157107016177873e-05,
+      "loss": 0.3893,
+      "step": 7736
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.615653026501716e-05,
+      "loss": 0.5356,
+      "step": 7737
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6155953513856448e-05,
+      "loss": 0.504,
+      "step": 7738
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6155376762695736e-05,
+      "loss": 0.4719,
+      "step": 7739
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6154800011535027e-05,
+      "loss": 0.498,
+      "step": 7740
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6154223260374314e-05,
+      "loss": 0.5573,
+      "step": 7741
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6153646509213602e-05,
+      "loss": 0.5494,
+      "step": 7742
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.615306975805289e-05,
+      "loss": 0.5546,
+      "step": 7743
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6152493006892177e-05,
+      "loss": 0.5307,
+      "step": 7744
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6151916255731465e-05,
+      "loss": 0.5358,
+      "step": 7745
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6151339504570753e-05,
+      "loss": 0.6737,
+      "step": 7746
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6150762753410044e-05,
+      "loss": 0.4611,
+      "step": 7747
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.615018600224933e-05,
+      "loss": 0.709,
+      "step": 7748
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614960925108862e-05,
+      "loss": 0.6239,
+      "step": 7749
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6149032499927907e-05,
+      "loss": 0.5102,
+      "step": 7750
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6148455748767195e-05,
+      "loss": 0.4918,
+      "step": 7751
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6147878997606482e-05,
+      "loss": 0.637,
+      "step": 7752
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6147302246445773e-05,
+      "loss": 0.5256,
+      "step": 7753
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614672549528506e-05,
+      "loss": 0.6819,
+      "step": 7754
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614614874412435e-05,
+      "loss": 0.6108,
+      "step": 7755
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6145571992963636e-05,
+      "loss": 0.5661,
+      "step": 7756
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6144995241802924e-05,
+      "loss": 0.5353,
+      "step": 7757
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6144418490642212e-05,
+      "loss": 0.6452,
+      "step": 7758
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6143841739481503e-05,
+      "loss": 0.6113,
+      "step": 7759
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614326498832079e-05,
+      "loss": 0.4792,
+      "step": 7760
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6142688237160078e-05,
+      "loss": 0.6204,
+      "step": 7761
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614211148599937e-05,
+      "loss": 0.495,
+      "step": 7762
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6141534734838654e-05,
+      "loss": 0.5143,
+      "step": 7763
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.614095798367794e-05,
+      "loss": 0.6312,
+      "step": 7764
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6140381232517232e-05,
+      "loss": 0.6134,
+      "step": 7765
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.613980448135652e-05,
+      "loss": 0.5192,
+      "step": 7766
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6139227730195808e-05,
+      "loss": 0.6639,
+      "step": 7767
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.61386509790351e-05,
+      "loss": 0.5758,
+      "step": 7768
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6138074227874387e-05,
+      "loss": 0.5654,
+      "step": 7769
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6137497476713674e-05,
+      "loss": 0.4294,
+      "step": 7770
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6136920725552962e-05,
+      "loss": 0.5035,
+      "step": 7771
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.613634397439225e-05,
+      "loss": 0.5482,
+      "step": 7772
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6135767223231537e-05,
+      "loss": 0.4501,
+      "step": 7773
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.613519047207083e-05,
+      "loss": 0.4151,
+      "step": 7774
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.6134613720910116e-05,
+      "loss": 0.4689,
+      "step": 7775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6134036969749404e-05,
+      "loss": 0.5564,
+      "step": 7776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.613346021858869e-05,
+      "loss": 0.5423,
+      "step": 7777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.613288346742798e-05,
+      "loss": 0.5094,
+      "step": 7778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6132306716267267e-05,
+      "loss": 0.6494,
+      "step": 7779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6131729965106554e-05,
+      "loss": 0.5704,
+      "step": 7780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6131153213945846e-05,
+      "loss": 0.6473,
+      "step": 7781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6130576462785133e-05,
+      "loss": 0.5899,
+      "step": 7782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612999971162442e-05,
+      "loss": 0.4877,
+      "step": 7783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612942296046371e-05,
+      "loss": 0.7453,
+      "step": 7784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6128846209302996e-05,
+      "loss": 0.5244,
+      "step": 7785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6128269458142284e-05,
+      "loss": 0.4018,
+      "step": 7786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6127692706981575e-05,
+      "loss": 0.576,
+      "step": 7787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6127115955820863e-05,
+      "loss": 0.4651,
+      "step": 7788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612653920466015e-05,
+      "loss": 0.7273,
+      "step": 7789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6125962453499438e-05,
+      "loss": 0.5554,
+      "step": 7790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6125385702338726e-05,
+      "loss": 0.5102,
+      "step": 7791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6124808951178013e-05,
+      "loss": 0.606,
+      "step": 7792
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6124232200017305e-05,
+      "loss": 0.5272,
+      "step": 7793
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6123655448856592e-05,
+      "loss": 0.509,
+      "step": 7794
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612307869769588e-05,
+      "loss": 0.6374,
+      "step": 7795
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612250194653517e-05,
+      "loss": 0.4477,
+      "step": 7796
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.612192519537446e-05,
+      "loss": 0.4808,
+      "step": 7797
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6121348444213746e-05,
+      "loss": 0.4664,
+      "step": 7798
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6120771693053034e-05,
+      "loss": 0.5762,
+      "step": 7799
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6120194941892322e-05,
+      "loss": 0.5465,
+      "step": 7800
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611961819073161e-05,
+      "loss": 0.433,
+      "step": 7801
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.61190414395709e-05,
+      "loss": 0.4959,
+      "step": 7802
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6118464688410188e-05,
+      "loss": 0.6161,
+      "step": 7803
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6117887937249476e-05,
+      "loss": 0.4897,
+      "step": 7804
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6117311186088764e-05,
+      "loss": 0.5146,
+      "step": 7805
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611673443492805e-05,
+      "loss": 0.5426,
+      "step": 7806
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611615768376734e-05,
+      "loss": 0.4835,
+      "step": 7807
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611558093260663e-05,
+      "loss": 0.5533,
+      "step": 7808
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6115004181445918e-05,
+      "loss": 0.6482,
+      "step": 7809
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6114427430285205e-05,
+      "loss": 0.4894,
+      "step": 7810
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6113850679124493e-05,
+      "loss": 0.4797,
+      "step": 7811
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611327392796378e-05,
+      "loss": 0.5961,
+      "step": 7812
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.611269717680307e-05,
+      "loss": 0.5106,
+      "step": 7813
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6112120425642356e-05,
+      "loss": 0.6384,
+      "step": 7814
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6111543674481647e-05,
+      "loss": 0.4853,
+      "step": 7815
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6110966923320935e-05,
+      "loss": 0.5243,
+      "step": 7816
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6110390172160223e-05,
+      "loss": 0.6034,
+      "step": 7817
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.610981342099951e-05,
+      "loss": 0.6178,
+      "step": 7818
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6109236669838798e-05,
+      "loss": 0.4989,
+      "step": 7819
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6108659918678086e-05,
+      "loss": 0.4456,
+      "step": 7820
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6108083167517377e-05,
+      "loss": 0.5943,
+      "step": 7821
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6107506416356664e-05,
+      "loss": 0.5316,
+      "step": 7822
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6106929665195952e-05,
+      "loss": 0.6443,
+      "step": 7823
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6106352914035243e-05,
+      "loss": 0.6102,
+      "step": 7824
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6105776162874527e-05,
+      "loss": 0.6297,
+      "step": 7825
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6105199411713815e-05,
+      "loss": 0.4928,
+      "step": 7826
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6104622660553106e-05,
+      "loss": 0.5583,
+      "step": 7827
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6104045909392394e-05,
+      "loss": 0.5353,
+      "step": 7828
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.610346915823168e-05,
+      "loss": 0.5624,
+      "step": 7829
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6102892407070973e-05,
+      "loss": 0.431,
+      "step": 7830
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.610231565591026e-05,
+      "loss": 0.5436,
+      "step": 7831
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6101738904749548e-05,
+      "loss": 0.6528,
+      "step": 7832
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6101162153588836e-05,
+      "loss": 0.634,
+      "step": 7833
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6100585402428123e-05,
+      "loss": 0.465,
+      "step": 7834
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.610000865126741e-05,
+      "loss": 0.5404,
+      "step": 7835
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6099431900106702e-05,
+      "loss": 0.5536,
+      "step": 7836
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609885514894599e-05,
+      "loss": 0.5515,
+      "step": 7837
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6098278397785277e-05,
+      "loss": 0.5562,
+      "step": 7838
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6097701646624565e-05,
+      "loss": 0.4157,
+      "step": 7839
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6097124895463853e-05,
+      "loss": 0.602,
+      "step": 7840
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609654814430314e-05,
+      "loss": 0.4997,
+      "step": 7841
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609597139314243e-05,
+      "loss": 0.539,
+      "step": 7842
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609539464198172e-05,
+      "loss": 0.5204,
+      "step": 7843
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6094817890821007e-05,
+      "loss": 0.5864,
+      "step": 7844
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6094241139660295e-05,
+      "loss": 0.5743,
+      "step": 7845
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6093664388499582e-05,
+      "loss": 0.4952,
+      "step": 7846
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609308763733887e-05,
+      "loss": 0.4416,
+      "step": 7847
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6092510886178158e-05,
+      "loss": 0.5415,
+      "step": 7848
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.609193413501745e-05,
+      "loss": 0.4009,
+      "step": 7849
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6091357383856736e-05,
+      "loss": 0.5878,
+      "step": 7850
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6090780632696024e-05,
+      "loss": 0.4356,
+      "step": 7851
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6090203881535312e-05,
+      "loss": 0.461,
+      "step": 7852
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.60896271303746e-05,
+      "loss": 0.6308,
+      "step": 7853
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6089050379213887e-05,
+      "loss": 0.4622,
+      "step": 7854
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6088473628053178e-05,
+      "loss": 0.4666,
+      "step": 7855
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6087896876892466e-05,
+      "loss": 0.5983,
+      "step": 7856
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6087320125731754e-05,
+      "loss": 0.5272,
+      "step": 7857
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6086743374571045e-05,
+      "loss": 0.4407,
+      "step": 7858
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6086166623410332e-05,
+      "loss": 0.482,
+      "step": 7859
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.608558987224962e-05,
+      "loss": 0.5895,
+      "step": 7860
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6085013121088908e-05,
+      "loss": 0.5972,
+      "step": 7861
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6084436369928195e-05,
+      "loss": 0.5579,
+      "step": 7862
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6083859618767483e-05,
+      "loss": 0.6711,
+      "step": 7863
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6083282867606774e-05,
+      "loss": 0.5024,
+      "step": 7864
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6082706116446062e-05,
+      "loss": 0.5766,
+      "step": 7865
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.608212936528535e-05,
+      "loss": 0.4496,
+      "step": 7866
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6081552614124637e-05,
+      "loss": 0.5388,
+      "step": 7867
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6080975862963925e-05,
+      "loss": 0.4864,
+      "step": 7868
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6080399111803213e-05,
+      "loss": 0.6103,
+      "step": 7869
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6079822360642504e-05,
+      "loss": 0.5823,
+      "step": 7870
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607924560948179e-05,
+      "loss": 0.5364,
+      "step": 7871
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607866885832108e-05,
+      "loss": 0.536,
+      "step": 7872
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6078092107160367e-05,
+      "loss": 0.5246,
+      "step": 7873
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6077515355999654e-05,
+      "loss": 0.5133,
+      "step": 7874
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6076938604838942e-05,
+      "loss": 0.5465,
+      "step": 7875
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6076361853678233e-05,
+      "loss": 0.5572,
+      "step": 7876
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607578510251752e-05,
+      "loss": 0.4854,
+      "step": 7877
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607520835135681e-05,
+      "loss": 0.5889,
+      "step": 7878
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6074631600196096e-05,
+      "loss": 0.7801,
+      "step": 7879
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6074054849035384e-05,
+      "loss": 0.5891,
+      "step": 7880
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607347809787467e-05,
+      "loss": 0.5401,
+      "step": 7881
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607290134671396e-05,
+      "loss": 0.6357,
+      "step": 7882
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.607232459555325e-05,
+      "loss": 0.5827,
+      "step": 7883
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6071747844392538e-05,
+      "loss": 0.6086,
+      "step": 7884
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6071171093231826e-05,
+      "loss": 0.58,
+      "step": 7885
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6070594342071117e-05,
+      "loss": 0.6658,
+      "step": 7886
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.60700175909104e-05,
+      "loss": 0.5107,
+      "step": 7887
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606944083974969e-05,
+      "loss": 0.5852,
+      "step": 7888
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606886408858898e-05,
+      "loss": 0.4823,
+      "step": 7889
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6068287337428268e-05,
+      "loss": 0.5278,
+      "step": 7890
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6067710586267555e-05,
+      "loss": 0.5484,
+      "step": 7891
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6067133835106846e-05,
+      "loss": 0.5689,
+      "step": 7892
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6066557083946134e-05,
+      "loss": 0.4734,
+      "step": 7893
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606598033278542e-05,
+      "loss": 0.5933,
+      "step": 7894
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606540358162471e-05,
+      "loss": 0.5992,
+      "step": 7895
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6064826830463997e-05,
+      "loss": 0.5792,
+      "step": 7896
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6064250079303285e-05,
+      "loss": 0.4915,
+      "step": 7897
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6063673328142576e-05,
+      "loss": 0.5697,
+      "step": 7898
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6063096576981864e-05,
+      "loss": 0.5071,
+      "step": 7899
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606251982582115e-05,
+      "loss": 0.5538,
+      "step": 7900
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.606194307466044e-05,
+      "loss": 0.5906,
+      "step": 7901
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6061366323499727e-05,
+      "loss": 0.6274,
+      "step": 7902
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6060789572339014e-05,
+      "loss": 0.534,
+      "step": 7903
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6060212821178305e-05,
+      "loss": 0.5592,
+      "step": 7904
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6059636070017593e-05,
+      "loss": 0.5312,
+      "step": 7905
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605905931885688e-05,
+      "loss": 0.4788,
+      "step": 7906
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605848256769617e-05,
+      "loss": 0.6226,
+      "step": 7907
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6057905816535456e-05,
+      "loss": 0.5281,
+      "step": 7908
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6057329065374744e-05,
+      "loss": 0.5938,
+      "step": 7909
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6056752314214035e-05,
+      "loss": 0.4892,
+      "step": 7910
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6056175563053323e-05,
+      "loss": 0.5286,
+      "step": 7911
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605559881189261e-05,
+      "loss": 0.5329,
+      "step": 7912
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.60550220607319e-05,
+      "loss": 0.4976,
+      "step": 7913
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6054445309571186e-05,
+      "loss": 0.628,
+      "step": 7914
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6053868558410473e-05,
+      "loss": 0.6108,
+      "step": 7915
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605329180724976e-05,
+      "loss": 0.6128,
+      "step": 7916
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6052715056089052e-05,
+      "loss": 0.6344,
+      "step": 7917
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605213830492834e-05,
+      "loss": 0.5747,
+      "step": 7918
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6051561553767627e-05,
+      "loss": 0.5919,
+      "step": 7919
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.605098480260692e-05,
+      "loss": 0.6064,
+      "step": 7920
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6050408051446206e-05,
+      "loss": 0.4052,
+      "step": 7921
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6049831300285494e-05,
+      "loss": 0.5512,
+      "step": 7922
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.604925454912478e-05,
+      "loss": 0.5421,
+      "step": 7923
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.604867779796407e-05,
+      "loss": 0.5247,
+      "step": 7924
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6048101046803357e-05,
+      "loss": 0.5187,
+      "step": 7925
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6047524295642648e-05,
+      "loss": 0.5469,
+      "step": 7926
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6046947544481936e-05,
+      "loss": 0.5105,
+      "step": 7927
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6046370793321223e-05,
+      "loss": 0.4628,
+      "step": 7928
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.604579404216051e-05,
+      "loss": 0.5766,
+      "step": 7929
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.60452172909998e-05,
+      "loss": 0.5714,
+      "step": 7930
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6044640539839086e-05,
+      "loss": 0.54,
+      "step": 7931
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6044063788678377e-05,
+      "loss": 0.6295,
+      "step": 7932
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6043487037517665e-05,
+      "loss": 0.6725,
+      "step": 7933
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6042910286356953e-05,
+      "loss": 0.5076,
+      "step": 7934
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.604233353519624e-05,
+      "loss": 0.4558,
+      "step": 7935
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6041756784035528e-05,
+      "loss": 0.4118,
+      "step": 7936
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6041180032874816e-05,
+      "loss": 0.5264,
+      "step": 7937
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6040603281714107e-05,
+      "loss": 0.6347,
+      "step": 7938
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6040026530553395e-05,
+      "loss": 0.614,
+      "step": 7939
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6039449779392682e-05,
+      "loss": 0.507,
+      "step": 7940
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.603887302823197e-05,
+      "loss": 0.5367,
+      "step": 7941
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6038296277071258e-05,
+      "loss": 0.572,
+      "step": 7942
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6037719525910545e-05,
+      "loss": 0.5944,
+      "step": 7943
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6037142774749836e-05,
+      "loss": 0.506,
+      "step": 7944
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6036566023589124e-05,
+      "loss": 0.5255,
+      "step": 7945
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6035989272428412e-05,
+      "loss": 0.5088,
+      "step": 7946
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6035412521267703e-05,
+      "loss": 0.517,
+      "step": 7947
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.603483577010699e-05,
+      "loss": 0.5092,
+      "step": 7948
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6034259018946278e-05,
+      "loss": 0.5639,
+      "step": 7949
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6033682267785563e-05,
+      "loss": 0.5071,
+      "step": 7950
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.6033105516624854e-05,
+      "loss": 0.5157,
+      "step": 7951
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.603252876546414e-05,
+      "loss": 0.3886,
+      "step": 7952
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.603195201430343e-05,
+      "loss": 0.4783,
+      "step": 7953
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.603137526314272e-05,
+      "loss": 0.6216,
+      "step": 7954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6030798511982008e-05,
+      "loss": 0.5356,
+      "step": 7955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6030221760821295e-05,
+      "loss": 0.535,
+      "step": 7956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6029645009660583e-05,
+      "loss": 0.4993,
+      "step": 7957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602906825849987e-05,
+      "loss": 0.4605,
+      "step": 7958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602849150733916e-05,
+      "loss": 0.557,
+      "step": 7959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602791475617845e-05,
+      "loss": 0.5838,
+      "step": 7960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6027338005017737e-05,
+      "loss": 0.4746,
+      "step": 7961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6026761253857025e-05,
+      "loss": 0.5021,
+      "step": 7962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6026184502696313e-05,
+      "loss": 0.55,
+      "step": 7963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.60256077515356e-05,
+      "loss": 0.4375,
+      "step": 7964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6025031000374888e-05,
+      "loss": 0.552,
+      "step": 7965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602445424921418e-05,
+      "loss": 0.5424,
+      "step": 7966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6023877498053467e-05,
+      "loss": 0.4281,
+      "step": 7967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6023300746892754e-05,
+      "loss": 0.6472,
+      "step": 7968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6022723995732042e-05,
+      "loss": 0.5833,
+      "step": 7969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602214724457133e-05,
+      "loss": 0.4874,
+      "step": 7970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6021570493410617e-05,
+      "loss": 0.5594,
+      "step": 7971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.602099374224991e-05,
+      "loss": 0.4857,
+      "step": 7972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6020416991089196e-05,
+      "loss": 0.7091,
+      "step": 7973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6019840239928484e-05,
+      "loss": 0.4645,
+      "step": 7974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6019263488767775e-05,
+      "loss": 0.5648,
+      "step": 7975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.601868673760706e-05,
+      "loss": 0.5718,
+      "step": 7976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6018109986446347e-05,
+      "loss": 0.5333,
+      "step": 7977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6017533235285638e-05,
+      "loss": 0.5267,
+      "step": 7978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6016956484124926e-05,
+      "loss": 0.5208,
+      "step": 7979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6016379732964213e-05,
+      "loss": 0.5439,
+      "step": 7980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6015802981803505e-05,
+      "loss": 0.5245,
+      "step": 7981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6015226230642792e-05,
+      "loss": 0.5216,
+      "step": 7982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.601464947948208e-05,
+      "loss": 0.5622,
+      "step": 7983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6014072728321368e-05,
+      "loss": 0.5834,
+      "step": 7984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6013495977160655e-05,
+      "loss": 0.5953,
+      "step": 7985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6012919225999943e-05,
+      "loss": 0.4919,
+      "step": 7986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.601234247483923e-05,
+      "loss": 0.6094,
+      "step": 7987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.601176572367852e-05,
+      "loss": 0.5246,
+      "step": 7988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.601118897251781e-05,
+      "loss": 0.6084,
+      "step": 7989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6010612221357097e-05,
+      "loss": 0.6467,
+      "step": 7990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6010035470196385e-05,
+      "loss": 0.6235,
+      "step": 7991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6009458719035672e-05,
+      "loss": 0.4836,
+      "step": 7992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600888196787496e-05,
+      "loss": 0.481,
+      "step": 7993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600830521671425e-05,
+      "loss": 0.5806,
+      "step": 7994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600772846555354e-05,
+      "loss": 0.5517,
+      "step": 7995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6007151714392827e-05,
+      "loss": 0.3965,
+      "step": 7996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6006574963232114e-05,
+      "loss": 0.5513,
+      "step": 7997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6005998212071402e-05,
+      "loss": 0.4998,
+      "step": 7998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600542146091069e-05,
+      "loss": 0.5627,
+      "step": 7999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600484470974998e-05,
+      "loss": 0.5227,
+      "step": 8000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600426795858927e-05,
+      "loss": 0.6438,
+      "step": 8001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6003691207428556e-05,
+      "loss": 0.6137,
+      "step": 8002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6003114456267844e-05,
+      "loss": 0.498,
+      "step": 8003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600253770510713e-05,
+      "loss": 0.552,
+      "step": 8004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600196095394642e-05,
+      "loss": 0.5137,
+      "step": 8005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.600138420278571e-05,
+      "loss": 0.5826,
+      "step": 8006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6000807451624998e-05,
+      "loss": 0.582,
+      "step": 8007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.6000230700464286e-05,
+      "loss": 0.5564,
+      "step": 8008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5999653949303577e-05,
+      "loss": 0.4359,
+      "step": 8009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5999077198142864e-05,
+      "loss": 0.6018,
+      "step": 8010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5998500446982152e-05,
+      "loss": 0.5701,
+      "step": 8011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.599792369582144e-05,
+      "loss": 0.4911,
+      "step": 8012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5997346944660727e-05,
+      "loss": 0.5816,
+      "step": 8013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5996770193500015e-05,
+      "loss": 0.6105,
+      "step": 8014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5996193442339306e-05,
+      "loss": 0.492,
+      "step": 8015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5995616691178594e-05,
+      "loss": 0.4619,
+      "step": 8016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.599503994001788e-05,
+      "loss": 0.4948,
+      "step": 8017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.599446318885717e-05,
+      "loss": 0.5026,
+      "step": 8018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5993886437696457e-05,
+      "loss": 0.5554,
+      "step": 8019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5993309686535745e-05,
+      "loss": 0.4689,
+      "step": 8020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5992732935375032e-05,
+      "loss": 0.5043,
+      "step": 8021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5992156184214323e-05,
+      "loss": 0.5356,
+      "step": 8022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.599157943305361e-05,
+      "loss": 0.4735,
+      "step": 8023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.59910026818929e-05,
+      "loss": 0.5651,
+      "step": 8024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5990425930732186e-05,
+      "loss": 0.518,
+      "step": 8025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5989849179571474e-05,
+      "loss": 0.6484,
+      "step": 8026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5989272428410762e-05,
+      "loss": 0.5635,
+      "step": 8027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5988695677250053e-05,
+      "loss": 0.4632,
+      "step": 8028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.598811892608934e-05,
+      "loss": 0.547,
+      "step": 8029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5987542174928628e-05,
+      "loss": 0.5519,
+      "step": 8030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5986965423767916e-05,
+      "loss": 0.556,
+      "step": 8031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5986388672607204e-05,
+      "loss": 0.5012,
+      "step": 8032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.598581192144649e-05,
+      "loss": 0.478,
+      "step": 8033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5985235170285782e-05,
+      "loss": 0.569,
+      "step": 8034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.598465841912507e-05,
+      "loss": 0.5278,
+      "step": 8035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5984081667964358e-05,
+      "loss": 0.5834,
+      "step": 8036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.598350491680365e-05,
+      "loss": 0.5142,
+      "step": 8037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5982928165642933e-05,
+      "loss": 0.516,
+      "step": 8038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.598235141448222e-05,
+      "loss": 0.5816,
+      "step": 8039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5981774663321512e-05,
+      "loss": 0.533,
+      "step": 8040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.59811979121608e-05,
+      "loss": 0.5473,
+      "step": 8041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5980621161000087e-05,
+      "loss": 0.4654,
+      "step": 8042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5980044409839378e-05,
+      "loss": 0.4828,
+      "step": 8043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5979467658678666e-05,
+      "loss": 0.4935,
+      "step": 8044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5978890907517954e-05,
+      "loss": 0.4958,
+      "step": 8045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.597831415635724e-05,
+      "loss": 0.6199,
+      "step": 8046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.597773740519653e-05,
+      "loss": 0.4152,
+      "step": 8047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5977160654035817e-05,
+      "loss": 0.614,
+      "step": 8048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5976583902875108e-05,
+      "loss": 0.5162,
+      "step": 8049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5976007151714395e-05,
+      "loss": 0.6244,
+      "step": 8050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5975430400553683e-05,
+      "loss": 0.5169,
+      "step": 8051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.597485364939297e-05,
+      "loss": 0.5506,
+      "step": 8052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.597427689823226e-05,
+      "loss": 0.5616,
+      "step": 8053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5973700147071546e-05,
+      "loss": 0.594,
+      "step": 8054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5973123395910834e-05,
+      "loss": 0.4668,
+      "step": 8055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5972546644750125e-05,
+      "loss": 0.5193,
+      "step": 8056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5971969893589413e-05,
+      "loss": 0.5077,
+      "step": 8057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.59713931424287e-05,
+      "loss": 0.5776,
+      "step": 8058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5970816391267988e-05,
+      "loss": 0.5377,
+      "step": 8059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5970239640107276e-05,
+      "loss": 0.5062,
+      "step": 8060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5969662888946563e-05,
+      "loss": 0.5135,
+      "step": 8061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5969086137785854e-05,
+      "loss": 0.5531,
+      "step": 8062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5968509386625142e-05,
+      "loss": 0.6572,
+      "step": 8063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596793263546443e-05,
+      "loss": 0.4982,
+      "step": 8064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5967355884303717e-05,
+      "loss": 0.4856,
+      "step": 8065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5966779133143005e-05,
+      "loss": 0.608,
+      "step": 8066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5966202381982293e-05,
+      "loss": 0.6423,
+      "step": 8067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5965625630821584e-05,
+      "loss": 0.5106,
+      "step": 8068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596504887966087e-05,
+      "loss": 0.5853,
+      "step": 8069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596447212850016e-05,
+      "loss": 0.4752,
+      "step": 8070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596389537733945e-05,
+      "loss": 0.5827,
+      "step": 8071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5963318626178738e-05,
+      "loss": 0.4108,
+      "step": 8072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5962741875018026e-05,
+      "loss": 0.52,
+      "step": 8073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5962165123857313e-05,
+      "loss": 0.4066,
+      "step": 8074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.59615883726966e-05,
+      "loss": 0.4243,
+      "step": 8075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596101162153589e-05,
+      "loss": 0.6294,
+      "step": 8076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.596043487037518e-05,
+      "loss": 0.4918,
+      "step": 8077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5959858119214468e-05,
+      "loss": 0.5708,
+      "step": 8078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5959281368053755e-05,
+      "loss": 0.5279,
+      "step": 8079
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5958704616893043e-05,
+      "loss": 0.5461,
+      "step": 8080
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.595812786573233e-05,
+      "loss": 0.5255,
+      "step": 8081
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5957551114571618e-05,
+      "loss": 0.5799,
+      "step": 8082
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.595697436341091e-05,
+      "loss": 0.4027,
+      "step": 8083
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5956397612250197e-05,
+      "loss": 0.4821,
+      "step": 8084
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5955820861089485e-05,
+      "loss": 0.6476,
+      "step": 8085
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5955244109928772e-05,
+      "loss": 0.473,
+      "step": 8086
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.595466735876806e-05,
+      "loss": 0.5084,
+      "step": 8087
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5954090607607348e-05,
+      "loss": 0.6367,
+      "step": 8088
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5953513856446635e-05,
+      "loss": 0.5222,
+      "step": 8089
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5952937105285927e-05,
+      "loss": 0.6051,
+      "step": 8090
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5952360354125214e-05,
+      "loss": 0.5394,
+      "step": 8091
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5951783602964502e-05,
+      "loss": 0.632,
+      "step": 8092
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.595120685180379e-05,
+      "loss": 0.5707,
+      "step": 8093
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5950630100643077e-05,
+      "loss": 0.6445,
+      "step": 8094
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5950053349482365e-05,
+      "loss": 0.4306,
+      "step": 8095
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5949476598321656e-05,
+      "loss": 0.4801,
+      "step": 8096
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5948899847160944e-05,
+      "loss": 0.4754,
+      "step": 8097
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594832309600023e-05,
+      "loss": 0.5128,
+      "step": 8098
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5947746344839522e-05,
+      "loss": 0.5694,
+      "step": 8099
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5947169593678807e-05,
+      "loss": 0.5852,
+      "step": 8100
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5946592842518094e-05,
+      "loss": 0.5876,
+      "step": 8101
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5946016091357386e-05,
+      "loss": 0.5379,
+      "step": 8102
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5945439340196673e-05,
+      "loss": 0.5892,
+      "step": 8103
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594486258903596e-05,
+      "loss": 0.4705,
+      "step": 8104
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5944285837875252e-05,
+      "loss": 0.6675,
+      "step": 8105
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594370908671454e-05,
+      "loss": 0.4879,
+      "step": 8106
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5943132335553827e-05,
+      "loss": 0.3821,
+      "step": 8107
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5942555584393115e-05,
+      "loss": 0.6877,
+      "step": 8108
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5941978833232403e-05,
+      "loss": 0.5072,
+      "step": 8109
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594140208207169e-05,
+      "loss": 0.5115,
+      "step": 8110
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594082533091098e-05,
+      "loss": 0.5707,
+      "step": 8111
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.594024857975027e-05,
+      "loss": 0.4217,
+      "step": 8112
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5939671828589557e-05,
+      "loss": 0.4482,
+      "step": 8113
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5939095077428845e-05,
+      "loss": 0.6391,
+      "step": 8114
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5938518326268132e-05,
+      "loss": 0.5537,
+      "step": 8115
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593794157510742e-05,
+      "loss": 0.4219,
+      "step": 8116
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593736482394671e-05,
+      "loss": 0.5265,
+      "step": 8117
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5936788072786e-05,
+      "loss": 0.5192,
+      "step": 8118
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5936211321625286e-05,
+      "loss": 0.5594,
+      "step": 8119
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5935634570464574e-05,
+      "loss": 0.5083,
+      "step": 8120
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5935057819303862e-05,
+      "loss": 0.4579,
+      "step": 8121
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593448106814315e-05,
+      "loss": 0.4887,
+      "step": 8122
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593390431698244e-05,
+      "loss": 0.4947,
+      "step": 8123
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5933327565821728e-05,
+      "loss": 0.425,
+      "step": 8124
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5932750814661016e-05,
+      "loss": 0.4048,
+      "step": 8125
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5932174063500304e-05,
+      "loss": 0.4901,
+      "step": 8126
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593159731233959e-05,
+      "loss": 0.5325,
+      "step": 8127
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.593102056117888e-05,
+      "loss": 0.5844,
+      "step": 8128
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5930443810018167e-05,
+      "loss": 0.6389,
+      "step": 8129
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5929867058857458e-05,
+      "loss": 0.5809,
+      "step": 8130
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5929290307696745e-05,
+      "loss": 0.6039,
+      "step": 8131
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5928713556536033e-05,
+      "loss": 0.5185,
+      "step": 8132
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.5928136805375324e-05,
+      "loss": 0.5754,
+      "step": 8133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5927560054214612e-05,
+      "loss": 0.5429,
+      "step": 8134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.59269833030539e-05,
+      "loss": 0.5662,
+      "step": 8135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5926406551893187e-05,
+      "loss": 0.4081,
+      "step": 8136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5925829800732475e-05,
+      "loss": 0.4731,
+      "step": 8137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5925253049571763e-05,
+      "loss": 0.566,
+      "step": 8138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5924676298411054e-05,
+      "loss": 0.5836,
+      "step": 8139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.592409954725034e-05,
+      "loss": 0.62,
+      "step": 8140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.592352279608963e-05,
+      "loss": 0.6123,
+      "step": 8141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5922946044928917e-05,
+      "loss": 0.5013,
+      "step": 8142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5922369293768204e-05,
+      "loss": 0.6004,
+      "step": 8143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5921792542607492e-05,
+      "loss": 0.463,
+      "step": 8144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5921215791446783e-05,
+      "loss": 0.4775,
+      "step": 8145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.592063904028607e-05,
+      "loss": 0.5357,
+      "step": 8146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.592006228912536e-05,
+      "loss": 0.57,
+      "step": 8147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5919485537964646e-05,
+      "loss": 0.5173,
+      "step": 8148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5918908786803934e-05,
+      "loss": 0.6718,
+      "step": 8149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.591833203564322e-05,
+      "loss": 0.5742,
+      "step": 8150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5917755284482513e-05,
+      "loss": 0.4943,
+      "step": 8151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.59171785333218e-05,
+      "loss": 0.5051,
+      "step": 8152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5916601782161088e-05,
+      "loss": 0.6071,
+      "step": 8153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5916025031000376e-05,
+      "loss": 0.5444,
+      "step": 8154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5915448279839663e-05,
+      "loss": 0.5555,
+      "step": 8155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.591487152867895e-05,
+      "loss": 0.5363,
+      "step": 8156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5914294777518242e-05,
+      "loss": 0.528,
+      "step": 8157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.591371802635753e-05,
+      "loss": 0.5376,
+      "step": 8158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5913141275196817e-05,
+      "loss": 0.4046,
+      "step": 8159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5912564524036105e-05,
+      "loss": 0.4344,
+      "step": 8160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5911987772875396e-05,
+      "loss": 0.5321,
+      "step": 8161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.591141102171468e-05,
+      "loss": 0.5844,
+      "step": 8162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5910834270553968e-05,
+      "loss": 0.6625,
+      "step": 8163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.591025751939326e-05,
+      "loss": 0.5509,
+      "step": 8164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5909680768232547e-05,
+      "loss": 0.5846,
+      "step": 8165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5909104017071835e-05,
+      "loss": 0.598,
+      "step": 8166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5908527265911126e-05,
+      "loss": 0.4486,
+      "step": 8167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5907950514750413e-05,
+      "loss": 0.5486,
+      "step": 8168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.59073737635897e-05,
+      "loss": 0.5635,
+      "step": 8169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.590679701242899e-05,
+      "loss": 0.4719,
+      "step": 8170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5906220261268276e-05,
+      "loss": 0.4647,
+      "step": 8171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5905643510107564e-05,
+      "loss": 0.5781,
+      "step": 8172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5905066758946855e-05,
+      "loss": 0.5476,
+      "step": 8173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5904490007786143e-05,
+      "loss": 0.6212,
+      "step": 8174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.590391325662543e-05,
+      "loss": 0.6415,
+      "step": 8175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5903336505464718e-05,
+      "loss": 0.4987,
+      "step": 8176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5902759754304006e-05,
+      "loss": 0.4685,
+      "step": 8177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5902183003143294e-05,
+      "loss": 0.5357,
+      "step": 8178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5901606251982585e-05,
+      "loss": 0.5395,
+      "step": 8179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5901029500821872e-05,
+      "loss": 0.5483,
+      "step": 8180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.590045274966116e-05,
+      "loss": 0.4568,
+      "step": 8181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5899875998500448e-05,
+      "loss": 0.4219,
+      "step": 8182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5899299247339735e-05,
+      "loss": 0.5234,
+      "step": 8183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5898722496179023e-05,
+      "loss": 0.4688,
+      "step": 8184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5898145745018314e-05,
+      "loss": 0.4342,
+      "step": 8185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5897568993857602e-05,
+      "loss": 0.5274,
+      "step": 8186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589699224269689e-05,
+      "loss": 0.5612,
+      "step": 8187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589641549153618e-05,
+      "loss": 0.5981,
+      "step": 8188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5895838740375465e-05,
+      "loss": 0.6751,
+      "step": 8189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5895261989214753e-05,
+      "loss": 0.5321,
+      "step": 8190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5894685238054044e-05,
+      "loss": 0.6287,
+      "step": 8191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589410848689333e-05,
+      "loss": 0.5198,
+      "step": 8192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589353173573262e-05,
+      "loss": 0.5491,
+      "step": 8193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5892954984571907e-05,
+      "loss": 0.6224,
+      "step": 8194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5892378233411198e-05,
+      "loss": 0.6303,
+      "step": 8195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5891801482250486e-05,
+      "loss": 0.5193,
+      "step": 8196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5891224731089773e-05,
+      "loss": 0.4655,
+      "step": 8197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589064797992906e-05,
+      "loss": 0.5728,
+      "step": 8198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.589007122876835e-05,
+      "loss": 0.4982,
+      "step": 8199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5889494477607636e-05,
+      "loss": 0.5206,
+      "step": 8200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5888917726446927e-05,
+      "loss": 0.4536,
+      "step": 8201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5888340975286215e-05,
+      "loss": 0.5414,
+      "step": 8202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5887764224125503e-05,
+      "loss": 0.646,
+      "step": 8203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.588718747296479e-05,
+      "loss": 0.5913,
+      "step": 8204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5886610721804078e-05,
+      "loss": 0.6119,
+      "step": 8205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5886033970643366e-05,
+      "loss": 0.4564,
+      "step": 8206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5885457219482657e-05,
+      "loss": 0.5236,
+      "step": 8207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5884880468321945e-05,
+      "loss": 0.5314,
+      "step": 8208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5884303717161232e-05,
+      "loss": 0.449,
+      "step": 8209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.588372696600052e-05,
+      "loss": 0.5166,
+      "step": 8210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5883150214839808e-05,
+      "loss": 0.515,
+      "step": 8211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5882573463679095e-05,
+      "loss": 0.5155,
+      "step": 8212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5881996712518386e-05,
+      "loss": 0.7079,
+      "step": 8213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5881419961357674e-05,
+      "loss": 0.5986,
+      "step": 8214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5880843210196962e-05,
+      "loss": 0.4612,
+      "step": 8215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.588026645903625e-05,
+      "loss": 0.7259,
+      "step": 8216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5879689707875537e-05,
+      "loss": 0.5612,
+      "step": 8217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5879112956714825e-05,
+      "loss": 0.4881,
+      "step": 8218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5878536205554116e-05,
+      "loss": 0.6637,
+      "step": 8219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5877959454393404e-05,
+      "loss": 0.4668,
+      "step": 8220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.587738270323269e-05,
+      "loss": 0.4429,
+      "step": 8221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5876805952071982e-05,
+      "loss": 0.5792,
+      "step": 8222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.587622920091127e-05,
+      "loss": 0.5156,
+      "step": 8223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5875652449750554e-05,
+      "loss": 0.5289,
+      "step": 8224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5875075698589845e-05,
+      "loss": 0.5167,
+      "step": 8225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5874498947429133e-05,
+      "loss": 0.4197,
+      "step": 8226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.587392219626842e-05,
+      "loss": 0.593,
+      "step": 8227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.587334544510771e-05,
+      "loss": 0.5362,
+      "step": 8228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5872768693947e-05,
+      "loss": 0.5758,
+      "step": 8229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5872191942786287e-05,
+      "loss": 0.5149,
+      "step": 8230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5871615191625575e-05,
+      "loss": 0.4754,
+      "step": 8231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5871038440464863e-05,
+      "loss": 0.7002,
+      "step": 8232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.587046168930415e-05,
+      "loss": 0.5224,
+      "step": 8233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5869884938143438e-05,
+      "loss": 0.5945,
+      "step": 8234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586930818698273e-05,
+      "loss": 0.6414,
+      "step": 8235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5868731435822017e-05,
+      "loss": 0.5045,
+      "step": 8236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5868154684661304e-05,
+      "loss": 0.5145,
+      "step": 8237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5867577933500592e-05,
+      "loss": 0.6361,
+      "step": 8238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586700118233988e-05,
+      "loss": 0.4739,
+      "step": 8239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5866424431179167e-05,
+      "loss": 0.5838,
+      "step": 8240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586584768001846e-05,
+      "loss": 0.6036,
+      "step": 8241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5865270928857746e-05,
+      "loss": 0.4754,
+      "step": 8242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5864694177697034e-05,
+      "loss": 0.4792,
+      "step": 8243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586411742653632e-05,
+      "loss": 0.4839,
+      "step": 8244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586354067537561e-05,
+      "loss": 0.5789,
+      "step": 8245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5862963924214897e-05,
+      "loss": 0.6723,
+      "step": 8246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5862387173054188e-05,
+      "loss": 0.5249,
+      "step": 8247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5861810421893476e-05,
+      "loss": 0.6908,
+      "step": 8248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5861233670732763e-05,
+      "loss": 0.4738,
+      "step": 8249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5860656919572054e-05,
+      "loss": 0.6602,
+      "step": 8250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.586008016841134e-05,
+      "loss": 0.5618,
+      "step": 8251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5859503417250626e-05,
+      "loss": 0.3682,
+      "step": 8252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5858926666089917e-05,
+      "loss": 0.6322,
+      "step": 8253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5858349914929205e-05,
+      "loss": 0.4817,
+      "step": 8254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5857773163768493e-05,
+      "loss": 0.4966,
+      "step": 8255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5857196412607784e-05,
+      "loss": 0.551,
+      "step": 8256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.585661966144707e-05,
+      "loss": 0.4003,
+      "step": 8257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.585604291028636e-05,
+      "loss": 0.6275,
+      "step": 8258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5855466159125647e-05,
+      "loss": 0.5121,
+      "step": 8259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5854889407964935e-05,
+      "loss": 0.6744,
+      "step": 8260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5854312656804222e-05,
+      "loss": 0.4662,
+      "step": 8261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5853735905643513e-05,
+      "loss": 0.591,
+      "step": 8262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.58531591544828e-05,
+      "loss": 0.4992,
+      "step": 8263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.585258240332209e-05,
+      "loss": 0.5323,
+      "step": 8264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5852005652161376e-05,
+      "loss": 0.4664,
+      "step": 8265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5851428901000664e-05,
+      "loss": 0.5626,
+      "step": 8266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5850852149839952e-05,
+      "loss": 0.4191,
+      "step": 8267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.585027539867924e-05,
+      "loss": 0.5904,
+      "step": 8268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584969864751853e-05,
+      "loss": 0.6146,
+      "step": 8269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5849121896357818e-05,
+      "loss": 0.6708,
+      "step": 8270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5848545145197106e-05,
+      "loss": 0.5707,
+      "step": 8271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5847968394036394e-05,
+      "loss": 0.6727,
+      "step": 8272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584739164287568e-05,
+      "loss": 0.4921,
+      "step": 8273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584681489171497e-05,
+      "loss": 0.562,
+      "step": 8274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584623814055426e-05,
+      "loss": 0.5723,
+      "step": 8275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5845661389393548e-05,
+      "loss": 0.5787,
+      "step": 8276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5845084638232835e-05,
+      "loss": 0.5496,
+      "step": 8277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5844507887072123e-05,
+      "loss": 0.5336,
+      "step": 8278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584393113591141e-05,
+      "loss": 0.4712,
+      "step": 8279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.58433543847507e-05,
+      "loss": 0.5644,
+      "step": 8280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.584277763358999e-05,
+      "loss": 0.5187,
+      "step": 8281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5842200882429277e-05,
+      "loss": 0.4651,
+      "step": 8282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5841624131268565e-05,
+      "loss": 0.6012,
+      "step": 8283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5841047380107856e-05,
+      "loss": 0.4434,
+      "step": 8284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5840470628947144e-05,
+      "loss": 0.5905,
+      "step": 8285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583989387778643e-05,
+      "loss": 0.4403,
+      "step": 8286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583931712662572e-05,
+      "loss": 0.667,
+      "step": 8287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5838740375465007e-05,
+      "loss": 0.5311,
+      "step": 8288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5838163624304294e-05,
+      "loss": 0.6504,
+      "step": 8289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5837586873143586e-05,
+      "loss": 0.5238,
+      "step": 8290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5837010121982873e-05,
+      "loss": 0.49,
+      "step": 8291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583643337082216e-05,
+      "loss": 0.4182,
+      "step": 8292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583585661966145e-05,
+      "loss": 0.4081,
+      "step": 8293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5835279868500736e-05,
+      "loss": 0.5583,
+      "step": 8294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5834703117340024e-05,
+      "loss": 0.6485,
+      "step": 8295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5834126366179315e-05,
+      "loss": 0.488,
+      "step": 8296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5833549615018603e-05,
+      "loss": 0.5672,
+      "step": 8297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583297286385789e-05,
+      "loss": 0.4867,
+      "step": 8298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5832396112697178e-05,
+      "loss": 0.544,
+      "step": 8299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5831819361536466e-05,
+      "loss": 0.6364,
+      "step": 8300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5831242610375753e-05,
+      "loss": 0.4464,
+      "step": 8301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.583066585921504e-05,
+      "loss": 0.5541,
+      "step": 8302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5830089108054332e-05,
+      "loss": 0.5944,
+      "step": 8303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.582951235689362e-05,
+      "loss": 0.5695,
+      "step": 8304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5828935605732908e-05,
+      "loss": 0.577,
+      "step": 8305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5828358854572195e-05,
+      "loss": 0.5222,
+      "step": 8306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5827782103411483e-05,
+      "loss": 0.5501,
+      "step": 8307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.582720535225077e-05,
+      "loss": 0.4062,
+      "step": 8308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5826628601090062e-05,
+      "loss": 0.5299,
+      "step": 8309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.582605184992935e-05,
+      "loss": 0.506,
+      "step": 8310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.5825475098768637e-05,
+      "loss": 0.5422,
+      "step": 8311
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5824898347607928e-05,
+      "loss": 0.6275,
+      "step": 8312
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5824321596447212e-05,
+      "loss": 0.4577,
+      "step": 8313
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.58237448452865e-05,
+      "loss": 0.683,
+      "step": 8314
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.582316809412579e-05,
+      "loss": 0.5252,
+      "step": 8315
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.582259134296508e-05,
+      "loss": 0.5128,
+      "step": 8316
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5822014591804367e-05,
+      "loss": 0.6492,
+      "step": 8317
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5821437840643658e-05,
+      "loss": 0.5631,
+      "step": 8318
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5820861089482945e-05,
+      "loss": 0.5386,
+      "step": 8319
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5820284338322233e-05,
+      "loss": 0.5126,
+      "step": 8320
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.581970758716152e-05,
+      "loss": 0.4834,
+      "step": 8321
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.581913083600081e-05,
+      "loss": 0.5826,
+      "step": 8322
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5818554084840096e-05,
+      "loss": 0.5384,
+      "step": 8323
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5817977333679387e-05,
+      "loss": 0.5071,
+      "step": 8324
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5817400582518675e-05,
+      "loss": 0.4523,
+      "step": 8325
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5816823831357963e-05,
+      "loss": 0.624,
+      "step": 8326
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.581624708019725e-05,
+      "loss": 0.631,
+      "step": 8327
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5815670329036538e-05,
+      "loss": 0.5184,
+      "step": 8328
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5815093577875826e-05,
+      "loss": 0.4414,
+      "step": 8329
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5814516826715117e-05,
+      "loss": 0.4996,
+      "step": 8330
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5813940075554404e-05,
+      "loss": 0.5988,
+      "step": 8331
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5813363324393692e-05,
+      "loss": 0.5636,
+      "step": 8332
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.581278657323298e-05,
+      "loss": 0.6062,
+      "step": 8333
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5812209822072267e-05,
+      "loss": 0.5148,
+      "step": 8334
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5811633070911555e-05,
+      "loss": 0.613,
+      "step": 8335
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5811056319750843e-05,
+      "loss": 0.4233,
+      "step": 8336
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5810479568590134e-05,
+      "loss": 0.5673,
+      "step": 8337
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580990281742942e-05,
+      "loss": 0.5295,
+      "step": 8338
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580932606626871e-05,
+      "loss": 0.6684,
+      "step": 8339
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5808749315107997e-05,
+      "loss": 0.5931,
+      "step": 8340
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5808172563947285e-05,
+      "loss": 0.419,
+      "step": 8341
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5807595812786572e-05,
+      "loss": 0.4782,
+      "step": 8342
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5807019061625863e-05,
+      "loss": 0.5247,
+      "step": 8343
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580644231046515e-05,
+      "loss": 0.3892,
+      "step": 8344
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580586555930444e-05,
+      "loss": 0.5761,
+      "step": 8345
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580528880814373e-05,
+      "loss": 0.5862,
+      "step": 8346
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5804712056983017e-05,
+      "loss": 0.6592,
+      "step": 8347
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5804135305822305e-05,
+      "loss": 0.4541,
+      "step": 8348
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5803558554661593e-05,
+      "loss": 0.4943,
+      "step": 8349
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580298180350088e-05,
+      "loss": 0.4233,
+      "step": 8350
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5802405052340168e-05,
+      "loss": 0.5617,
+      "step": 8351
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.580182830117946e-05,
+      "loss": 0.5064,
+      "step": 8352
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5801251550018747e-05,
+      "loss": 0.4847,
+      "step": 8353
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5800674798858035e-05,
+      "loss": 0.5187,
+      "step": 8354
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5800098047697322e-05,
+      "loss": 0.643,
+      "step": 8355
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.579952129653661e-05,
+      "loss": 0.6818,
+      "step": 8356
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5798944545375898e-05,
+      "loss": 0.5999,
+      "step": 8357
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.579836779421519e-05,
+      "loss": 0.5189,
+      "step": 8358
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5797791043054476e-05,
+      "loss": 0.6006,
+      "step": 8359
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5797214291893764e-05,
+      "loss": 0.699,
+      "step": 8360
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5796637540733052e-05,
+      "loss": 0.4756,
+      "step": 8361
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.579606078957234e-05,
+      "loss": 0.472,
+      "step": 8362
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5795484038411627e-05,
+      "loss": 0.6362,
+      "step": 8363
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5794907287250918e-05,
+      "loss": 0.5296,
+      "step": 8364
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5794330536090206e-05,
+      "loss": 0.4196,
+      "step": 8365
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5793753784929494e-05,
+      "loss": 0.5337,
+      "step": 8366
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.579317703376878e-05,
+      "loss": 0.5933,
+      "step": 8367
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.579260028260807e-05,
+      "loss": 0.6133,
+      "step": 8368
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5792023531447357e-05,
+      "loss": 0.5219,
+      "step": 8369
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5791446780286644e-05,
+      "loss": 0.4813,
+      "step": 8370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5790870029125935e-05,
+      "loss": 0.5522,
+      "step": 8371
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5790293277965223e-05,
+      "loss": 0.6071,
+      "step": 8372
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578971652680451e-05,
+      "loss": 0.4421,
+      "step": 8373
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5789139775643802e-05,
+      "loss": 0.6016,
+      "step": 8374
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5788563024483086e-05,
+      "loss": 0.5805,
+      "step": 8375
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5787986273322374e-05,
+      "loss": 0.5194,
+      "step": 8376
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5787409522161665e-05,
+      "loss": 0.4647,
+      "step": 8377
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5786832771000953e-05,
+      "loss": 0.5203,
+      "step": 8378
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578625601984024e-05,
+      "loss": 0.5215,
+      "step": 8379
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578567926867953e-05,
+      "loss": 0.5246,
+      "step": 8380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578510251751882e-05,
+      "loss": 0.5949,
+      "step": 8381
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5784525766358107e-05,
+      "loss": 0.6063,
+      "step": 8382
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5783949015197394e-05,
+      "loss": 0.5929,
+      "step": 8383
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5783372264036682e-05,
+      "loss": 0.5695,
+      "step": 8384
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578279551287597e-05,
+      "loss": 0.5813,
+      "step": 8385
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578221876171526e-05,
+      "loss": 0.5723,
+      "step": 8386
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.578164201055455e-05,
+      "loss": 0.5968,
+      "step": 8387
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5781065259393836e-05,
+      "loss": 0.5115,
+      "step": 8388
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5780488508233124e-05,
+      "loss": 0.5997,
+      "step": 8389
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577991175707241e-05,
+      "loss": 0.5539,
+      "step": 8390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.57793350059117e-05,
+      "loss": 0.5349,
+      "step": 8391
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577875825475099e-05,
+      "loss": 0.6033,
+      "step": 8392
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5778181503590278e-05,
+      "loss": 0.4814,
+      "step": 8393
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5777604752429566e-05,
+      "loss": 0.5314,
+      "step": 8394
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5777028001268853e-05,
+      "loss": 0.6071,
+      "step": 8395
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577645125010814e-05,
+      "loss": 0.4903,
+      "step": 8396
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577587449894743e-05,
+      "loss": 0.509,
+      "step": 8397
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577529774778672e-05,
+      "loss": 0.4788,
+      "step": 8398
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5774720996626008e-05,
+      "loss": 0.568,
+      "step": 8399
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5774144245465295e-05,
+      "loss": 0.4546,
+      "step": 8400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5773567494304586e-05,
+      "loss": 0.556,
+      "step": 8401
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577299074314387e-05,
+      "loss": 0.5593,
+      "step": 8402
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.577241399198316e-05,
+      "loss": 0.6351,
+      "step": 8403
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5771837240822446e-05,
+      "loss": 0.7197,
+      "step": 8404
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5771260489661737e-05,
+      "loss": 0.5263,
+      "step": 8405
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5770683738501025e-05,
+      "loss": 0.4953,
+      "step": 8406
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5770106987340312e-05,
+      "loss": 0.5724,
+      "step": 8407
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5769530236179604e-05,
+      "loss": 0.4393,
+      "step": 8408
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576895348501889e-05,
+      "loss": 0.4503,
+      "step": 8409
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576837673385818e-05,
+      "loss": 0.5724,
+      "step": 8410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5767799982697467e-05,
+      "loss": 0.5483,
+      "step": 8411
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5767223231536754e-05,
+      "loss": 0.6664,
+      "step": 8412
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5766646480376042e-05,
+      "loss": 0.4924,
+      "step": 8413
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5766069729215333e-05,
+      "loss": 0.3822,
+      "step": 8414
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576549297805462e-05,
+      "loss": 0.52,
+      "step": 8415
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576491622689391e-05,
+      "loss": 0.5942,
+      "step": 8416
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5764339475733196e-05,
+      "loss": 0.5905,
+      "step": 8417
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5763762724572484e-05,
+      "loss": 0.532,
+      "step": 8418
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576318597341177e-05,
+      "loss": 0.508,
+      "step": 8419
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5762609222251063e-05,
+      "loss": 0.5838,
+      "step": 8420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.576203247109035e-05,
+      "loss": 0.5991,
+      "step": 8421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5761455719929638e-05,
+      "loss": 0.6099,
+      "step": 8422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5760878968768926e-05,
+      "loss": 0.5545,
+      "step": 8423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5760302217608213e-05,
+      "loss": 0.521,
+      "step": 8424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.57597254664475e-05,
+      "loss": 0.5104,
+      "step": 8425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5759148715286792e-05,
+      "loss": 0.6336,
+      "step": 8426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575857196412608e-05,
+      "loss": 0.5504,
+      "step": 8427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5757995212965367e-05,
+      "loss": 0.6168,
+      "step": 8428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5757418461804655e-05,
+      "loss": 0.5505,
+      "step": 8429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5756841710643943e-05,
+      "loss": 0.5905,
+      "step": 8430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575626495948323e-05,
+      "loss": 0.543,
+      "step": 8431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575568820832252e-05,
+      "loss": 0.6913,
+      "step": 8432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575511145716181e-05,
+      "loss": 0.5085,
+      "step": 8433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5754534706001097e-05,
+      "loss": 0.5617,
+      "step": 8434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5753957954840388e-05,
+      "loss": 0.6244,
+      "step": 8435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5753381203679676e-05,
+      "loss": 0.4515,
+      "step": 8436
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575280445251896e-05,
+      "loss": 0.5557,
+      "step": 8437
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5752227701358248e-05,
+      "loss": 0.4848,
+      "step": 8438
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.575165095019754e-05,
+      "loss": 0.4351,
+      "step": 8439
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5751074199036826e-05,
+      "loss": 0.5652,
+      "step": 8440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5750497447876114e-05,
+      "loss": 0.5416,
+      "step": 8441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5749920696715405e-05,
+      "loss": 0.6485,
+      "step": 8442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5749343945554693e-05,
+      "loss": 0.6952,
+      "step": 8443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.574876719439398e-05,
+      "loss": 0.6201,
+      "step": 8444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5748190443233268e-05,
+      "loss": 0.4712,
+      "step": 8445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5747613692072556e-05,
+      "loss": 0.4545,
+      "step": 8446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5747036940911844e-05,
+      "loss": 0.5254,
+      "step": 8447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5746460189751135e-05,
+      "loss": 0.4382,
+      "step": 8448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5745883438590422e-05,
+      "loss": 0.5019,
+      "step": 8449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.574530668742971e-05,
+      "loss": 0.5769,
+      "step": 8450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5744729936268998e-05,
+      "loss": 0.5209,
+      "step": 8451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5744153185108285e-05,
+      "loss": 0.3559,
+      "step": 8452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5743576433947573e-05,
+      "loss": 0.4669,
+      "step": 8453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5742999682786864e-05,
+      "loss": 0.5084,
+      "step": 8454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5742422931626152e-05,
+      "loss": 0.5526,
+      "step": 8455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.574184618046544e-05,
+      "loss": 0.4842,
+      "step": 8456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5741269429304727e-05,
+      "loss": 0.568,
+      "step": 8457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5740692678144015e-05,
+      "loss": 0.5148,
+      "step": 8458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5740115926983303e-05,
+      "loss": 0.4798,
+      "step": 8459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5739539175822594e-05,
+      "loss": 0.5011,
+      "step": 8460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573896242466188e-05,
+      "loss": 0.4559,
+      "step": 8461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573838567350117e-05,
+      "loss": 0.4064,
+      "step": 8462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573780892234046e-05,
+      "loss": 0.5721,
+      "step": 8463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5737232171179744e-05,
+      "loss": 0.4927,
+      "step": 8464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5736655420019032e-05,
+      "loss": 0.5958,
+      "step": 8465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5736078668858323e-05,
+      "loss": 0.4699,
+      "step": 8466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573550191769761e-05,
+      "loss": 0.5807,
+      "step": 8467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.57349251665369e-05,
+      "loss": 0.5642,
+      "step": 8468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573434841537619e-05,
+      "loss": 0.5876,
+      "step": 8469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5733771664215477e-05,
+      "loss": 0.6805,
+      "step": 8470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5733194913054765e-05,
+      "loss": 0.6481,
+      "step": 8471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5732618161894053e-05,
+      "loss": 0.4549,
+      "step": 8472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.573204141073334e-05,
+      "loss": 0.6447,
+      "step": 8473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5731464659572628e-05,
+      "loss": 0.6646,
+      "step": 8474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5730887908411916e-05,
+      "loss": 0.6092,
+      "step": 8475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5730311157251207e-05,
+      "loss": 0.669,
+      "step": 8476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5729734406090494e-05,
+      "loss": 0.4819,
+      "step": 8477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5729157654929782e-05,
+      "loss": 0.473,
+      "step": 8478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.572858090376907e-05,
+      "loss": 0.5724,
+      "step": 8479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5728004152608358e-05,
+      "loss": 0.5512,
+      "step": 8480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5727427401447645e-05,
+      "loss": 0.5654,
+      "step": 8481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5726850650286936e-05,
+      "loss": 0.6155,
+      "step": 8482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5726273899126224e-05,
+      "loss": 0.5601,
+      "step": 8483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.572569714796551e-05,
+      "loss": 0.6043,
+      "step": 8484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.57251203968048e-05,
+      "loss": 0.426,
+      "step": 8485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5724543645644087e-05,
+      "loss": 0.4471,
+      "step": 8486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5723966894483375e-05,
+      "loss": 0.6164,
+      "step": 8487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5723390143322666e-05,
+      "loss": 0.4026,
+      "step": 8488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.5722813392161953e-05,
+      "loss": 0.5778,
+      "step": 8489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.572223664100124e-05,
+      "loss": 0.5686,
+      "step": 8490
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.572165988984053e-05,
+      "loss": 0.5191,
+      "step": 8491
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5721083138679817e-05,
+      "loss": 0.5972,
+      "step": 8492
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5720506387519104e-05,
+      "loss": 0.5286,
+      "step": 8493
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5719929636358395e-05,
+      "loss": 0.6669,
+      "step": 8494
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5719352885197683e-05,
+      "loss": 0.3898,
+      "step": 8495
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571877613403697e-05,
+      "loss": 0.5447,
+      "step": 8496
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5718199382876262e-05,
+      "loss": 0.5177,
+      "step": 8497
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571762263171555e-05,
+      "loss": 0.4964,
+      "step": 8498
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5717045880554834e-05,
+      "loss": 0.69,
+      "step": 8499
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5716469129394125e-05,
+      "loss": 0.5316,
+      "step": 8500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5715892378233412e-05,
+      "loss": 0.5633,
+      "step": 8501
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.57153156270727e-05,
+      "loss": 0.5141,
+      "step": 8502
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571473887591199e-05,
+      "loss": 0.5079,
+      "step": 8503
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571416212475128e-05,
+      "loss": 0.4631,
+      "step": 8504
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5713585373590567e-05,
+      "loss": 0.4584,
+      "step": 8505
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5713008622429854e-05,
+      "loss": 0.6805,
+      "step": 8506
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5712431871269142e-05,
+      "loss": 0.4982,
+      "step": 8507
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571185512010843e-05,
+      "loss": 0.4565,
+      "step": 8508
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5711278368947717e-05,
+      "loss": 0.4247,
+      "step": 8509
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.571070161778701e-05,
+      "loss": 0.5495,
+      "step": 8510
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5710124866626296e-05,
+      "loss": 0.5651,
+      "step": 8511
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5709548115465584e-05,
+      "loss": 0.5361,
+      "step": 8512
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.570897136430487e-05,
+      "loss": 0.5037,
+      "step": 8513
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.570839461314416e-05,
+      "loss": 0.5004,
+      "step": 8514
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5707817861983447e-05,
+      "loss": 0.6227,
+      "step": 8515
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5707241110822738e-05,
+      "loss": 0.5385,
+      "step": 8516
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5706664359662026e-05,
+      "loss": 0.5037,
+      "step": 8517
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5706087608501313e-05,
+      "loss": 0.4554,
+      "step": 8518
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.57055108573406e-05,
+      "loss": 0.5087,
+      "step": 8519
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.570493410617989e-05,
+      "loss": 0.5215,
+      "step": 8520
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5704357355019176e-05,
+      "loss": 0.4379,
+      "step": 8521
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5703780603858467e-05,
+      "loss": 0.5193,
+      "step": 8522
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5703203852697755e-05,
+      "loss": 0.5262,
+      "step": 8523
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5702627101537043e-05,
+      "loss": 0.538,
+      "step": 8524
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5702050350376334e-05,
+      "loss": 0.6029,
+      "step": 8525
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5701473599215618e-05,
+      "loss": 0.5566,
+      "step": 8526
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5700896848054906e-05,
+      "loss": 0.7241,
+      "step": 8527
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5700320096894197e-05,
+      "loss": 0.5549,
+      "step": 8528
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5699743345733485e-05,
+      "loss": 0.4893,
+      "step": 8529
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5699166594572772e-05,
+      "loss": 0.488,
+      "step": 8530
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5698589843412063e-05,
+      "loss": 0.4775,
+      "step": 8531
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569801309225135e-05,
+      "loss": 0.6273,
+      "step": 8532
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569743634109064e-05,
+      "loss": 0.658,
+      "step": 8533
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5696859589929926e-05,
+      "loss": 0.609,
+      "step": 8534
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5696282838769214e-05,
+      "loss": 0.5696,
+      "step": 8535
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5695706087608502e-05,
+      "loss": 0.6056,
+      "step": 8536
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5695129336447793e-05,
+      "loss": 0.4578,
+      "step": 8537
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569455258528708e-05,
+      "loss": 0.5897,
+      "step": 8538
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5693975834126368e-05,
+      "loss": 0.38,
+      "step": 8539
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5693399082965656e-05,
+      "loss": 0.6408,
+      "step": 8540
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5692822331804944e-05,
+      "loss": 0.4887,
+      "step": 8541
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569224558064423e-05,
+      "loss": 0.6939,
+      "step": 8542
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569166882948352e-05,
+      "loss": 0.5857,
+      "step": 8543
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.569109207832281e-05,
+      "loss": 0.5939,
+      "step": 8544
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5690515327162098e-05,
+      "loss": 0.4485,
+      "step": 8545
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5689938576001385e-05,
+      "loss": 0.5397,
+      "step": 8546
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5689361824840673e-05,
+      "loss": 0.4623,
+      "step": 8547
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568878507367996e-05,
+      "loss": 0.5058,
+      "step": 8548
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568820832251925e-05,
+      "loss": 0.5399,
+      "step": 8549
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568763157135854e-05,
+      "loss": 0.5618,
+      "step": 8550
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5687054820197827e-05,
+      "loss": 0.5612,
+      "step": 8551
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5686478069037115e-05,
+      "loss": 0.5124,
+      "step": 8552
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5685901317876403e-05,
+      "loss": 0.6072,
+      "step": 8553
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568532456671569e-05,
+      "loss": 0.5404,
+      "step": 8554
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5684747815554978e-05,
+      "loss": 0.5716,
+      "step": 8555
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568417106439427e-05,
+      "loss": 0.4442,
+      "step": 8556
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5683594313233557e-05,
+      "loss": 0.6082,
+      "step": 8557
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5683017562072844e-05,
+      "loss": 0.4715,
+      "step": 8558
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5682440810912135e-05,
+      "loss": 0.6506,
+      "step": 8559
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5681864059751423e-05,
+      "loss": 0.5932,
+      "step": 8560
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5681287308590707e-05,
+      "loss": 0.4846,
+      "step": 8561
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.568071055743e-05,
+      "loss": 0.5511,
+      "step": 8562
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5680133806269286e-05,
+      "loss": 0.5453,
+      "step": 8563
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5679557055108574e-05,
+      "loss": 0.4955,
+      "step": 8564
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5678980303947865e-05,
+      "loss": 0.5686,
+      "step": 8565
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5678403552787153e-05,
+      "loss": 0.3903,
+      "step": 8566
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.567782680162644e-05,
+      "loss": 0.5194,
+      "step": 8567
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5677250050465728e-05,
+      "loss": 0.4415,
+      "step": 8568
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5676673299305016e-05,
+      "loss": 0.5173,
+      "step": 8569
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5676096548144303e-05,
+      "loss": 0.4767,
+      "step": 8570
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5675519796983594e-05,
+      "loss": 0.5903,
+      "step": 8571
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5674943045822882e-05,
+      "loss": 0.6191,
+      "step": 8572
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.567436629466217e-05,
+      "loss": 0.5313,
+      "step": 8573
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5673789543501458e-05,
+      "loss": 0.3619,
+      "step": 8574
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5673212792340745e-05,
+      "loss": 0.6258,
+      "step": 8575
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5672636041180033e-05,
+      "loss": 0.4528,
+      "step": 8576
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.567205929001932e-05,
+      "loss": 0.5126,
+      "step": 8577
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.567148253885861e-05,
+      "loss": 0.6348,
+      "step": 8578
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.56709057876979e-05,
+      "loss": 0.6584,
+      "step": 8579
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5670329036537187e-05,
+      "loss": 0.5765,
+      "step": 8580
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5669752285376475e-05,
+      "loss": 0.637,
+      "step": 8581
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5669175534215762e-05,
+      "loss": 0.3416,
+      "step": 8582
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566859878305505e-05,
+      "loss": 0.5002,
+      "step": 8583
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566802203189434e-05,
+      "loss": 0.4407,
+      "step": 8584
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566744528073363e-05,
+      "loss": 0.4663,
+      "step": 8585
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5666868529572917e-05,
+      "loss": 0.4665,
+      "step": 8586
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5666291778412208e-05,
+      "loss": 0.563,
+      "step": 8587
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5665715027251492e-05,
+      "loss": 0.4968,
+      "step": 8588
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566513827609078e-05,
+      "loss": 0.5569,
+      "step": 8589
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566456152493007e-05,
+      "loss": 0.5797,
+      "step": 8590
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.566398477376936e-05,
+      "loss": 0.5047,
+      "step": 8591
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5663408022608646e-05,
+      "loss": 0.4737,
+      "step": 8592
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5662831271447937e-05,
+      "loss": 0.4767,
+      "step": 8593
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5662254520287225e-05,
+      "loss": 0.5442,
+      "step": 8594
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5661677769126512e-05,
+      "loss": 0.605,
+      "step": 8595
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.56611010179658e-05,
+      "loss": 0.4613,
+      "step": 8596
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5660524266805088e-05,
+      "loss": 0.567,
+      "step": 8597
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5659947515644376e-05,
+      "loss": 0.4971,
+      "step": 8598
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5659370764483667e-05,
+      "loss": 0.4955,
+      "step": 8599
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5658794013322954e-05,
+      "loss": 0.358,
+      "step": 8600
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5658217262162242e-05,
+      "loss": 0.601,
+      "step": 8601
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.565764051100153e-05,
+      "loss": 0.473,
+      "step": 8602
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5657063759840817e-05,
+      "loss": 0.5136,
+      "step": 8603
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5656487008680105e-05,
+      "loss": 0.3849,
+      "step": 8604
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5655910257519396e-05,
+      "loss": 0.4834,
+      "step": 8605
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5655333506358684e-05,
+      "loss": 0.7052,
+      "step": 8606
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.565475675519797e-05,
+      "loss": 0.6619,
+      "step": 8607
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.565418000403726e-05,
+      "loss": 0.5007,
+      "step": 8608
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5653603252876547e-05,
+      "loss": 0.5047,
+      "step": 8609
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5653026501715835e-05,
+      "loss": 0.5407,
+      "step": 8610
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5652449750555122e-05,
+      "loss": 0.6016,
+      "step": 8611
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5651872999394413e-05,
+      "loss": 0.361,
+      "step": 8612
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.56512962482337e-05,
+      "loss": 0.6467,
+      "step": 8613
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.565071949707299e-05,
+      "loss": 0.6367,
+      "step": 8614
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5650142745912276e-05,
+      "loss": 0.5291,
+      "step": 8615
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5649565994751564e-05,
+      "loss": 0.4319,
+      "step": 8616
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5648989243590852e-05,
+      "loss": 0.5796,
+      "step": 8617
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5648412492430143e-05,
+      "loss": 0.5367,
+      "step": 8618
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564783574126943e-05,
+      "loss": 0.5505,
+      "step": 8619
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5647258990108718e-05,
+      "loss": 0.434,
+      "step": 8620
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564668223894801e-05,
+      "loss": 0.4972,
+      "step": 8621
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5646105487787297e-05,
+      "loss": 0.6071,
+      "step": 8622
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564552873662658e-05,
+      "loss": 0.4649,
+      "step": 8623
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5644951985465872e-05,
+      "loss": 0.5096,
+      "step": 8624
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564437523430516e-05,
+      "loss": 0.6,
+      "step": 8625
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5643798483144448e-05,
+      "loss": 0.5528,
+      "step": 8626
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564322173198374e-05,
+      "loss": 0.4884,
+      "step": 8627
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5642644980823026e-05,
+      "loss": 0.4669,
+      "step": 8628
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5642068229662314e-05,
+      "loss": 0.6756,
+      "step": 8629
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5641491478501602e-05,
+      "loss": 0.4378,
+      "step": 8630
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.564091472734089e-05,
+      "loss": 0.606,
+      "step": 8631
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5640337976180177e-05,
+      "loss": 0.4836,
+      "step": 8632
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5639761225019468e-05,
+      "loss": 0.5404,
+      "step": 8633
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5639184473858756e-05,
+      "loss": 0.6502,
+      "step": 8634
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5638607722698044e-05,
+      "loss": 0.5334,
+      "step": 8635
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563803097153733e-05,
+      "loss": 0.4661,
+      "step": 8636
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563745422037662e-05,
+      "loss": 0.5438,
+      "step": 8637
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5636877469215907e-05,
+      "loss": 0.4284,
+      "step": 8638
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5636300718055198e-05,
+      "loss": 0.5202,
+      "step": 8639
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5635723966894485e-05,
+      "loss": 0.5158,
+      "step": 8640
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5635147215733773e-05,
+      "loss": 0.487,
+      "step": 8641
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563457046457306e-05,
+      "loss": 0.6023,
+      "step": 8642
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563399371341235e-05,
+      "loss": 0.5538,
+      "step": 8643
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5633416962251636e-05,
+      "loss": 0.5336,
+      "step": 8644
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5632840211090927e-05,
+      "loss": 0.6181,
+      "step": 8645
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5632263459930215e-05,
+      "loss": 0.5808,
+      "step": 8646
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5631686708769503e-05,
+      "loss": 0.5798,
+      "step": 8647
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563110995760879e-05,
+      "loss": 0.4957,
+      "step": 8648
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.563053320644808e-05,
+      "loss": 0.5171,
+      "step": 8649
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5629956455287366e-05,
+      "loss": 0.5223,
+      "step": 8650
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5629379704126653e-05,
+      "loss": 0.5808,
+      "step": 8651
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5628802952965944e-05,
+      "loss": 0.5661,
+      "step": 8652
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5628226201805232e-05,
+      "loss": 0.4398,
+      "step": 8653
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562764945064452e-05,
+      "loss": 0.5447,
+      "step": 8654
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562707269948381e-05,
+      "loss": 0.4837,
+      "step": 8655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.56264959483231e-05,
+      "loss": 0.4604,
+      "step": 8656
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5625919197162386e-05,
+      "loss": 0.5221,
+      "step": 8657
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5625342446001674e-05,
+      "loss": 0.506,
+      "step": 8658
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562476569484096e-05,
+      "loss": 0.406,
+      "step": 8659
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562418894368025e-05,
+      "loss": 0.5014,
+      "step": 8660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562361219251954e-05,
+      "loss": 0.5377,
+      "step": 8661
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5623035441358828e-05,
+      "loss": 0.7092,
+      "step": 8662
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5622458690198116e-05,
+      "loss": 0.4641,
+      "step": 8663
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5621881939037403e-05,
+      "loss": 0.5049,
+      "step": 8664
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562130518787669e-05,
+      "loss": 0.4462,
+      "step": 8665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562072843671598e-05,
+      "loss": 0.5294,
+      "step": 8666
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.562015168555527e-05,
+      "loss": 0.5266,
+      "step": 8667
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5619574934394558e-05,
+      "loss": 0.5072,
+      "step": 8668
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.5618998183233845e-05,
+      "loss": 0.4349,
+      "step": 8669
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5618421432073133e-05,
+      "loss": 0.5948,
+      "step": 8670
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.561784468091242e-05,
+      "loss": 0.4808,
+      "step": 8671
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5617267929751708e-05,
+      "loss": 0.4814,
+      "step": 8672
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5616691178591e-05,
+      "loss": 0.5248,
+      "step": 8673
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5616114427430287e-05,
+      "loss": 0.5697,
+      "step": 8674
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5615537676269575e-05,
+      "loss": 0.5403,
+      "step": 8675
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5614960925108862e-05,
+      "loss": 0.6097,
+      "step": 8676
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.561438417394815e-05,
+      "loss": 0.6353,
+      "step": 8677
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5613807422787438e-05,
+      "loss": 0.4681,
+      "step": 8678
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.561323067162673e-05,
+      "loss": 0.516,
+      "step": 8679
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5612653920466017e-05,
+      "loss": 0.5593,
+      "step": 8680
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5612077169305304e-05,
+      "loss": 0.4908,
+      "step": 8681
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5611500418144592e-05,
+      "loss": 0.5009,
+      "step": 8682
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5610923666983883e-05,
+      "loss": 0.4653,
+      "step": 8683
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.561034691582317e-05,
+      "loss": 0.5612,
+      "step": 8684
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560977016466246e-05,
+      "loss": 0.5734,
+      "step": 8685
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5609193413501746e-05,
+      "loss": 0.5692,
+      "step": 8686
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5608616662341034e-05,
+      "loss": 0.5476,
+      "step": 8687
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560803991118032e-05,
+      "loss": 0.414,
+      "step": 8688
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5607463160019612e-05,
+      "loss": 0.5986,
+      "step": 8689
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.56068864088589e-05,
+      "loss": 0.5678,
+      "step": 8690
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5606309657698188e-05,
+      "loss": 0.4881,
+      "step": 8691
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5605732906537476e-05,
+      "loss": 0.5894,
+      "step": 8692
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5605156155376763e-05,
+      "loss": 0.7676,
+      "step": 8693
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560457940421605e-05,
+      "loss": 0.5829,
+      "step": 8694
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5604002653055342e-05,
+      "loss": 0.3965,
+      "step": 8695
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560342590189463e-05,
+      "loss": 0.5083,
+      "step": 8696
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5602849150733917e-05,
+      "loss": 0.562,
+      "step": 8697
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5602272399573205e-05,
+      "loss": 0.5893,
+      "step": 8698
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5601695648412493e-05,
+      "loss": 0.5909,
+      "step": 8699
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560111889725178e-05,
+      "loss": 0.4479,
+      "step": 8700
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.560054214609107e-05,
+      "loss": 0.6376,
+      "step": 8701
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559996539493036e-05,
+      "loss": 0.5309,
+      "step": 8702
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5599388643769647e-05,
+      "loss": 0.4637,
+      "step": 8703
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5598811892608935e-05,
+      "loss": 0.5444,
+      "step": 8704
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5598235141448222e-05,
+      "loss": 0.5685,
+      "step": 8705
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559765839028751e-05,
+      "loss": 0.5706,
+      "step": 8706
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.55970816391268e-05,
+      "loss": 0.4903,
+      "step": 8707
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559650488796609e-05,
+      "loss": 0.7356,
+      "step": 8708
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5595928136805376e-05,
+      "loss": 0.5506,
+      "step": 8709
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5595351385644667e-05,
+      "loss": 0.4973,
+      "step": 8710
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5594774634483955e-05,
+      "loss": 0.463,
+      "step": 8711
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559419788332324e-05,
+      "loss": 0.5698,
+      "step": 8712
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559362113216253e-05,
+      "loss": 0.4581,
+      "step": 8713
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5593044381001818e-05,
+      "loss": 0.6366,
+      "step": 8714
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5592467629841106e-05,
+      "loss": 0.5867,
+      "step": 8715
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5591890878680394e-05,
+      "loss": 0.54,
+      "step": 8716
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5591314127519685e-05,
+      "loss": 0.4997,
+      "step": 8717
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5590737376358972e-05,
+      "loss": 0.548,
+      "step": 8718
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.559016062519826e-05,
+      "loss": 0.442,
+      "step": 8719
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5589583874037548e-05,
+      "loss": 0.5542,
+      "step": 8720
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5589007122876835e-05,
+      "loss": 0.609,
+      "step": 8721
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5588430371716123e-05,
+      "loss": 0.6637,
+      "step": 8722
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5587853620555414e-05,
+      "loss": 0.5694,
+      "step": 8723
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5587276869394702e-05,
+      "loss": 0.5871,
+      "step": 8724
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.558670011823399e-05,
+      "loss": 0.5359,
+      "step": 8725
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5586123367073277e-05,
+      "loss": 0.5781,
+      "step": 8726
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5585546615912565e-05,
+      "loss": 0.573,
+      "step": 8727
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5584969864751853e-05,
+      "loss": 0.5113,
+      "step": 8728
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5584393113591144e-05,
+      "loss": 0.6246,
+      "step": 8729
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.558381636243043e-05,
+      "loss": 0.456,
+      "step": 8730
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.558323961126972e-05,
+      "loss": 0.3955,
+      "step": 8731
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5582662860109007e-05,
+      "loss": 0.3428,
+      "step": 8732
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5582086108948294e-05,
+      "loss": 0.534,
+      "step": 8733
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5581509357787582e-05,
+      "loss": 0.5212,
+      "step": 8734
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5580932606626873e-05,
+      "loss": 0.7012,
+      "step": 8735
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.558035585546616e-05,
+      "loss": 0.6187,
+      "step": 8736
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557977910430545e-05,
+      "loss": 0.5068,
+      "step": 8737
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5579202353144736e-05,
+      "loss": 0.5423,
+      "step": 8738
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5578625601984024e-05,
+      "loss": 0.6061,
+      "step": 8739
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557804885082331e-05,
+      "loss": 0.5664,
+      "step": 8740
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5577472099662603e-05,
+      "loss": 0.5897,
+      "step": 8741
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557689534850189e-05,
+      "loss": 0.6453,
+      "step": 8742
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5576318597341178e-05,
+      "loss": 0.4533,
+      "step": 8743
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557574184618047e-05,
+      "loss": 0.4302,
+      "step": 8744
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5575165095019757e-05,
+      "loss": 0.5604,
+      "step": 8745
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5574588343859044e-05,
+      "loss": 0.5839,
+      "step": 8746
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5574011592698332e-05,
+      "loss": 0.5364,
+      "step": 8747
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557343484153762e-05,
+      "loss": 0.5332,
+      "step": 8748
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5572858090376907e-05,
+      "loss": 0.5691,
+      "step": 8749
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5572281339216195e-05,
+      "loss": 0.5411,
+      "step": 8750
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5571704588055486e-05,
+      "loss": 0.5478,
+      "step": 8751
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5571127836894774e-05,
+      "loss": 0.6004,
+      "step": 8752
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.557055108573406e-05,
+      "loss": 0.5623,
+      "step": 8753
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.556997433457335e-05,
+      "loss": 0.4583,
+      "step": 8754
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5569397583412637e-05,
+      "loss": 0.5324,
+      "step": 8755
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5568820832251925e-05,
+      "loss": 0.4535,
+      "step": 8756
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5568244081091216e-05,
+      "loss": 0.4968,
+      "step": 8757
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5567667329930503e-05,
+      "loss": 0.5941,
+      "step": 8758
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.556709057876979e-05,
+      "loss": 0.4762,
+      "step": 8759
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.556651382760908e-05,
+      "loss": 0.4759,
+      "step": 8760
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5565937076448366e-05,
+      "loss": 0.5345,
+      "step": 8761
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5565360325287654e-05,
+      "loss": 0.3208,
+      "step": 8762
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5564783574126945e-05,
+      "loss": 0.5563,
+      "step": 8763
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5564206822966233e-05,
+      "loss": 0.5197,
+      "step": 8764
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.556363007180552e-05,
+      "loss": 0.55,
+      "step": 8765
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5563053320644808e-05,
+      "loss": 0.5136,
+      "step": 8766
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5562476569484096e-05,
+      "loss": 0.5886,
+      "step": 8767
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5561899818323384e-05,
+      "loss": 0.459,
+      "step": 8768
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5561323067162675e-05,
+      "loss": 0.4912,
+      "step": 8769
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5560746316001962e-05,
+      "loss": 0.6402,
+      "step": 8770
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.556016956484125e-05,
+      "loss": 0.5021,
+      "step": 8771
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555959281368054e-05,
+      "loss": 0.5117,
+      "step": 8772
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555901606251983e-05,
+      "loss": 0.4307,
+      "step": 8773
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5558439311359113e-05,
+      "loss": 0.6328,
+      "step": 8774
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5557862560198404e-05,
+      "loss": 0.4792,
+      "step": 8775
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5557285809037692e-05,
+      "loss": 0.5086,
+      "step": 8776
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555670905787698e-05,
+      "loss": 0.5652,
+      "step": 8777
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555613230671627e-05,
+      "loss": 0.5387,
+      "step": 8778
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555555555555556e-05,
+      "loss": 0.4683,
+      "step": 8779
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5554978804394846e-05,
+      "loss": 0.6369,
+      "step": 8780
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5554402053234134e-05,
+      "loss": 0.723,
+      "step": 8781
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555382530207342e-05,
+      "loss": 0.4119,
+      "step": 8782
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555324855091271e-05,
+      "loss": 0.5,
+      "step": 8783
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5552671799752e-05,
+      "loss": 0.4968,
+      "step": 8784
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5552095048591288e-05,
+      "loss": 0.4683,
+      "step": 8785
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5551518297430576e-05,
+      "loss": 0.5716,
+      "step": 8786
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5550941546269863e-05,
+      "loss": 0.5194,
+      "step": 8787
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.555036479510915e-05,
+      "loss": 0.5174,
+      "step": 8788
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.554978804394844e-05,
+      "loss": 0.4721,
+      "step": 8789
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5549211292787726e-05,
+      "loss": 0.5954,
+      "step": 8790
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5548634541627017e-05,
+      "loss": 0.6046,
+      "step": 8791
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5548057790466305e-05,
+      "loss": 0.658,
+      "step": 8792
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5547481039305593e-05,
+      "loss": 0.6455,
+      "step": 8793
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.554690428814488e-05,
+      "loss": 0.4818,
+      "step": 8794
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5546327536984168e-05,
+      "loss": 0.5388,
+      "step": 8795
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5545750785823456e-05,
+      "loss": 0.5621,
+      "step": 8796
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5545174034662747e-05,
+      "loss": 0.511,
+      "step": 8797
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5544597283502035e-05,
+      "loss": 0.5633,
+      "step": 8798
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5544020532341322e-05,
+      "loss": 0.5357,
+      "step": 8799
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5543443781180613e-05,
+      "loss": 0.5895,
+      "step": 8800
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5542867030019898e-05,
+      "loss": 0.5026,
+      "step": 8801
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5542290278859185e-05,
+      "loss": 0.6281,
+      "step": 8802
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5541713527698476e-05,
+      "loss": 0.5774,
+      "step": 8803
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5541136776537764e-05,
+      "loss": 0.5432,
+      "step": 8804
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5540560025377052e-05,
+      "loss": 0.4364,
+      "step": 8805
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5539983274216343e-05,
+      "loss": 0.7434,
+      "step": 8806
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553940652305563e-05,
+      "loss": 0.6773,
+      "step": 8807
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5538829771894918e-05,
+      "loss": 0.5765,
+      "step": 8808
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5538253020734206e-05,
+      "loss": 0.5408,
+      "step": 8809
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5537676269573494e-05,
+      "loss": 0.5056,
+      "step": 8810
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553709951841278e-05,
+      "loss": 0.4841,
+      "step": 8811
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5536522767252072e-05,
+      "loss": 0.5632,
+      "step": 8812
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553594601609136e-05,
+      "loss": 0.5752,
+      "step": 8813
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5535369264930648e-05,
+      "loss": 0.4857,
+      "step": 8814
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5534792513769935e-05,
+      "loss": 0.6784,
+      "step": 8815
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5534215762609223e-05,
+      "loss": 0.5381,
+      "step": 8816
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553363901144851e-05,
+      "loss": 0.5223,
+      "step": 8817
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5533062260287802e-05,
+      "loss": 0.5045,
+      "step": 8818
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553248550912709e-05,
+      "loss": 0.5892,
+      "step": 8819
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5531908757966377e-05,
+      "loss": 0.5647,
+      "step": 8820
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5531332006805665e-05,
+      "loss": 0.4453,
+      "step": 8821
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5530755255644953e-05,
+      "loss": 0.569,
+      "step": 8822
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.553017850448424e-05,
+      "loss": 0.6382,
+      "step": 8823
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5529601753323528e-05,
+      "loss": 0.656,
+      "step": 8824
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.552902500216282e-05,
+      "loss": 0.501,
+      "step": 8825
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5528448251002107e-05,
+      "loss": 0.5208,
+      "step": 8826
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5527871499841394e-05,
+      "loss": 0.576,
+      "step": 8827
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5527294748680682e-05,
+      "loss": 0.5793,
+      "step": 8828
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.552671799751997e-05,
+      "loss": 0.5249,
+      "step": 8829
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5526141246359257e-05,
+      "loss": 0.574,
+      "step": 8830
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.552556449519855e-05,
+      "loss": 0.5861,
+      "step": 8831
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5524987744037836e-05,
+      "loss": 0.5254,
+      "step": 8832
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5524410992877124e-05,
+      "loss": 0.5857,
+      "step": 8833
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5523834241716415e-05,
+      "loss": 0.4261,
+      "step": 8834
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5523257490555703e-05,
+      "loss": 0.5254,
+      "step": 8835
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5522680739394987e-05,
+      "loss": 0.4904,
+      "step": 8836
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5522103988234278e-05,
+      "loss": 0.6584,
+      "step": 8837
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5521527237073566e-05,
+      "loss": 0.5132,
+      "step": 8838
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5520950485912853e-05,
+      "loss": 0.6227,
+      "step": 8839
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5520373734752144e-05,
+      "loss": 0.5132,
+      "step": 8840
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5519796983591432e-05,
+      "loss": 0.5991,
+      "step": 8841
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.551922023243072e-05,
+      "loss": 0.5508,
+      "step": 8842
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5518643481270007e-05,
+      "loss": 0.5714,
+      "step": 8843
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5518066730109295e-05,
+      "loss": 0.5272,
+      "step": 8844
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5517489978948583e-05,
+      "loss": 0.385,
+      "step": 8845
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.5516913227787874e-05,
+      "loss": 0.5425,
+      "step": 8846
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.551633647662716e-05,
+      "loss": 0.5318,
+      "step": 8847
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.551575972546645e-05,
+      "loss": 0.4148,
+      "step": 8848
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5515182974305737e-05,
+      "loss": 0.4858,
+      "step": 8849
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5514606223145025e-05,
+      "loss": 0.5901,
+      "step": 8850
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5514029471984312e-05,
+      "loss": 0.4661,
+      "step": 8851
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5513452720823603e-05,
+      "loss": 0.4094,
+      "step": 8852
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.551287596966289e-05,
+      "loss": 0.5545,
+      "step": 8853
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.551229921850218e-05,
+      "loss": 0.6001,
+      "step": 8854
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5511722467341466e-05,
+      "loss": 0.6706,
+      "step": 8855
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5511145716180754e-05,
+      "loss": 0.5087,
+      "step": 8856
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5510568965020042e-05,
+      "loss": 0.5613,
+      "step": 8857
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550999221385933e-05,
+      "loss": 0.575,
+      "step": 8858
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550941546269862e-05,
+      "loss": 0.5439,
+      "step": 8859
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5508838711537908e-05,
+      "loss": 0.4792,
+      "step": 8860
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5508261960377196e-05,
+      "loss": 0.5432,
+      "step": 8861
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5507685209216487e-05,
+      "loss": 0.4891,
+      "step": 8862
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550710845805577e-05,
+      "loss": 0.543,
+      "step": 8863
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550653170689506e-05,
+      "loss": 0.5107,
+      "step": 8864
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550595495573435e-05,
+      "loss": 0.5342,
+      "step": 8865
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5505378204573638e-05,
+      "loss": 0.6124,
+      "step": 8866
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5504801453412925e-05,
+      "loss": 0.4739,
+      "step": 8867
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5504224702252217e-05,
+      "loss": 0.4363,
+      "step": 8868
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5503647951091504e-05,
+      "loss": 0.4365,
+      "step": 8869
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5503071199930792e-05,
+      "loss": 0.4954,
+      "step": 8870
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.550249444877008e-05,
+      "loss": 0.6582,
+      "step": 8871
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5501917697609367e-05,
+      "loss": 0.5342,
+      "step": 8872
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5501340946448655e-05,
+      "loss": 0.5204,
+      "step": 8873
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5500764195287946e-05,
+      "loss": 0.4903,
+      "step": 8874
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5500187444127234e-05,
+      "loss": 0.6075,
+      "step": 8875
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549961069296652e-05,
+      "loss": 0.4974,
+      "step": 8876
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549903394180581e-05,
+      "loss": 0.6583,
+      "step": 8877
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5498457190645097e-05,
+      "loss": 0.5117,
+      "step": 8878
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5497880439484384e-05,
+      "loss": 0.556,
+      "step": 8879
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5497303688323676e-05,
+      "loss": 0.5523,
+      "step": 8880
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5496726937162963e-05,
+      "loss": 0.5367,
+      "step": 8881
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549615018600225e-05,
+      "loss": 0.6372,
+      "step": 8882
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549557343484154e-05,
+      "loss": 0.4999,
+      "step": 8883
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5494996683680826e-05,
+      "loss": 0.41,
+      "step": 8884
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5494419932520114e-05,
+      "loss": 0.6258,
+      "step": 8885
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5493843181359405e-05,
+      "loss": 0.5078,
+      "step": 8886
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5493266430198693e-05,
+      "loss": 0.6444,
+      "step": 8887
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549268967903798e-05,
+      "loss": 0.4986,
+      "step": 8888
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5492112927877268e-05,
+      "loss": 0.5774,
+      "step": 8889
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5491536176716556e-05,
+      "loss": 0.6226,
+      "step": 8890
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5490959425555843e-05,
+      "loss": 0.546,
+      "step": 8891
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.549038267439513e-05,
+      "loss": 0.5281,
+      "step": 8892
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5489805923234422e-05,
+      "loss": 0.4983,
+      "step": 8893
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548922917207371e-05,
+      "loss": 0.6055,
+      "step": 8894
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5488652420912998e-05,
+      "loss": 0.4516,
+      "step": 8895
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548807566975229e-05,
+      "loss": 0.6678,
+      "step": 8896
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5487498918591576e-05,
+      "loss": 0.4514,
+      "step": 8897
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548692216743086e-05,
+      "loss": 0.6469,
+      "step": 8898
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5486345416270152e-05,
+      "loss": 0.6087,
+      "step": 8899
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548576866510944e-05,
+      "loss": 0.6157,
+      "step": 8900
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5485191913948727e-05,
+      "loss": 0.5209,
+      "step": 8901
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5484615162788018e-05,
+      "loss": 0.4743,
+      "step": 8902
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5484038411627306e-05,
+      "loss": 0.529,
+      "step": 8903
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5483461660466594e-05,
+      "loss": 0.6196,
+      "step": 8904
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548288490930588e-05,
+      "loss": 0.5104,
+      "step": 8905
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.548230815814517e-05,
+      "loss": 0.5869,
+      "step": 8906
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5481731406984457e-05,
+      "loss": 0.536,
+      "step": 8907
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5481154655823748e-05,
+      "loss": 0.5697,
+      "step": 8908
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5480577904663035e-05,
+      "loss": 0.4924,
+      "step": 8909
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5480001153502323e-05,
+      "loss": 0.592,
+      "step": 8910
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.547942440234161e-05,
+      "loss": 0.4184,
+      "step": 8911
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54788476511809e-05,
+      "loss": 0.5109,
+      "step": 8912
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5478270900020186e-05,
+      "loss": 0.4586,
+      "step": 8913
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5477694148859477e-05,
+      "loss": 0.5054,
+      "step": 8914
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5477117397698765e-05,
+      "loss": 0.4285,
+      "step": 8915
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5476540646538053e-05,
+      "loss": 0.4895,
+      "step": 8916
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.547596389537734e-05,
+      "loss": 0.5255,
+      "step": 8917
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5475387144216628e-05,
+      "loss": 0.6031,
+      "step": 8918
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5474810393055916e-05,
+      "loss": 0.5316,
+      "step": 8919
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5474233641895207e-05,
+      "loss": 0.509,
+      "step": 8920
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5473656890734494e-05,
+      "loss": 0.6312,
+      "step": 8921
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5473080139573782e-05,
+      "loss": 0.5208,
+      "step": 8922
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5472503388413073e-05,
+      "loss": 0.384,
+      "step": 8923
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.547192663725236e-05,
+      "loss": 0.5248,
+      "step": 8924
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5471349886091645e-05,
+      "loss": 0.602,
+      "step": 8925
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5470773134930933e-05,
+      "loss": 0.5333,
+      "step": 8926
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5470196383770224e-05,
+      "loss": 0.5028,
+      "step": 8927
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546961963260951e-05,
+      "loss": 0.5305,
+      "step": 8928
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54690428814488e-05,
+      "loss": 0.5168,
+      "step": 8929
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546846613028809e-05,
+      "loss": 0.4883,
+      "step": 8930
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5467889379127378e-05,
+      "loss": 0.6166,
+      "step": 8931
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5467312627966666e-05,
+      "loss": 0.6765,
+      "step": 8932
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5466735876805953e-05,
+      "loss": 0.4445,
+      "step": 8933
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546615912564524e-05,
+      "loss": 0.6128,
+      "step": 8934
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546558237448453e-05,
+      "loss": 0.5787,
+      "step": 8935
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546500562332382e-05,
+      "loss": 0.5333,
+      "step": 8936
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5464428872163107e-05,
+      "loss": 0.5063,
+      "step": 8937
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5463852121002395e-05,
+      "loss": 0.4902,
+      "step": 8938
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5463275369841683e-05,
+      "loss": 0.7253,
+      "step": 8939
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546269861868097e-05,
+      "loss": 0.6135,
+      "step": 8940
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5462121867520258e-05,
+      "loss": 0.5087,
+      "step": 8941
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.546154511635955e-05,
+      "loss": 0.5127,
+      "step": 8942
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5460968365198837e-05,
+      "loss": 0.4226,
+      "step": 8943
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5460391614038125e-05,
+      "loss": 0.6328,
+      "step": 8944
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5459814862877412e-05,
+      "loss": 0.4512,
+      "step": 8945
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54592381117167e-05,
+      "loss": 0.5674,
+      "step": 8946
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5458661360555988e-05,
+      "loss": 0.4611,
+      "step": 8947
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.545808460939528e-05,
+      "loss": 0.6553,
+      "step": 8948
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5457507858234566e-05,
+      "loss": 0.4532,
+      "step": 8949
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5456931107073854e-05,
+      "loss": 0.5884,
+      "step": 8950
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5456354355913142e-05,
+      "loss": 0.7297,
+      "step": 8951
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.545577760475243e-05,
+      "loss": 0.5138,
+      "step": 8952
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5455200853591717e-05,
+      "loss": 0.6583,
+      "step": 8953
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5454624102431008e-05,
+      "loss": 0.5253,
+      "step": 8954
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5454047351270296e-05,
+      "loss": 0.515,
+      "step": 8955
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5453470600109584e-05,
+      "loss": 0.4827,
+      "step": 8956
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5452893848948875e-05,
+      "loss": 0.5308,
+      "step": 8957
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5452317097788162e-05,
+      "loss": 0.4801,
+      "step": 8958
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.545174034662745e-05,
+      "loss": 0.4925,
+      "step": 8959
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5451163595466734e-05,
+      "loss": 0.5043,
+      "step": 8960
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5450586844306025e-05,
+      "loss": 0.4958,
+      "step": 8961
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5450010093145313e-05,
+      "loss": 0.5266,
+      "step": 8962
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54494333419846e-05,
+      "loss": 0.5063,
+      "step": 8963
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5448856590823892e-05,
+      "loss": 0.5486,
+      "step": 8964
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544827983966318e-05,
+      "loss": 0.4633,
+      "step": 8965
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5447703088502467e-05,
+      "loss": 0.553,
+      "step": 8966
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5447126337341755e-05,
+      "loss": 0.5581,
+      "step": 8967
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5446549586181043e-05,
+      "loss": 0.6047,
+      "step": 8968
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544597283502033e-05,
+      "loss": 0.4845,
+      "step": 8969
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544539608385962e-05,
+      "loss": 0.5852,
+      "step": 8970
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544481933269891e-05,
+      "loss": 0.5492,
+      "step": 8971
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5444242581538197e-05,
+      "loss": 0.5469,
+      "step": 8972
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5443665830377484e-05,
+      "loss": 0.5472,
+      "step": 8973
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5443089079216772e-05,
+      "loss": 0.6513,
+      "step": 8974
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544251232805606e-05,
+      "loss": 0.5814,
+      "step": 8975
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544193557689535e-05,
+      "loss": 0.4861,
+      "step": 8976
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.544135882573464e-05,
+      "loss": 0.5284,
+      "step": 8977
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5440782074573926e-05,
+      "loss": 0.5825,
+      "step": 8978
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5440205323413214e-05,
+      "loss": 0.6118,
+      "step": 8979
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54396285722525e-05,
+      "loss": 0.3266,
+      "step": 8980
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.543905182109179e-05,
+      "loss": 0.469,
+      "step": 8981
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.543847506993108e-05,
+      "loss": 0.4558,
+      "step": 8982
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5437898318770368e-05,
+      "loss": 0.5407,
+      "step": 8983
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5437321567609656e-05,
+      "loss": 0.5732,
+      "step": 8984
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5436744816448947e-05,
+      "loss": 0.4012,
+      "step": 8985
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5436168065288235e-05,
+      "loss": 0.4207,
+      "step": 8986
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.543559131412752e-05,
+      "loss": 0.5965,
+      "step": 8987
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.543501456296681e-05,
+      "loss": 0.6158,
+      "step": 8988
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5434437811806098e-05,
+      "loss": 0.665,
+      "step": 8989
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5433861060645385e-05,
+      "loss": 0.5261,
+      "step": 8990
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5433284309484676e-05,
+      "loss": 0.5367,
+      "step": 8991
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5432707558323964e-05,
+      "loss": 0.5374,
+      "step": 8992
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5432130807163252e-05,
+      "loss": 0.5254,
+      "step": 8993
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.543155405600254e-05,
+      "loss": 0.536,
+      "step": 8994
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5430977304841827e-05,
+      "loss": 0.5563,
+      "step": 8995
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5430400553681115e-05,
+      "loss": 0.6249,
+      "step": 8996
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5429823802520402e-05,
+      "loss": 0.4927,
+      "step": 8997
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5429247051359694e-05,
+      "loss": 0.4744,
+      "step": 8998
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.542867030019898e-05,
+      "loss": 0.7024,
+      "step": 8999
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.542809354903827e-05,
+      "loss": 0.5055,
+      "step": 9000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5427516797877557e-05,
+      "loss": 0.6238,
+      "step": 9001
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5426940046716844e-05,
+      "loss": 0.641,
+      "step": 9002
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5426363295556132e-05,
+      "loss": 0.5708,
+      "step": 9003
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5425786544395423e-05,
+      "loss": 0.416,
+      "step": 9004
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.542520979323471e-05,
+      "loss": 0.6028,
+      "step": 9005
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5424633042074e-05,
+      "loss": 0.5215,
+      "step": 9006
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5424056290913286e-05,
+      "loss": 0.4881,
+      "step": 9007
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5423479539752574e-05,
+      "loss": 0.5696,
+      "step": 9008
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.542290278859186e-05,
+      "loss": 0.4801,
+      "step": 9009
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5422326037431153e-05,
+      "loss": 0.5243,
+      "step": 9010
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.542174928627044e-05,
+      "loss": 0.4875,
+      "step": 9011
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5421172535109728e-05,
+      "loss": 0.3793,
+      "step": 9012
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5420595783949016e-05,
+      "loss": 0.5857,
+      "step": 9013
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5420019032788303e-05,
+      "loss": 0.5736,
+      "step": 9014
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.541944228162759e-05,
+      "loss": 0.5251,
+      "step": 9015
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5418865530466882e-05,
+      "loss": 0.5875,
+      "step": 9016
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.541828877930617e-05,
+      "loss": 0.4813,
+      "step": 9017
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5417712028145457e-05,
+      "loss": 0.5,
+      "step": 9018
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.541713527698475e-05,
+      "loss": 0.4313,
+      "step": 9019
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5416558525824036e-05,
+      "loss": 0.3901,
+      "step": 9020
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5415981774663324e-05,
+      "loss": 0.5073,
+      "step": 9021
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.541540502350261e-05,
+      "loss": 0.5706,
+      "step": 9022
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.54148282723419e-05,
+      "loss": 0.6099,
+      "step": 9023
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5414251521181187e-05,
+      "loss": 0.5559,
+      "step": 9024
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5413674770020478e-05,
+      "loss": 0.4977,
+      "step": 9025
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.5413098018859766e-05,
+      "loss": 0.6655,
+      "step": 9026
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5412521267699053e-05,
+      "loss": 0.5073,
+      "step": 9027
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.541194451653834e-05,
+      "loss": 0.6286,
+      "step": 9028
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.541136776537763e-05,
+      "loss": 0.6051,
+      "step": 9029
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5410791014216916e-05,
+      "loss": 0.489,
+      "step": 9030
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5410214263056204e-05,
+      "loss": 0.4618,
+      "step": 9031
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5409637511895495e-05,
+      "loss": 0.6005,
+      "step": 9032
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5409060760734783e-05,
+      "loss": 0.5357,
+      "step": 9033
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.540848400957407e-05,
+      "loss": 0.5602,
+      "step": 9034
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5407907258413358e-05,
+      "loss": 0.6199,
+      "step": 9035
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5407330507252646e-05,
+      "loss": 0.6309,
+      "step": 9036
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5406753756091934e-05,
+      "loss": 0.7453,
+      "step": 9037
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5406177004931225e-05,
+      "loss": 0.467,
+      "step": 9038
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5405600253770512e-05,
+      "loss": 0.4388,
+      "step": 9039
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.54050235026098e-05,
+      "loss": 0.4669,
+      "step": 9040
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5404446751449088e-05,
+      "loss": 0.5124,
+      "step": 9041
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5403870000288375e-05,
+      "loss": 0.4842,
+      "step": 9042
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5403293249127663e-05,
+      "loss": 0.517,
+      "step": 9043
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5402716497966954e-05,
+      "loss": 0.4913,
+      "step": 9044
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5402139746806242e-05,
+      "loss": 0.5436,
+      "step": 9045
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.540156299564553e-05,
+      "loss": 0.5317,
+      "step": 9046
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.540098624448482e-05,
+      "loss": 0.5614,
+      "step": 9047
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5400409493324108e-05,
+      "loss": 0.4661,
+      "step": 9048
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5399832742163393e-05,
+      "loss": 0.3795,
+      "step": 9049
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5399255991002684e-05,
+      "loss": 0.5395,
+      "step": 9050
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539867923984197e-05,
+      "loss": 0.5286,
+      "step": 9051
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539810248868126e-05,
+      "loss": 0.4964,
+      "step": 9052
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539752573752055e-05,
+      "loss": 0.5568,
+      "step": 9053
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5396948986359838e-05,
+      "loss": 0.5342,
+      "step": 9054
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5396372235199125e-05,
+      "loss": 0.5253,
+      "step": 9055
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5395795484038413e-05,
+      "loss": 0.5973,
+      "step": 9056
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53952187328777e-05,
+      "loss": 0.4348,
+      "step": 9057
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539464198171699e-05,
+      "loss": 0.6559,
+      "step": 9058
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539406523055628e-05,
+      "loss": 0.4903,
+      "step": 9059
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5393488479395567e-05,
+      "loss": 0.4992,
+      "step": 9060
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5392911728234855e-05,
+      "loss": 0.6053,
+      "step": 9061
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5392334977074143e-05,
+      "loss": 0.4693,
+      "step": 9062
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.539175822591343e-05,
+      "loss": 0.607,
+      "step": 9063
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5391181474752718e-05,
+      "loss": 0.6528,
+      "step": 9064
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5390604723592006e-05,
+      "loss": 0.5247,
+      "step": 9065
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5390027972431297e-05,
+      "loss": 0.5186,
+      "step": 9066
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5389451221270584e-05,
+      "loss": 0.5899,
+      "step": 9067
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5388874470109872e-05,
+      "loss": 0.4857,
+      "step": 9068
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.538829771894916e-05,
+      "loss": 0.5896,
+      "step": 9069
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5387720967788447e-05,
+      "loss": 0.5245,
+      "step": 9070
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5387144216627735e-05,
+      "loss": 0.4524,
+      "step": 9071
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5386567465467026e-05,
+      "loss": 0.554,
+      "step": 9072
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5385990714306314e-05,
+      "loss": 0.5826,
+      "step": 9073
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53854139631456e-05,
+      "loss": 0.5731,
+      "step": 9074
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.538483721198489e-05,
+      "loss": 0.5661,
+      "step": 9075
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5384260460824177e-05,
+      "loss": 0.4995,
+      "step": 9076
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5383683709663465e-05,
+      "loss": 0.623,
+      "step": 9077
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5383106958502756e-05,
+      "loss": 0.5659,
+      "step": 9078
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5382530207342043e-05,
+      "loss": 0.435,
+      "step": 9079
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.538195345618133e-05,
+      "loss": 0.5961,
+      "step": 9080
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5381376705020622e-05,
+      "loss": 0.4496,
+      "step": 9081
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.538079995385991e-05,
+      "loss": 0.5306,
+      "step": 9082
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5380223202699198e-05,
+      "loss": 0.5198,
+      "step": 9083
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5379646451538485e-05,
+      "loss": 0.503,
+      "step": 9084
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5379069700377773e-05,
+      "loss": 0.6257,
+      "step": 9085
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537849294921706e-05,
+      "loss": 0.6026,
+      "step": 9086
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537791619805635e-05,
+      "loss": 0.5663,
+      "step": 9087
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537733944689564e-05,
+      "loss": 0.4619,
+      "step": 9088
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5376762695734927e-05,
+      "loss": 0.4442,
+      "step": 9089
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5376185944574215e-05,
+      "loss": 0.5402,
+      "step": 9090
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5375609193413502e-05,
+      "loss": 0.4563,
+      "step": 9091
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537503244225279e-05,
+      "loss": 0.5676,
+      "step": 9092
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537445569109208e-05,
+      "loss": 0.6063,
+      "step": 9093
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537387893993137e-05,
+      "loss": 0.5691,
+      "step": 9094
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5373302188770657e-05,
+      "loss": 0.5544,
+      "step": 9095
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5372725437609944e-05,
+      "loss": 0.6511,
+      "step": 9096
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5372148686449232e-05,
+      "loss": 0.6272,
+      "step": 9097
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.537157193528852e-05,
+      "loss": 0.5255,
+      "step": 9098
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5370995184127807e-05,
+      "loss": 0.6452,
+      "step": 9099
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53704184329671e-05,
+      "loss": 0.4547,
+      "step": 9100
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5369841681806386e-05,
+      "loss": 0.5351,
+      "step": 9101
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5369264930645674e-05,
+      "loss": 0.5352,
+      "step": 9102
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.536868817948496e-05,
+      "loss": 0.4783,
+      "step": 9103
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.536811142832425e-05,
+      "loss": 0.5034,
+      "step": 9104
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5367534677163537e-05,
+      "loss": 0.6169,
+      "step": 9105
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5366957926002828e-05,
+      "loss": 0.6228,
+      "step": 9106
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5366381174842116e-05,
+      "loss": 0.5694,
+      "step": 9107
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5365804423681403e-05,
+      "loss": 0.5336,
+      "step": 9108
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5365227672520694e-05,
+      "loss": 0.5669,
+      "step": 9109
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5364650921359982e-05,
+      "loss": 0.5264,
+      "step": 9110
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5364074170199266e-05,
+      "loss": 0.3696,
+      "step": 9111
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5363497419038557e-05,
+      "loss": 0.577,
+      "step": 9112
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5362920667877845e-05,
+      "loss": 0.5521,
+      "step": 9113
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5362343916717133e-05,
+      "loss": 0.6164,
+      "step": 9114
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5361767165556424e-05,
+      "loss": 0.5832,
+      "step": 9115
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.536119041439571e-05,
+      "loss": 0.5635,
+      "step": 9116
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5360613663235e-05,
+      "loss": 0.5392,
+      "step": 9117
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5360036912074287e-05,
+      "loss": 0.5057,
+      "step": 9118
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5359460160913575e-05,
+      "loss": 0.5755,
+      "step": 9119
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5358883409752862e-05,
+      "loss": 0.4218,
+      "step": 9120
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5358306658592153e-05,
+      "loss": 0.4212,
+      "step": 9121
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.535772990743144e-05,
+      "loss": 0.6404,
+      "step": 9122
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.535715315627073e-05,
+      "loss": 0.3935,
+      "step": 9123
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5356576405110016e-05,
+      "loss": 0.4882,
+      "step": 9124
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5355999653949304e-05,
+      "loss": 0.485,
+      "step": 9125
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5355422902788592e-05,
+      "loss": 0.5303,
+      "step": 9126
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5354846151627883e-05,
+      "loss": 0.591,
+      "step": 9127
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.535426940046717e-05,
+      "loss": 0.5708,
+      "step": 9128
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5353692649306458e-05,
+      "loss": 0.5241,
+      "step": 9129
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5353115898145746e-05,
+      "loss": 0.6959,
+      "step": 9130
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5352539146985034e-05,
+      "loss": 0.5175,
+      "step": 9131
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.535196239582432e-05,
+      "loss": 0.5596,
+      "step": 9132
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.535138564466361e-05,
+      "loss": 0.6792,
+      "step": 9133
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53508088935029e-05,
+      "loss": 0.5475,
+      "step": 9134
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5350232142342188e-05,
+      "loss": 0.4526,
+      "step": 9135
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5349655391181475e-05,
+      "loss": 0.6472,
+      "step": 9136
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5349078640020766e-05,
+      "loss": 0.455,
+      "step": 9137
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534850188886005e-05,
+      "loss": 0.6359,
+      "step": 9138
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534792513769934e-05,
+      "loss": 0.5993,
+      "step": 9139
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534734838653863e-05,
+      "loss": 0.6336,
+      "step": 9140
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5346771635377917e-05,
+      "loss": 0.5608,
+      "step": 9141
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5346194884217205e-05,
+      "loss": 0.3866,
+      "step": 9142
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5345618133056496e-05,
+      "loss": 0.5505,
+      "step": 9143
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5345041381895784e-05,
+      "loss": 0.4792,
+      "step": 9144
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534446463073507e-05,
+      "loss": 0.4766,
+      "step": 9145
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534388787957436e-05,
+      "loss": 0.5004,
+      "step": 9146
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5343311128413647e-05,
+      "loss": 0.5434,
+      "step": 9147
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5342734377252934e-05,
+      "loss": 0.4712,
+      "step": 9148
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5342157626092225e-05,
+      "loss": 0.5957,
+      "step": 9149
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5341580874931513e-05,
+      "loss": 0.555,
+      "step": 9150
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53410041237708e-05,
+      "loss": 0.4506,
+      "step": 9151
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.534042737261009e-05,
+      "loss": 0.4902,
+      "step": 9152
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5339850621449376e-05,
+      "loss": 0.4675,
+      "step": 9153
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5339273870288664e-05,
+      "loss": 0.5902,
+      "step": 9154
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5338697119127955e-05,
+      "loss": 0.5854,
+      "step": 9155
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5338120367967243e-05,
+      "loss": 0.4785,
+      "step": 9156
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.533754361680653e-05,
+      "loss": 0.6202,
+      "step": 9157
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5336966865645818e-05,
+      "loss": 0.5045,
+      "step": 9158
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5336390114485106e-05,
+      "loss": 0.4715,
+      "step": 9159
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5335813363324393e-05,
+      "loss": 0.5015,
+      "step": 9160
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5335236612163684e-05,
+      "loss": 0.3866,
+      "step": 9161
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5334659861002972e-05,
+      "loss": 0.638,
+      "step": 9162
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.533408310984226e-05,
+      "loss": 0.5902,
+      "step": 9163
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5333506358681547e-05,
+      "loss": 0.5169,
+      "step": 9164
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5332929607520835e-05,
+      "loss": 0.5386,
+      "step": 9165
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5332352856360123e-05,
+      "loss": 0.548,
+      "step": 9166
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5331776105199414e-05,
+      "loss": 0.5074,
+      "step": 9167
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.53311993540387e-05,
+      "loss": 0.5516,
+      "step": 9168
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.533062260287799e-05,
+      "loss": 0.4395,
+      "step": 9169
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5330045851717277e-05,
+      "loss": 0.5812,
+      "step": 9170
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5329469100556568e-05,
+      "loss": 0.4578,
+      "step": 9171
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5328892349395856e-05,
+      "loss": 0.4999,
+      "step": 9172
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.532831559823514e-05,
+      "loss": 0.4912,
+      "step": 9173
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.532773884707443e-05,
+      "loss": 0.5431,
+      "step": 9174
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.532716209591372e-05,
+      "loss": 0.4515,
+      "step": 9175
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5326585344753006e-05,
+      "loss": 0.6329,
+      "step": 9176
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5326008593592298e-05,
+      "loss": 0.6403,
+      "step": 9177
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5325431842431585e-05,
+      "loss": 0.5169,
+      "step": 9178
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5324855091270873e-05,
+      "loss": 0.4945,
+      "step": 9179
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.532427834011016e-05,
+      "loss": 0.424,
+      "step": 9180
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5323701588949448e-05,
+      "loss": 0.6453,
+      "step": 9181
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5323124837788736e-05,
+      "loss": 0.4891,
+      "step": 9182
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5322548086628027e-05,
+      "loss": 0.5041,
+      "step": 9183
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5321971335467315e-05,
+      "loss": 0.5658,
+      "step": 9184
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5321394584306602e-05,
+      "loss": 0.4967,
+      "step": 9185
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.532081783314589e-05,
+      "loss": 0.6064,
+      "step": 9186
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5320241081985178e-05,
+      "loss": 0.4637,
+      "step": 9187
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5319664330824465e-05,
+      "loss": 0.481,
+      "step": 9188
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5319087579663757e-05,
+      "loss": 0.6283,
+      "step": 9189
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5318510828503044e-05,
+      "loss": 0.5803,
+      "step": 9190
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5317934077342332e-05,
+      "loss": 0.5598,
+      "step": 9191
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.531735732618162e-05,
+      "loss": 0.5106,
+      "step": 9192
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5316780575020907e-05,
+      "loss": 0.5384,
+      "step": 9193
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5316203823860195e-05,
+      "loss": 0.5015,
+      "step": 9194
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5315627072699486e-05,
+      "loss": 0.5063,
+      "step": 9195
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5315050321538774e-05,
+      "loss": 0.5531,
+      "step": 9196
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.531447357037806e-05,
+      "loss": 0.4468,
+      "step": 9197
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5313896819217353e-05,
+      "loss": 0.5632,
+      "step": 9198
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.531332006805664e-05,
+      "loss": 0.5999,
+      "step": 9199
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5312743316895924e-05,
+      "loss": 0.4705,
+      "step": 9200
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5312166565735216e-05,
+      "loss": 0.4713,
+      "step": 9201
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.5311589814574503e-05,
+      "loss": 0.6549,
+      "step": 9202
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.531101306341379e-05,
+      "loss": 0.5727,
+      "step": 9203
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.531043631225308e-05,
+      "loss": 0.4833,
+      "step": 9204
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.530985956109237e-05,
+      "loss": 0.5152,
+      "step": 9205
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5309282809931657e-05,
+      "loss": 0.5358,
+      "step": 9206
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5308706058770945e-05,
+      "loss": 0.639,
+      "step": 9207
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5308129307610233e-05,
+      "loss": 0.5995,
+      "step": 9208
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.530755255644952e-05,
+      "loss": 0.6091,
+      "step": 9209
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5306975805288808e-05,
+      "loss": 0.5775,
+      "step": 9210
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.53063990541281e-05,
+      "loss": 0.5237,
+      "step": 9211
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5305822302967387e-05,
+      "loss": 0.5094,
+      "step": 9212
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5305245551806675e-05,
+      "loss": 0.483,
+      "step": 9213
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5304668800645962e-05,
+      "loss": 0.5979,
+      "step": 9214
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.530409204948525e-05,
+      "loss": 0.498,
+      "step": 9215
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5303515298324538e-05,
+      "loss": 0.5008,
+      "step": 9216
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.530293854716383e-05,
+      "loss": 0.461,
+      "step": 9217
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5302361796003116e-05,
+      "loss": 0.5224,
+      "step": 9218
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5301785044842404e-05,
+      "loss": 0.5971,
+      "step": 9219
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5301208293681692e-05,
+      "loss": 0.4898,
+      "step": 9220
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.530063154252098e-05,
+      "loss": 0.5784,
+      "step": 9221
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5300054791360267e-05,
+      "loss": 0.458,
+      "step": 9222
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5299478040199558e-05,
+      "loss": 0.6008,
+      "step": 9223
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5298901289038846e-05,
+      "loss": 0.5334,
+      "step": 9224
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5298324537878134e-05,
+      "loss": 0.5307,
+      "step": 9225
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.529774778671742e-05,
+      "loss": 0.5563,
+      "step": 9226
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.529717103555671e-05,
+      "loss": 0.5259,
+      "step": 9227
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5296594284395997e-05,
+      "loss": 0.4861,
+      "step": 9228
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5296017533235288e-05,
+      "loss": 0.6388,
+      "step": 9229
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5295440782074575e-05,
+      "loss": 0.503,
+      "step": 9230
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5294864030913863e-05,
+      "loss": 0.4413,
+      "step": 9231
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5294287279753154e-05,
+      "loss": 0.5321,
+      "step": 9232
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5293710528592442e-05,
+      "loss": 0.5184,
+      "step": 9233
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.529313377743173e-05,
+      "loss": 0.5488,
+      "step": 9234
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5292557026271017e-05,
+      "loss": 0.4771,
+      "step": 9235
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5291980275110305e-05,
+      "loss": 0.5356,
+      "step": 9236
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5291403523949593e-05,
+      "loss": 0.6234,
+      "step": 9237
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.529082677278888e-05,
+      "loss": 0.704,
+      "step": 9238
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.529025002162817e-05,
+      "loss": 0.4223,
+      "step": 9239
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528967327046746e-05,
+      "loss": 0.5112,
+      "step": 9240
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5289096519306747e-05,
+      "loss": 0.5738,
+      "step": 9241
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5288519768146034e-05,
+      "loss": 0.4763,
+      "step": 9242
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5287943016985322e-05,
+      "loss": 0.4646,
+      "step": 9243
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528736626582461e-05,
+      "loss": 0.4356,
+      "step": 9244
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.52867895146639e-05,
+      "loss": 0.4989,
+      "step": 9245
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528621276350319e-05,
+      "loss": 0.5732,
+      "step": 9246
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5285636012342476e-05,
+      "loss": 0.4934,
+      "step": 9247
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5285059261181764e-05,
+      "loss": 0.5436,
+      "step": 9248
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528448251002105e-05,
+      "loss": 0.6014,
+      "step": 9249
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528390575886034e-05,
+      "loss": 0.5512,
+      "step": 9250
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528332900769963e-05,
+      "loss": 0.6316,
+      "step": 9251
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5282752256538918e-05,
+      "loss": 0.5836,
+      "step": 9252
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5282175505378206e-05,
+      "loss": 0.535,
+      "step": 9253
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5281598754217493e-05,
+      "loss": 0.5401,
+      "step": 9254
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528102200305678e-05,
+      "loss": 0.4288,
+      "step": 9255
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.528044525189607e-05,
+      "loss": 0.6586,
+      "step": 9256
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.527986850073536e-05,
+      "loss": 0.5376,
+      "step": 9257
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5279291749574647e-05,
+      "loss": 0.4558,
+      "step": 9258
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5278714998413935e-05,
+      "loss": 0.5433,
+      "step": 9259
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5278138247253226e-05,
+      "loss": 0.5105,
+      "step": 9260
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5277561496092514e-05,
+      "loss": 0.4738,
+      "step": 9261
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5276984744931798e-05,
+      "loss": 0.5988,
+      "step": 9262
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.527640799377109e-05,
+      "loss": 0.5039,
+      "step": 9263
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5275831242610377e-05,
+      "loss": 0.4406,
+      "step": 9264
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5275254491449665e-05,
+      "loss": 0.5395,
+      "step": 9265
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5274677740288956e-05,
+      "loss": 0.4848,
+      "step": 9266
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5274100989128243e-05,
+      "loss": 0.5005,
+      "step": 9267
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.527352423796753e-05,
+      "loss": 0.5706,
+      "step": 9268
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.527294748680682e-05,
+      "loss": 0.4411,
+      "step": 9269
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5272370735646106e-05,
+      "loss": 0.549,
+      "step": 9270
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5271793984485394e-05,
+      "loss": 0.4103,
+      "step": 9271
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5271217233324685e-05,
+      "loss": 0.4715,
+      "step": 9272
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5270640482163973e-05,
+      "loss": 0.4959,
+      "step": 9273
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.527006373100326e-05,
+      "loss": 0.6307,
+      "step": 9274
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5269486979842548e-05,
+      "loss": 0.5551,
+      "step": 9275
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5268910228681836e-05,
+      "loss": 0.5563,
+      "step": 9276
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5268333477521124e-05,
+      "loss": 0.5543,
+      "step": 9277
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526775672636041e-05,
+      "loss": 0.5241,
+      "step": 9278
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5267179975199702e-05,
+      "loss": 0.5356,
+      "step": 9279
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526660322403899e-05,
+      "loss": 0.604,
+      "step": 9280
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5266026472878278e-05,
+      "loss": 0.4885,
+      "step": 9281
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5265449721717565e-05,
+      "loss": 0.5301,
+      "step": 9282
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5264872970556853e-05,
+      "loss": 0.5615,
+      "step": 9283
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526429621939614e-05,
+      "loss": 0.4987,
+      "step": 9284
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5263719468235432e-05,
+      "loss": 0.4757,
+      "step": 9285
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526314271707472e-05,
+      "loss": 0.5116,
+      "step": 9286
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5262565965914007e-05,
+      "loss": 0.5022,
+      "step": 9287
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5261989214753295e-05,
+      "loss": 0.5829,
+      "step": 9288
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5261412463592583e-05,
+      "loss": 0.4806,
+      "step": 9289
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526083571243187e-05,
+      "loss": 0.5716,
+      "step": 9290
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.526025896127116e-05,
+      "loss": 0.6868,
+      "step": 9291
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.525968221011045e-05,
+      "loss": 0.6147,
+      "step": 9292
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5259105458949737e-05,
+      "loss": 0.6019,
+      "step": 9293
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5258528707789026e-05,
+      "loss": 0.57,
+      "step": 9294
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5257951956628314e-05,
+      "loss": 0.5805,
+      "step": 9295
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5257375205467602e-05,
+      "loss": 0.5484,
+      "step": 9296
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5256798454306891e-05,
+      "loss": 0.5273,
+      "step": 9297
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5256221703146179e-05,
+      "loss": 0.5613,
+      "step": 9298
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5255644951985466e-05,
+      "loss": 0.5918,
+      "step": 9299
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5255068200824756e-05,
+      "loss": 0.5727,
+      "step": 9300
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5254491449664043e-05,
+      "loss": 0.3385,
+      "step": 9301
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5253914698503331e-05,
+      "loss": 0.6424,
+      "step": 9302
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5253337947342622e-05,
+      "loss": 0.4399,
+      "step": 9303
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5252761196181908e-05,
+      "loss": 0.7147,
+      "step": 9304
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5252184445021196e-05,
+      "loss": 0.4867,
+      "step": 9305
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5251607693860487e-05,
+      "loss": 0.4996,
+      "step": 9306
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5251030942699775e-05,
+      "loss": 0.3675,
+      "step": 9307
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.525045419153906e-05,
+      "loss": 0.4377,
+      "step": 9308
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5249877440378348e-05,
+      "loss": 0.5208,
+      "step": 9309
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.524930068921764e-05,
+      "loss": 0.489,
+      "step": 9310
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5248723938056927e-05,
+      "loss": 0.4438,
+      "step": 9311
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5248147186896215e-05,
+      "loss": 0.4831,
+      "step": 9312
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5247570435735504e-05,
+      "loss": 0.4605,
+      "step": 9313
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5246993684574792e-05,
+      "loss": 0.4833,
+      "step": 9314
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.524641693341408e-05,
+      "loss": 0.4783,
+      "step": 9315
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5245840182253369e-05,
+      "loss": 0.4893,
+      "step": 9316
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5245263431092656e-05,
+      "loss": 0.4727,
+      "step": 9317
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5244686679931944e-05,
+      "loss": 0.5272,
+      "step": 9318
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5244109928771234e-05,
+      "loss": 0.6113,
+      "step": 9319
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5243533177610521e-05,
+      "loss": 0.679,
+      "step": 9320
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5242956426449809e-05,
+      "loss": 0.614,
+      "step": 9321
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5242379675289098e-05,
+      "loss": 0.4926,
+      "step": 9322
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5241802924128386e-05,
+      "loss": 0.6261,
+      "step": 9323
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5241226172967674e-05,
+      "loss": 0.5509,
+      "step": 9324
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5240649421806963e-05,
+      "loss": 0.4798,
+      "step": 9325
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.524007267064625e-05,
+      "loss": 0.5333,
+      "step": 9326
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5239495919485538e-05,
+      "loss": 0.6319,
+      "step": 9327
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5238919168324828e-05,
+      "loss": 0.4075,
+      "step": 9328
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5238342417164115e-05,
+      "loss": 0.4614,
+      "step": 9329
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5237765666003403e-05,
+      "loss": 0.5785,
+      "step": 9330
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5237188914842693e-05,
+      "loss": 0.4535,
+      "step": 9331
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.523661216368198e-05,
+      "loss": 0.5625,
+      "step": 9332
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5236035412521268e-05,
+      "loss": 0.5412,
+      "step": 9333
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5235458661360559e-05,
+      "loss": 0.4931,
+      "step": 9334
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5234881910199845e-05,
+      "loss": 0.4844,
+      "step": 9335
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5234305159039133e-05,
+      "loss": 0.4628,
+      "step": 9336
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5233728407878424e-05,
+      "loss": 0.7092,
+      "step": 9337
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5233151656717711e-05,
+      "loss": 0.5091,
+      "step": 9338
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5232574905556997e-05,
+      "loss": 0.5645,
+      "step": 9339
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5231998154396288e-05,
+      "loss": 0.55,
+      "step": 9340
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5231421403235576e-05,
+      "loss": 0.5614,
+      "step": 9341
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5230844652074864e-05,
+      "loss": 0.6279,
+      "step": 9342
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5230267900914152e-05,
+      "loss": 0.4978,
+      "step": 9343
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5229691149753441e-05,
+      "loss": 0.4758,
+      "step": 9344
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5229114398592729e-05,
+      "loss": 0.452,
+      "step": 9345
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5228537647432016e-05,
+      "loss": 0.5933,
+      "step": 9346
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5227960896271306e-05,
+      "loss": 0.5412,
+      "step": 9347
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5227384145110593e-05,
+      "loss": 0.543,
+      "step": 9348
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5226807393949881e-05,
+      "loss": 0.4987,
+      "step": 9349
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.522623064278917e-05,
+      "loss": 0.804,
+      "step": 9350
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5225653891628458e-05,
+      "loss": 0.5475,
+      "step": 9351
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5225077140467746e-05,
+      "loss": 0.5667,
+      "step": 9352
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5224500389307035e-05,
+      "loss": 0.376,
+      "step": 9353
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5223923638146323e-05,
+      "loss": 0.6511,
+      "step": 9354
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.522334688698561e-05,
+      "loss": 0.4757,
+      "step": 9355
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.52227701358249e-05,
+      "loss": 0.6144,
+      "step": 9356
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5222193384664188e-05,
+      "loss": 0.6621,
+      "step": 9357
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5221616633503475e-05,
+      "loss": 0.4938,
+      "step": 9358
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5221039882342765e-05,
+      "loss": 0.5653,
+      "step": 9359
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5220463131182052e-05,
+      "loss": 0.5611,
+      "step": 9360
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.521988638002134e-05,
+      "loss": 0.5906,
+      "step": 9361
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.521930962886063e-05,
+      "loss": 0.5776,
+      "step": 9362
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5218732877699917e-05,
+      "loss": 0.534,
+      "step": 9363
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5218156126539205e-05,
+      "loss": 0.6197,
+      "step": 9364
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5217579375378496e-05,
+      "loss": 0.581,
+      "step": 9365
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5217002624217782e-05,
+      "loss": 0.7642,
+      "step": 9366
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.521642587305707e-05,
+      "loss": 0.6257,
+      "step": 9367
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.521584912189636e-05,
+      "loss": 0.4909,
+      "step": 9368
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5215272370735648e-05,
+      "loss": 0.4972,
+      "step": 9369
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5214695619574934e-05,
+      "loss": 0.5298,
+      "step": 9370
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5214118868414225e-05,
+      "loss": 0.5621,
+      "step": 9371
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5213542117253513e-05,
+      "loss": 0.548,
+      "step": 9372
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.52129653660928e-05,
+      "loss": 0.522,
+      "step": 9373
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.521238861493209e-05,
+      "loss": 0.5714,
+      "step": 9374
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5211811863771378e-05,
+      "loss": 0.5253,
+      "step": 9375
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5211235112610665e-05,
+      "loss": 0.689,
+      "step": 9376
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5210658361449953e-05,
+      "loss": 0.5008,
+      "step": 9377
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5210081610289243e-05,
+      "loss": 0.4963,
+      "step": 9378
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.520950485912853e-05,
+      "loss": 0.5315,
+      "step": 9379
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5208928107967818e-05,
+      "loss": 0.5056,
+      "step": 9380
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5208351356807107e-05,
+      "loss": 0.5507,
+      "step": 9381
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5207774605646395e-05,
+      "loss": 0.5464,
+      "step": 9382
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5207197854485683e-05,
+      "loss": 0.4415,
+      "step": 9383
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.5206621103324972e-05,
+      "loss": 0.6409,
+      "step": 9384
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.520604435216426e-05,
+      "loss": 0.5562,
+      "step": 9385
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5205467601003547e-05,
+      "loss": 0.5019,
+      "step": 9386
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5204890849842837e-05,
+      "loss": 0.443,
+      "step": 9387
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5204314098682124e-05,
+      "loss": 0.4925,
+      "step": 9388
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5203737347521412e-05,
+      "loss": 0.3904,
+      "step": 9389
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5203160596360702e-05,
+      "loss": 0.5597,
+      "step": 9390
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.520258384519999e-05,
+      "loss": 0.611,
+      "step": 9391
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5202007094039277e-05,
+      "loss": 0.5205,
+      "step": 9392
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5201430342878566e-05,
+      "loss": 0.4135,
+      "step": 9393
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5200853591717854e-05,
+      "loss": 0.5878,
+      "step": 9394
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5200276840557142e-05,
+      "loss": 0.6007,
+      "step": 9395
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5199700089396433e-05,
+      "loss": 0.4217,
+      "step": 9396
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5199123338235719e-05,
+      "loss": 0.5527,
+      "step": 9397
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5198546587075006e-05,
+      "loss": 0.62,
+      "step": 9398
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5197969835914297e-05,
+      "loss": 0.5692,
+      "step": 9399
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5197393084753585e-05,
+      "loss": 0.4892,
+      "step": 9400
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5196816333592871e-05,
+      "loss": 0.5529,
+      "step": 9401
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5196239582432162e-05,
+      "loss": 0.5112,
+      "step": 9402
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.519566283127145e-05,
+      "loss": 0.5572,
+      "step": 9403
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5195086080110738e-05,
+      "loss": 0.5901,
+      "step": 9404
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5194509328950027e-05,
+      "loss": 0.5591,
+      "step": 9405
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5193932577789315e-05,
+      "loss": 0.6128,
+      "step": 9406
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5193355826628602e-05,
+      "loss": 0.585,
+      "step": 9407
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5192779075467892e-05,
+      "loss": 0.5096,
+      "step": 9408
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.519220232430718e-05,
+      "loss": 0.6366,
+      "step": 9409
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5191625573146467e-05,
+      "loss": 0.4605,
+      "step": 9410
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5191048821985756e-05,
+      "loss": 0.4584,
+      "step": 9411
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5190472070825044e-05,
+      "loss": 0.5878,
+      "step": 9412
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5189895319664332e-05,
+      "loss": 0.5064,
+      "step": 9413
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.518931856850362e-05,
+      "loss": 0.4583,
+      "step": 9414
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5188741817342909e-05,
+      "loss": 0.6485,
+      "step": 9415
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5188165066182197e-05,
+      "loss": 0.5448,
+      "step": 9416
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5187588315021484e-05,
+      "loss": 0.5409,
+      "step": 9417
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5187011563860774e-05,
+      "loss": 0.5259,
+      "step": 9418
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5186434812700061e-05,
+      "loss": 0.5293,
+      "step": 9419
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5185858061539349e-05,
+      "loss": 0.4821,
+      "step": 9420
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5185281310378638e-05,
+      "loss": 0.667,
+      "step": 9421
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5184704559217926e-05,
+      "loss": 0.441,
+      "step": 9422
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5184127808057214e-05,
+      "loss": 0.4747,
+      "step": 9423
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5183551056896503e-05,
+      "loss": 0.5686,
+      "step": 9424
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5182974305735791e-05,
+      "loss": 0.491,
+      "step": 9425
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5182397554575079e-05,
+      "loss": 0.5269,
+      "step": 9426
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.518182080341437e-05,
+      "loss": 0.395,
+      "step": 9427
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5181244052253656e-05,
+      "loss": 0.4038,
+      "step": 9428
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5180667301092943e-05,
+      "loss": 0.675,
+      "step": 9429
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5180090549932234e-05,
+      "loss": 0.5457,
+      "step": 9430
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5179513798771522e-05,
+      "loss": 0.605,
+      "step": 9431
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.517893704761081e-05,
+      "loss": 0.5311,
+      "step": 9432
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5178360296450099e-05,
+      "loss": 0.4808,
+      "step": 9433
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5177783545289387e-05,
+      "loss": 0.5364,
+      "step": 9434
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5177206794128674e-05,
+      "loss": 0.5286,
+      "step": 9435
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5176630042967964e-05,
+      "loss": 0.5088,
+      "step": 9436
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5176053291807252e-05,
+      "loss": 0.5532,
+      "step": 9437
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.517547654064654e-05,
+      "loss": 0.5761,
+      "step": 9438
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5174899789485829e-05,
+      "loss": 0.4971,
+      "step": 9439
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5174323038325116e-05,
+      "loss": 0.4558,
+      "step": 9440
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5173746287164404e-05,
+      "loss": 0.4801,
+      "step": 9441
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5173169536003693e-05,
+      "loss": 0.668,
+      "step": 9442
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5172592784842981e-05,
+      "loss": 0.4528,
+      "step": 9443
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5172016033682269e-05,
+      "loss": 0.5267,
+      "step": 9444
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5171439282521558e-05,
+      "loss": 0.5818,
+      "step": 9445
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5170862531360846e-05,
+      "loss": 0.5002,
+      "step": 9446
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5170285780200133e-05,
+      "loss": 0.5884,
+      "step": 9447
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5169709029039421e-05,
+      "loss": 0.5021,
+      "step": 9448
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.516913227787871e-05,
+      "loss": 0.4999,
+      "step": 9449
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5168555526717998e-05,
+      "loss": 0.5322,
+      "step": 9450
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5167978775557286e-05,
+      "loss": 0.554,
+      "step": 9451
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5167402024396575e-05,
+      "loss": 0.532,
+      "step": 9452
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5166825273235863e-05,
+      "loss": 0.6137,
+      "step": 9453
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.516624852207515e-05,
+      "loss": 0.5892,
+      "step": 9454
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.516567177091444e-05,
+      "loss": 0.5991,
+      "step": 9455
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5165095019753728e-05,
+      "loss": 0.4268,
+      "step": 9456
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5164518268593015e-05,
+      "loss": 0.4455,
+      "step": 9457
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5163941517432306e-05,
+      "loss": 0.5832,
+      "step": 9458
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5163364766271592e-05,
+      "loss": 0.6235,
+      "step": 9459
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.516278801511088e-05,
+      "loss": 0.4236,
+      "step": 9460
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5162211263950171e-05,
+      "loss": 0.4799,
+      "step": 9461
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5161634512789459e-05,
+      "loss": 0.5406,
+      "step": 9462
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5161057761628747e-05,
+      "loss": 0.5755,
+      "step": 9463
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5160481010468036e-05,
+      "loss": 0.5208,
+      "step": 9464
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5159904259307324e-05,
+      "loss": 0.4948,
+      "step": 9465
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5159327508146611e-05,
+      "loss": 0.511,
+      "step": 9466
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.51587507569859e-05,
+      "loss": 0.4103,
+      "step": 9467
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5158174005825188e-05,
+      "loss": 0.4865,
+      "step": 9468
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5157597254664476e-05,
+      "loss": 0.6373,
+      "step": 9469
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5157020503503765e-05,
+      "loss": 0.4946,
+      "step": 9470
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5156443752343053e-05,
+      "loss": 0.6002,
+      "step": 9471
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5155867001182341e-05,
+      "loss": 0.4941,
+      "step": 9472
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.515529025002163e-05,
+      "loss": 0.6143,
+      "step": 9473
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5154713498860918e-05,
+      "loss": 0.4351,
+      "step": 9474
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5154136747700206e-05,
+      "loss": 0.4365,
+      "step": 9475
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5153559996539495e-05,
+      "loss": 0.4825,
+      "step": 9476
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5152983245378783e-05,
+      "loss": 0.4847,
+      "step": 9477
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.515240649421807e-05,
+      "loss": 0.6471,
+      "step": 9478
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.515182974305736e-05,
+      "loss": 0.454,
+      "step": 9479
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5151252991896647e-05,
+      "loss": 0.5708,
+      "step": 9480
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5150676240735935e-05,
+      "loss": 0.5998,
+      "step": 9481
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5150099489575223e-05,
+      "loss": 0.5135,
+      "step": 9482
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5149522738414512e-05,
+      "loss": 0.5257,
+      "step": 9483
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.51489459872538e-05,
+      "loss": 0.5424,
+      "step": 9484
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5148369236093088e-05,
+      "loss": 0.5648,
+      "step": 9485
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5147792484932377e-05,
+      "loss": 0.5411,
+      "step": 9486
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5147215733771665e-05,
+      "loss": 0.464,
+      "step": 9487
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5146638982610952e-05,
+      "loss": 0.4506,
+      "step": 9488
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5146062231450243e-05,
+      "loss": 0.4706,
+      "step": 9489
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.514548548028953e-05,
+      "loss": 0.4397,
+      "step": 9490
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5144908729128817e-05,
+      "loss": 0.554,
+      "step": 9491
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5144331977968108e-05,
+      "loss": 0.5873,
+      "step": 9492
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5143755226807396e-05,
+      "loss": 0.6018,
+      "step": 9493
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5143178475646683e-05,
+      "loss": 0.4774,
+      "step": 9494
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5142601724485973e-05,
+      "loss": 0.4512,
+      "step": 9495
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.514202497332526e-05,
+      "loss": 0.4576,
+      "step": 9496
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5141448222164548e-05,
+      "loss": 0.4279,
+      "step": 9497
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5140871471003838e-05,
+      "loss": 0.4287,
+      "step": 9498
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5140294719843125e-05,
+      "loss": 0.5842,
+      "step": 9499
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5139717968682413e-05,
+      "loss": 0.5541,
+      "step": 9500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5139141217521702e-05,
+      "loss": 0.5842,
+      "step": 9501
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.513856446636099e-05,
+      "loss": 0.4573,
+      "step": 9502
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5137987715200278e-05,
+      "loss": 0.5944,
+      "step": 9503
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5137410964039567e-05,
+      "loss": 0.6455,
+      "step": 9504
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5136834212878855e-05,
+      "loss": 0.5525,
+      "step": 9505
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5136257461718142e-05,
+      "loss": 0.5715,
+      "step": 9506
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5135680710557432e-05,
+      "loss": 0.6304,
+      "step": 9507
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.513510395939672e-05,
+      "loss": 0.523,
+      "step": 9508
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5134527208236007e-05,
+      "loss": 0.4841,
+      "step": 9509
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5133950457075297e-05,
+      "loss": 0.4045,
+      "step": 9510
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5133373705914584e-05,
+      "loss": 0.5887,
+      "step": 9511
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5132796954753872e-05,
+      "loss": 0.573,
+      "step": 9512
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5132220203593161e-05,
+      "loss": 0.4672,
+      "step": 9513
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5131643452432449e-05,
+      "loss": 0.5988,
+      "step": 9514
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5131066701271737e-05,
+      "loss": 0.4541,
+      "step": 9515
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5130489950111024e-05,
+      "loss": 0.4482,
+      "step": 9516
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5129913198950314e-05,
+      "loss": 0.4834,
+      "step": 9517
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5129336447789601e-05,
+      "loss": 0.6392,
+      "step": 9518
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512875969662889e-05,
+      "loss": 0.5031,
+      "step": 9519
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512818294546818e-05,
+      "loss": 0.526,
+      "step": 9520
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5127606194307466e-05,
+      "loss": 0.5581,
+      "step": 9521
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5127029443146754e-05,
+      "loss": 0.4793,
+      "step": 9522
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5126452691986045e-05,
+      "loss": 0.4985,
+      "step": 9523
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5125875940825333e-05,
+      "loss": 0.5385,
+      "step": 9524
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512529918966462e-05,
+      "loss": 0.5857,
+      "step": 9525
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512472243850391e-05,
+      "loss": 0.6303,
+      "step": 9526
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5124145687343197e-05,
+      "loss": 0.5071,
+      "step": 9527
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5123568936182485e-05,
+      "loss": 0.6555,
+      "step": 9528
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5122992185021774e-05,
+      "loss": 0.394,
+      "step": 9529
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5122415433861062e-05,
+      "loss": 0.5488,
+      "step": 9530
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512183868270035e-05,
+      "loss": 0.6511,
+      "step": 9531
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.512126193153964e-05,
+      "loss": 0.4923,
+      "step": 9532
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5120685180378927e-05,
+      "loss": 0.6782,
+      "step": 9533
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5120108429218215e-05,
+      "loss": 0.5804,
+      "step": 9534
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5119531678057504e-05,
+      "loss": 0.6004,
+      "step": 9535
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5118954926896792e-05,
+      "loss": 0.4266,
+      "step": 9536
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.511837817573608e-05,
+      "loss": 0.4226,
+      "step": 9537
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5117801424575369e-05,
+      "loss": 0.5314,
+      "step": 9538
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5117224673414656e-05,
+      "loss": 0.5452,
+      "step": 9539
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5116647922253944e-05,
+      "loss": 0.4796,
+      "step": 9540
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5116071171093233e-05,
+      "loss": 0.4478,
+      "step": 9541
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5115494419932521e-05,
+      "loss": 0.4211,
+      "step": 9542
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5114917668771809e-05,
+      "loss": 0.6307,
+      "step": 9543
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5114340917611098e-05,
+      "loss": 0.5479,
+      "step": 9544
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5113764166450386e-05,
+      "loss": 0.4652,
+      "step": 9545
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5113187415289674e-05,
+      "loss": 0.5379,
+      "step": 9546
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5112610664128965e-05,
+      "loss": 0.5439,
+      "step": 9547
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.511203391296825e-05,
+      "loss": 0.4887,
+      "step": 9548
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5111457161807538e-05,
+      "loss": 0.4034,
+      "step": 9549
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.511088041064683e-05,
+      "loss": 0.5768,
+      "step": 9550
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5110303659486117e-05,
+      "loss": 0.5737,
+      "step": 9551
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5109726908325403e-05,
+      "loss": 0.5288,
+      "step": 9552
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.510915015716469e-05,
+      "loss": 0.5679,
+      "step": 9553
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5108573406003982e-05,
+      "loss": 0.5048,
+      "step": 9554
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.510799665484327e-05,
+      "loss": 0.5478,
+      "step": 9555
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5107419903682557e-05,
+      "loss": 0.4977,
+      "step": 9556
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5106843152521847e-05,
+      "loss": 0.6362,
+      "step": 9557
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5106266401361134e-05,
+      "loss": 0.5246,
+      "step": 9558
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5105689650200422e-05,
+      "loss": 0.4181,
+      "step": 9559
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5105112899039711e-05,
+      "loss": 0.6696,
+      "step": 9560
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5104536147878999e-05,
+      "loss": 0.655,
+      "step": 9561
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5103959396718287e-05,
+      "loss": 0.5837,
+      "step": 9562
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 1.5103382645557576e-05,
+      "loss": 0.5441,
+      "step": 9563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5102805894396864e-05,
+      "loss": 0.5446,
+      "step": 9564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5102229143236151e-05,
+      "loss": 0.5856,
+      "step": 9565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5101652392075441e-05,
+      "loss": 0.5125,
+      "step": 9566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5101075640914729e-05,
+      "loss": 0.5416,
+      "step": 9567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5100498889754016e-05,
+      "loss": 0.5509,
+      "step": 9568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5099922138593306e-05,
+      "loss": 0.5385,
+      "step": 9569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5099345387432593e-05,
+      "loss": 0.4676,
+      "step": 9570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5098768636271881e-05,
+      "loss": 0.5317,
+      "step": 9571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.509819188511117e-05,
+      "loss": 0.5665,
+      "step": 9572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5097615133950458e-05,
+      "loss": 0.4636,
+      "step": 9573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5097038382789746e-05,
+      "loss": 0.5313,
+      "step": 9574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5096461631629035e-05,
+      "loss": 0.5201,
+      "step": 9575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5095884880468323e-05,
+      "loss": 0.5751,
+      "step": 9576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.509530812930761e-05,
+      "loss": 0.5029,
+      "step": 9577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5094731378146902e-05,
+      "loss": 0.6069,
+      "step": 9578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5094154626986188e-05,
+      "loss": 0.4791,
+      "step": 9579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5093577875825475e-05,
+      "loss": 0.599,
+      "step": 9580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5093001124664766e-05,
+      "loss": 0.5114,
+      "step": 9581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5092424373504054e-05,
+      "loss": 0.5198,
+      "step": 9582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.509184762234334e-05,
+      "loss": 0.3816,
+      "step": 9583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5091270871182631e-05,
+      "loss": 0.4854,
+      "step": 9584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5090694120021919e-05,
+      "loss": 0.5453,
+      "step": 9585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5090117368861206e-05,
+      "loss": 0.5629,
+      "step": 9586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5089540617700494e-05,
+      "loss": 0.44,
+      "step": 9587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5088963866539783e-05,
+      "loss": 0.5673,
+      "step": 9588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5088387115379071e-05,
+      "loss": 0.5534,
+      "step": 9589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5087810364218359e-05,
+      "loss": 0.5687,
+      "step": 9590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5087233613057648e-05,
+      "loss": 0.5218,
+      "step": 9591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5086656861896936e-05,
+      "loss": 0.465,
+      "step": 9592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5086080110736224e-05,
+      "loss": 0.5338,
+      "step": 9593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5085503359575513e-05,
+      "loss": 0.4173,
+      "step": 9594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.50849266084148e-05,
+      "loss": 0.5671,
+      "step": 9595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5084349857254088e-05,
+      "loss": 0.4416,
+      "step": 9596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5083773106093378e-05,
+      "loss": 0.4601,
+      "step": 9597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5083196354932665e-05,
+      "loss": 0.6258,
+      "step": 9598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5082619603771953e-05,
+      "loss": 0.4109,
+      "step": 9599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5082042852611242e-05,
+      "loss": 0.5098,
+      "step": 9600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.508146610145053e-05,
+      "loss": 0.5107,
+      "step": 9601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5080889350289818e-05,
+      "loss": 0.4836,
+      "step": 9602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5080312599129107e-05,
+      "loss": 0.7183,
+      "step": 9603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5079735847968395e-05,
+      "loss": 0.4811,
+      "step": 9604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5079159096807683e-05,
+      "loss": 0.5808,
+      "step": 9605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5078582345646972e-05,
+      "loss": 0.6013,
+      "step": 9606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.507800559448626e-05,
+      "loss": 0.5354,
+      "step": 9607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5077428843325547e-05,
+      "loss": 0.5394,
+      "step": 9608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5076852092164838e-05,
+      "loss": 0.4325,
+      "step": 9609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5076275341004124e-05,
+      "loss": 0.4079,
+      "step": 9610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5075698589843412e-05,
+      "loss": 0.646,
+      "step": 9611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5075121838682703e-05,
+      "loss": 0.5765,
+      "step": 9612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5074545087521991e-05,
+      "loss": 0.6339,
+      "step": 9613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5073968336361277e-05,
+      "loss": 0.46,
+      "step": 9614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5073391585200568e-05,
+      "loss": 0.4934,
+      "step": 9615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5072814834039856e-05,
+      "loss": 0.5117,
+      "step": 9616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5072238082879143e-05,
+      "loss": 0.5792,
+      "step": 9617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5071661331718433e-05,
+      "loss": 0.4862,
+      "step": 9618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.507108458055772e-05,
+      "loss": 0.6122,
+      "step": 9619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5070507829397008e-05,
+      "loss": 0.5966,
+      "step": 9620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5069931078236296e-05,
+      "loss": 0.4407,
+      "step": 9621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5069354327075585e-05,
+      "loss": 0.5862,
+      "step": 9622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5068777575914873e-05,
+      "loss": 0.4525,
+      "step": 9623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.506820082475416e-05,
+      "loss": 0.6927,
+      "step": 9624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.506762407359345e-05,
+      "loss": 0.6359,
+      "step": 9625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5067047322432738e-05,
+      "loss": 0.5024,
+      "step": 9626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5066470571272025e-05,
+      "loss": 0.6006,
+      "step": 9627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5065893820111315e-05,
+      "loss": 0.5491,
+      "step": 9628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5065317068950602e-05,
+      "loss": 0.4247,
+      "step": 9629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.506474031778989e-05,
+      "loss": 0.5206,
+      "step": 9630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.506416356662918e-05,
+      "loss": 0.5089,
+      "step": 9631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5063586815468467e-05,
+      "loss": 0.4767,
+      "step": 9632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5063010064307755e-05,
+      "loss": 0.6299,
+      "step": 9633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5062433313147044e-05,
+      "loss": 0.5412,
+      "step": 9634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5061856561986332e-05,
+      "loss": 0.5372,
+      "step": 9635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.506127981082562e-05,
+      "loss": 0.4789,
+      "step": 9636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5060703059664909e-05,
+      "loss": 0.6383,
+      "step": 9637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5060126308504197e-05,
+      "loss": 0.5711,
+      "step": 9638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5059549557343484e-05,
+      "loss": 0.5783,
+      "step": 9639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5058972806182775e-05,
+      "loss": 0.4379,
+      "step": 9640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5058396055022061e-05,
+      "loss": 0.4961,
+      "step": 9641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5057819303861349e-05,
+      "loss": 0.479,
+      "step": 9642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.505724255270064e-05,
+      "loss": 0.6351,
+      "step": 9643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5056665801539928e-05,
+      "loss": 0.4042,
+      "step": 9644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5056089050379214e-05,
+      "loss": 0.4524,
+      "step": 9645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5055512299218505e-05,
+      "loss": 0.4967,
+      "step": 9646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5054935548057792e-05,
+      "loss": 0.5272,
+      "step": 9647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.505435879689708e-05,
+      "loss": 0.5541,
+      "step": 9648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.505378204573637e-05,
+      "loss": 0.4714,
+      "step": 9649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5053205294575657e-05,
+      "loss": 0.5249,
+      "step": 9650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5052628543414945e-05,
+      "loss": 0.5169,
+      "step": 9651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5052051792254234e-05,
+      "loss": 0.5348,
+      "step": 9652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5051475041093522e-05,
+      "loss": 0.6333,
+      "step": 9653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.505089828993281e-05,
+      "loss": 0.5163,
+      "step": 9654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5050321538772099e-05,
+      "loss": 0.5081,
+      "step": 9655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5049744787611387e-05,
+      "loss": 0.5939,
+      "step": 9656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5049168036450674e-05,
+      "loss": 0.4637,
+      "step": 9657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5048591285289962e-05,
+      "loss": 0.657,
+      "step": 9658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5048014534129251e-05,
+      "loss": 0.5306,
+      "step": 9659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5047437782968539e-05,
+      "loss": 0.451,
+      "step": 9660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5046861031807827e-05,
+      "loss": 0.4662,
+      "step": 9661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5046284280647116e-05,
+      "loss": 0.6123,
+      "step": 9662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5045707529486404e-05,
+      "loss": 0.5854,
+      "step": 9663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5045130778325692e-05,
+      "loss": 0.5266,
+      "step": 9664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5044554027164981e-05,
+      "loss": 0.5146,
+      "step": 9665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5043977276004269e-05,
+      "loss": 0.4899,
+      "step": 9666
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5043400524843556e-05,
+      "loss": 0.5753,
+      "step": 9667
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5042823773682846e-05,
+      "loss": 0.4419,
+      "step": 9668
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5042247022522133e-05,
+      "loss": 0.5638,
+      "step": 9669
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5041670271361421e-05,
+      "loss": 0.5191,
+      "step": 9670
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5041093520200712e-05,
+      "loss": 0.5843,
+      "step": 9671
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5040516769039998e-05,
+      "loss": 0.4672,
+      "step": 9672
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5039940017879286e-05,
+      "loss": 0.3832,
+      "step": 9673
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5039363266718577e-05,
+      "loss": 0.4661,
+      "step": 9674
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5038786515557865e-05,
+      "loss": 0.5558,
+      "step": 9675
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.503820976439715e-05,
+      "loss": 0.5638,
+      "step": 9676
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5037633013236442e-05,
+      "loss": 0.4955,
+      "step": 9677
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.503705626207573e-05,
+      "loss": 0.4865,
+      "step": 9678
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5036479510915017e-05,
+      "loss": 0.5666,
+      "step": 9679
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5035902759754306e-05,
+      "loss": 0.4974,
+      "step": 9680
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5035326008593594e-05,
+      "loss": 0.4626,
+      "step": 9681
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5034749257432882e-05,
+      "loss": 0.495,
+      "step": 9682
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5034172506272171e-05,
+      "loss": 0.5617,
+      "step": 9683
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5033595755111459e-05,
+      "loss": 0.5836,
+      "step": 9684
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5033019003950747e-05,
+      "loss": 0.646,
+      "step": 9685
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5032442252790036e-05,
+      "loss": 0.5148,
+      "step": 9686
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5031865501629324e-05,
+      "loss": 0.6064,
+      "step": 9687
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5031288750468611e-05,
+      "loss": 0.4387,
+      "step": 9688
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.50307119993079e-05,
+      "loss": 0.4427,
+      "step": 9689
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5030135248147188e-05,
+      "loss": 0.478,
+      "step": 9690
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5029558496986476e-05,
+      "loss": 0.433,
+      "step": 9691
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5028981745825764e-05,
+      "loss": 0.4855,
+      "step": 9692
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5028404994665053e-05,
+      "loss": 0.4658,
+      "step": 9693
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.502782824350434e-05,
+      "loss": 0.6728,
+      "step": 9694
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5027251492343628e-05,
+      "loss": 0.5787,
+      "step": 9695
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5026674741182918e-05,
+      "loss": 0.5756,
+      "step": 9696
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5026097990022206e-05,
+      "loss": 0.3677,
+      "step": 9697
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5025521238861493e-05,
+      "loss": 0.6001,
+      "step": 9698
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5024944487700783e-05,
+      "loss": 0.5446,
+      "step": 9699
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.502436773654007e-05,
+      "loss": 0.5605,
+      "step": 9700
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5023790985379358e-05,
+      "loss": 0.5693,
+      "step": 9701
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5023214234218649e-05,
+      "loss": 0.4682,
+      "step": 9702
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5022637483057935e-05,
+      "loss": 0.4999,
+      "step": 9703
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5022060731897223e-05,
+      "loss": 0.534,
+      "step": 9704
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5021483980736514e-05,
+      "loss": 0.4537,
+      "step": 9705
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5020907229575801e-05,
+      "loss": 0.541,
+      "step": 9706
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5020330478415087e-05,
+      "loss": 0.5089,
+      "step": 9707
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5019753727254379e-05,
+      "loss": 0.5695,
+      "step": 9708
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5019176976093666e-05,
+      "loss": 0.5405,
+      "step": 9709
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5018600224932954e-05,
+      "loss": 0.5125,
+      "step": 9710
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5018023473772243e-05,
+      "loss": 0.6035,
+      "step": 9711
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5017446722611531e-05,
+      "loss": 0.5239,
+      "step": 9712
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5016869971450819e-05,
+      "loss": 0.6208,
+      "step": 9713
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5016293220290108e-05,
+      "loss": 0.5521,
+      "step": 9714
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5015716469129396e-05,
+      "loss": 0.568,
+      "step": 9715
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5015139717968683e-05,
+      "loss": 0.6063,
+      "step": 9716
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5014562966807973e-05,
+      "loss": 0.4967,
+      "step": 9717
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.501398621564726e-05,
+      "loss": 0.5722,
+      "step": 9718
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5013409464486548e-05,
+      "loss": 0.4536,
+      "step": 9719
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5012832713325838e-05,
+      "loss": 0.498,
+      "step": 9720
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5012255962165125e-05,
+      "loss": 0.5519,
+      "step": 9721
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5011679211004413e-05,
+      "loss": 0.3851,
+      "step": 9722
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5011102459843702e-05,
+      "loss": 0.5167,
+      "step": 9723
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.501052570868299e-05,
+      "loss": 0.5465,
+      "step": 9724
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5009948957522278e-05,
+      "loss": 0.5853,
+      "step": 9725
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5009372206361565e-05,
+      "loss": 0.555,
+      "step": 9726
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5008795455200855e-05,
+      "loss": 0.5083,
+      "step": 9727
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5008218704040142e-05,
+      "loss": 0.5674,
+      "step": 9728
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.500764195287943e-05,
+      "loss": 0.4873,
+      "step": 9729
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.500706520171872e-05,
+      "loss": 0.5408,
+      "step": 9730
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5006488450558007e-05,
+      "loss": 0.562,
+      "step": 9731
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5005911699397295e-05,
+      "loss": 0.4785,
+      "step": 9732
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5005334948236586e-05,
+      "loss": 0.4475,
+      "step": 9733
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5004758197075872e-05,
+      "loss": 0.6521,
+      "step": 9734
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.500418144591516e-05,
+      "loss": 0.5987,
+      "step": 9735
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.500360469475445e-05,
+      "loss": 0.5621,
+      "step": 9736
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5003027943593738e-05,
+      "loss": 0.4454,
+      "step": 9737
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5002451192433024e-05,
+      "loss": 0.4961,
+      "step": 9738
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5001874441272315e-05,
+      "loss": 0.4093,
+      "step": 9739
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.5001297690111603e-05,
+      "loss": 0.4219,
+      "step": 9740
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.500072093895089e-05,
+      "loss": 0.4734,
+      "step": 9741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.500014418779018e-05,
+      "loss": 0.5615,
+      "step": 9742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4999567436629468e-05,
+      "loss": 0.521,
+      "step": 9743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4998990685468756e-05,
+      "loss": 0.6461,
+      "step": 9744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4998413934308045e-05,
+      "loss": 0.5715,
+      "step": 9745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4997837183147333e-05,
+      "loss": 0.4609,
+      "step": 9746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.499726043198662e-05,
+      "loss": 0.4365,
+      "step": 9747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.499668368082591e-05,
+      "loss": 0.4345,
+      "step": 9748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4996106929665197e-05,
+      "loss": 0.4954,
+      "step": 9749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4995530178504485e-05,
+      "loss": 0.5104,
+      "step": 9750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4994953427343774e-05,
+      "loss": 0.6276,
+      "step": 9751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4994376676183062e-05,
+      "loss": 0.3942,
+      "step": 9752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.499379992502235e-05,
+      "loss": 0.5888,
+      "step": 9753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4993223173861639e-05,
+      "loss": 0.4804,
+      "step": 9754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4992646422700927e-05,
+      "loss": 0.5703,
+      "step": 9755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4992069671540215e-05,
+      "loss": 0.4523,
+      "step": 9756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4991492920379504e-05,
+      "loss": 0.5992,
+      "step": 9757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4990916169218792e-05,
+      "loss": 0.464,
+      "step": 9758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.499033941805808e-05,
+      "loss": 0.6299,
+      "step": 9759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4989762666897367e-05,
+      "loss": 0.5176,
+      "step": 9760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4989185915736656e-05,
+      "loss": 0.446,
+      "step": 9761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4988609164575944e-05,
+      "loss": 0.4547,
+      "step": 9762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4988032413415232e-05,
+      "loss": 0.6699,
+      "step": 9763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4987455662254523e-05,
+      "loss": 0.5452,
+      "step": 9764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4986878911093809e-05,
+      "loss": 0.5721,
+      "step": 9765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4986302159933096e-05,
+      "loss": 0.5755,
+      "step": 9766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4985725408772388e-05,
+      "loss": 0.442,
+      "step": 9767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4985148657611675e-05,
+      "loss": 0.6548,
+      "step": 9768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4984571906450961e-05,
+      "loss": 0.6131,
+      "step": 9769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4983995155290252e-05,
+      "loss": 0.5655,
+      "step": 9770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.498341840412954e-05,
+      "loss": 0.5335,
+      "step": 9771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4982841652968828e-05,
+      "loss": 0.4289,
+      "step": 9772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4982264901808117e-05,
+      "loss": 0.4688,
+      "step": 9773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4981688150647405e-05,
+      "loss": 0.4106,
+      "step": 9774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4981111399486692e-05,
+      "loss": 0.6701,
+      "step": 9775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4980534648325982e-05,
+      "loss": 0.5092,
+      "step": 9776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.497995789716527e-05,
+      "loss": 0.6319,
+      "step": 9777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4979381146004557e-05,
+      "loss": 0.5929,
+      "step": 9778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4978804394843847e-05,
+      "loss": 0.4617,
+      "step": 9779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4978227643683134e-05,
+      "loss": 0.4841,
+      "step": 9780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4977650892522422e-05,
+      "loss": 0.4293,
+      "step": 9781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4977074141361711e-05,
+      "loss": 0.5749,
+      "step": 9782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4976497390200999e-05,
+      "loss": 0.4077,
+      "step": 9783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4975920639040287e-05,
+      "loss": 0.4912,
+      "step": 9784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4975343887879576e-05,
+      "loss": 0.5721,
+      "step": 9785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4974767136718864e-05,
+      "loss": 0.5911,
+      "step": 9786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4974190385558151e-05,
+      "loss": 0.5524,
+      "step": 9787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.497361363439744e-05,
+      "loss": 0.5499,
+      "step": 9788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4973036883236728e-05,
+      "loss": 0.4501,
+      "step": 9789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4972460132076016e-05,
+      "loss": 0.3743,
+      "step": 9790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4971883380915306e-05,
+      "loss": 0.562,
+      "step": 9791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4971306629754593e-05,
+      "loss": 0.4887,
+      "step": 9792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4970729878593881e-05,
+      "loss": 0.4974,
+      "step": 9793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4970153127433172e-05,
+      "loss": 0.4705,
+      "step": 9794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.496957637627246e-05,
+      "loss": 0.5345,
+      "step": 9795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4968999625111746e-05,
+      "loss": 0.4525,
+      "step": 9796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4968422873951033e-05,
+      "loss": 0.4554,
+      "step": 9797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4967846122790324e-05,
+      "loss": 0.5698,
+      "step": 9798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4967269371629612e-05,
+      "loss": 0.5978,
+      "step": 9799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.49666926204689e-05,
+      "loss": 0.4471,
+      "step": 9800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4966115869308189e-05,
+      "loss": 0.5273,
+      "step": 9801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4965539118147477e-05,
+      "loss": 0.3959,
+      "step": 9802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4964962366986765e-05,
+      "loss": 0.4853,
+      "step": 9803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4964385615826054e-05,
+      "loss": 0.5404,
+      "step": 9804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4963808864665342e-05,
+      "loss": 0.5957,
+      "step": 9805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.496323211350463e-05,
+      "loss": 0.4364,
+      "step": 9806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4962655362343919e-05,
+      "loss": 0.5841,
+      "step": 9807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4962078611183206e-05,
+      "loss": 0.5054,
+      "step": 9808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4961501860022494e-05,
+      "loss": 0.5181,
+      "step": 9809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4960925108861783e-05,
+      "loss": 0.5002,
+      "step": 9810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4960348357701071e-05,
+      "loss": 0.4428,
+      "step": 9811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4959771606540359e-05,
+      "loss": 0.4081,
+      "step": 9812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4959194855379648e-05,
+      "loss": 0.4899,
+      "step": 9813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4958618104218936e-05,
+      "loss": 0.4755,
+      "step": 9814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4958041353058224e-05,
+      "loss": 0.5244,
+      "step": 9815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4957464601897513e-05,
+      "loss": 0.5626,
+      "step": 9816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.49568878507368e-05,
+      "loss": 0.5509,
+      "step": 9817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4956311099576088e-05,
+      "loss": 0.5084,
+      "step": 9818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4955734348415378e-05,
+      "loss": 0.5495,
+      "step": 9819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4955157597254665e-05,
+      "loss": 0.417,
+      "step": 9820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4954580846093953e-05,
+      "loss": 0.5039,
+      "step": 9821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4954004094933242e-05,
+      "loss": 0.5483,
+      "step": 9822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.495342734377253e-05,
+      "loss": 0.5254,
+      "step": 9823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4952850592611818e-05,
+      "loss": 0.4487,
+      "step": 9824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4952273841451109e-05,
+      "loss": 0.5225,
+      "step": 9825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4951697090290397e-05,
+      "loss": 0.5406,
+      "step": 9826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4951120339129683e-05,
+      "loss": 0.5202,
+      "step": 9827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4950543587968974e-05,
+      "loss": 0.5782,
+      "step": 9828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4949966836808261e-05,
+      "loss": 0.5625,
+      "step": 9829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4949390085647549e-05,
+      "loss": 0.4162,
+      "step": 9830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4948813334486837e-05,
+      "loss": 0.4998,
+      "step": 9831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4948236583326126e-05,
+      "loss": 0.4864,
+      "step": 9832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4947659832165414e-05,
+      "loss": 0.5364,
+      "step": 9833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4947083081004701e-05,
+      "loss": 0.4538,
+      "step": 9834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.494650632984399e-05,
+      "loss": 0.5203,
+      "step": 9835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4945929578683278e-05,
+      "loss": 0.6445,
+      "step": 9836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4945352827522566e-05,
+      "loss": 0.4294,
+      "step": 9837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4944776076361856e-05,
+      "loss": 0.5439,
+      "step": 9838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4944199325201143e-05,
+      "loss": 0.5273,
+      "step": 9839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4943622574040431e-05,
+      "loss": 0.5254,
+      "step": 9840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.494304582287972e-05,
+      "loss": 0.5045,
+      "step": 9841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4942469071719008e-05,
+      "loss": 0.4866,
+      "step": 9842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4941892320558296e-05,
+      "loss": 0.4982,
+      "step": 9843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4941315569397585e-05,
+      "loss": 0.5085,
+      "step": 9844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4940738818236873e-05,
+      "loss": 0.4908,
+      "step": 9845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.494016206707616e-05,
+      "loss": 0.4849,
+      "step": 9846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.493958531591545e-05,
+      "loss": 0.5189,
+      "step": 9847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4939008564754737e-05,
+      "loss": 0.5382,
+      "step": 9848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4938431813594025e-05,
+      "loss": 0.6063,
+      "step": 9849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4937855062433315e-05,
+      "loss": 0.4649,
+      "step": 9850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4937278311272602e-05,
+      "loss": 0.5955,
+      "step": 9851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.493670156011189e-05,
+      "loss": 0.5097,
+      "step": 9852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.493612480895118e-05,
+      "loss": 0.5837,
+      "step": 9853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4935548057790467e-05,
+      "loss": 0.5387,
+      "step": 9854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4934971306629755e-05,
+      "loss": 0.4361,
+      "step": 9855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4934394555469046e-05,
+      "loss": 0.5103,
+      "step": 9856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4933817804308333e-05,
+      "loss": 0.4239,
+      "step": 9857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.493324105314762e-05,
+      "loss": 0.4823,
+      "step": 9858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.493266430198691e-05,
+      "loss": 0.4932,
+      "step": 9859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4932087550826198e-05,
+      "loss": 0.5757,
+      "step": 9860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4931510799665486e-05,
+      "loss": 0.6223,
+      "step": 9861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4930934048504775e-05,
+      "loss": 0.5816,
+      "step": 9862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4930357297344063e-05,
+      "loss": 0.4595,
+      "step": 9863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.492978054618335e-05,
+      "loss": 0.5408,
+      "step": 9864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4929203795022638e-05,
+      "loss": 0.5123,
+      "step": 9865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4928627043861928e-05,
+      "loss": 0.5457,
+      "step": 9866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4928050292701215e-05,
+      "loss": 0.5445,
+      "step": 9867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4927473541540503e-05,
+      "loss": 0.5787,
+      "step": 9868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4926896790379792e-05,
+      "loss": 0.6398,
+      "step": 9869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.492632003921908e-05,
+      "loss": 0.4668,
+      "step": 9870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4925743288058368e-05,
+      "loss": 0.7129,
+      "step": 9871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4925166536897657e-05,
+      "loss": 0.5711,
+      "step": 9872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4924589785736945e-05,
+      "loss": 0.6519,
+      "step": 9873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4924013034576233e-05,
+      "loss": 0.4907,
+      "step": 9874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4923436283415522e-05,
+      "loss": 0.579,
+      "step": 9875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.492285953225481e-05,
+      "loss": 0.647,
+      "step": 9876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4922282781094097e-05,
+      "loss": 0.6176,
+      "step": 9877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4921706029933387e-05,
+      "loss": 0.4264,
+      "step": 9878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4921129278772674e-05,
+      "loss": 0.6756,
+      "step": 9879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4920552527611962e-05,
+      "loss": 0.656,
+      "step": 9880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4919975776451251e-05,
+      "loss": 0.5401,
+      "step": 9881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4919399025290539e-05,
+      "loss": 0.6912,
+      "step": 9882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4918822274129827e-05,
+      "loss": 0.4914,
+      "step": 9883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4918245522969118e-05,
+      "loss": 0.4855,
+      "step": 9884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4917668771808404e-05,
+      "loss": 0.533,
+      "step": 9885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4917092020647692e-05,
+      "loss": 0.4627,
+      "step": 9886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4916515269486983e-05,
+      "loss": 0.5442,
+      "step": 9887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.491593851832627e-05,
+      "loss": 0.6269,
+      "step": 9888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4915361767165556e-05,
+      "loss": 0.6004,
+      "step": 9889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4914785016004847e-05,
+      "loss": 0.4879,
+      "step": 9890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4914208264844135e-05,
+      "loss": 0.5635,
+      "step": 9891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4913631513683423e-05,
+      "loss": 0.5343,
+      "step": 9892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4913054762522712e-05,
+      "loss": 0.5519,
+      "step": 9893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4912478011362e-05,
+      "loss": 0.5448,
+      "step": 9894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4911901260201287e-05,
+      "loss": 0.5233,
+      "step": 9895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4911324509040577e-05,
+      "loss": 0.5413,
+      "step": 9896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4910747757879865e-05,
+      "loss": 0.4196,
+      "step": 9897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4910171006719152e-05,
+      "loss": 0.4948,
+      "step": 9898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.490959425555844e-05,
+      "loss": 0.6364,
+      "step": 9899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.490901750439773e-05,
+      "loss": 0.575,
+      "step": 9900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4908440753237017e-05,
+      "loss": 0.5611,
+      "step": 9901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4907864002076305e-05,
+      "loss": 0.4872,
+      "step": 9902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4907287250915594e-05,
+      "loss": 0.5517,
+      "step": 9903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4906710499754882e-05,
+      "loss": 0.3868,
+      "step": 9904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.490613374859417e-05,
+      "loss": 0.5251,
+      "step": 9905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4905556997433459e-05,
+      "loss": 0.6145,
+      "step": 9906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4904980246272746e-05,
+      "loss": 0.3648,
+      "step": 9907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4904403495112034e-05,
+      "loss": 0.5747,
+      "step": 9908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4903826743951324e-05,
+      "loss": 0.618,
+      "step": 9909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4903249992790611e-05,
+      "loss": 0.6408,
+      "step": 9910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4902673241629899e-05,
+      "loss": 0.4359,
+      "step": 9911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4902096490469188e-05,
+      "loss": 0.5602,
+      "step": 9912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4901519739308476e-05,
+      "loss": 0.5004,
+      "step": 9913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4900942988147764e-05,
+      "loss": 0.5853,
+      "step": 9914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4900366236987055e-05,
+      "loss": 0.5949,
+      "step": 9915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.489978948582634e-05,
+      "loss": 0.5482,
+      "step": 9916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4899212734665628e-05,
+      "loss": 0.4783,
+      "step": 9917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.489863598350492e-05,
+      "loss": 0.5699,
+      "step": 9918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4898059232344207e-05,
+      "loss": 0.4017,
+      "step": 9919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 1.4897482481183493e-05,
+      "loss": 0.4974,
+      "step": 9920
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4896905730022784e-05,
+      "loss": 0.3923,
+      "step": 9921
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4896328978862072e-05,
+      "loss": 0.5262,
+      "step": 9922
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.489575222770136e-05,
+      "loss": 0.4173,
+      "step": 9923
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4895175476540649e-05,
+      "loss": 0.5309,
+      "step": 9924
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4894598725379937e-05,
+      "loss": 0.5492,
+      "step": 9925
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4894021974219224e-05,
+      "loss": 0.5178,
+      "step": 9926
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4893445223058514e-05,
+      "loss": 0.6028,
+      "step": 9927
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4892868471897801e-05,
+      "loss": 0.4749,
+      "step": 9928
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4892291720737089e-05,
+      "loss": 0.4825,
+      "step": 9929
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4891714969576378e-05,
+      "loss": 0.4764,
+      "step": 9930
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4891138218415666e-05,
+      "loss": 0.4797,
+      "step": 9931
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4890561467254954e-05,
+      "loss": 0.5565,
+      "step": 9932
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4889984716094243e-05,
+      "loss": 0.5697,
+      "step": 9933
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4889407964933531e-05,
+      "loss": 0.5689,
+      "step": 9934
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4888831213772819e-05,
+      "loss": 0.4917,
+      "step": 9935
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4888254462612106e-05,
+      "loss": 0.6341,
+      "step": 9936
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4887677711451396e-05,
+      "loss": 0.5004,
+      "step": 9937
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4887100960290683e-05,
+      "loss": 0.6448,
+      "step": 9938
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4886524209129971e-05,
+      "loss": 0.4722,
+      "step": 9939
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.488594745796926e-05,
+      "loss": 0.5539,
+      "step": 9940
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4885370706808548e-05,
+      "loss": 0.5359,
+      "step": 9941
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4884793955647836e-05,
+      "loss": 0.5221,
+      "step": 9942
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4884217204487125e-05,
+      "loss": 0.5476,
+      "step": 9943
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4883640453326413e-05,
+      "loss": 0.6073,
+      "step": 9944
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.48830637021657e-05,
+      "loss": 0.5466,
+      "step": 9945
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4882486951004992e-05,
+      "loss": 0.4171,
+      "step": 9946
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4881910199844278e-05,
+      "loss": 0.4968,
+      "step": 9947
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4881333448683565e-05,
+      "loss": 0.5989,
+      "step": 9948
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4880756697522856e-05,
+      "loss": 0.537,
+      "step": 9949
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4880179946362144e-05,
+      "loss": 0.6712,
+      "step": 9950
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.487960319520143e-05,
+      "loss": 0.4777,
+      "step": 9951
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4879026444040721e-05,
+      "loss": 0.5092,
+      "step": 9952
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4878449692880009e-05,
+      "loss": 0.5042,
+      "step": 9953
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4877872941719296e-05,
+      "loss": 0.6199,
+      "step": 9954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4877296190558586e-05,
+      "loss": 0.5298,
+      "step": 9955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4876719439397874e-05,
+      "loss": 0.5943,
+      "step": 9956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4876142688237161e-05,
+      "loss": 0.5333,
+      "step": 9957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.487556593707645e-05,
+      "loss": 0.5498,
+      "step": 9958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4874989185915738e-05,
+      "loss": 0.5332,
+      "step": 9959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4874412434755026e-05,
+      "loss": 0.534,
+      "step": 9960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4873835683594315e-05,
+      "loss": 0.5112,
+      "step": 9961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4873258932433603e-05,
+      "loss": 0.4624,
+      "step": 9962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.487268218127289e-05,
+      "loss": 0.5881,
+      "step": 9963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.487210543011218e-05,
+      "loss": 0.5999,
+      "step": 9964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4871528678951468e-05,
+      "loss": 0.3862,
+      "step": 9965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4870951927790755e-05,
+      "loss": 0.472,
+      "step": 9966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4870375176630045e-05,
+      "loss": 0.6159,
+      "step": 9967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4869798425469333e-05,
+      "loss": 0.6278,
+      "step": 9968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.486922167430862e-05,
+      "loss": 0.4539,
+      "step": 9969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4868644923147908e-05,
+      "loss": 0.5581,
+      "step": 9970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4868068171987197e-05,
+      "loss": 0.5973,
+      "step": 9971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4867491420826485e-05,
+      "loss": 0.5118,
+      "step": 9972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4866914669665773e-05,
+      "loss": 0.5402,
+      "step": 9973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4866337918505062e-05,
+      "loss": 0.5971,
+      "step": 9974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.486576116734435e-05,
+      "loss": 0.5938,
+      "step": 9975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4865184416183637e-05,
+      "loss": 0.5026,
+      "step": 9976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4864607665022928e-05,
+      "loss": 0.536,
+      "step": 9977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4864030913862214e-05,
+      "loss": 0.5628,
+      "step": 9978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4863454162701502e-05,
+      "loss": 0.5038,
+      "step": 9979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4862877411540793e-05,
+      "loss": 0.5327,
+      "step": 9980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4862300660380081e-05,
+      "loss": 0.5515,
+      "step": 9981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4861723909219367e-05,
+      "loss": 0.6459,
+      "step": 9982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4861147158058658e-05,
+      "loss": 0.5098,
+      "step": 9983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4860570406897946e-05,
+      "loss": 0.5177,
+      "step": 9984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4859993655737233e-05,
+      "loss": 0.5255,
+      "step": 9985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4859416904576523e-05,
+      "loss": 0.5106,
+      "step": 9986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.485884015341581e-05,
+      "loss": 0.5066,
+      "step": 9987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4858263402255098e-05,
+      "loss": 0.4369,
+      "step": 9988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4857686651094387e-05,
+      "loss": 0.4618,
+      "step": 9989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4857109899933675e-05,
+      "loss": 0.4221,
+      "step": 9990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4856533148772963e-05,
+      "loss": 0.4813,
+      "step": 9991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4855956397612252e-05,
+      "loss": 0.5628,
+      "step": 9992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.485537964645154e-05,
+      "loss": 0.4991,
+      "step": 9993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4854802895290828e-05,
+      "loss": 0.57,
+      "step": 9994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4854226144130117e-05,
+      "loss": 0.5444,
+      "step": 9995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4853649392969405e-05,
+      "loss": 0.4863,
+      "step": 9996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4853072641808692e-05,
+      "loss": 0.6416,
+      "step": 9997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4852495890647982e-05,
+      "loss": 0.4398,
+      "step": 9998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.485191913948727e-05,
+      "loss": 0.5338,
+      "step": 9999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4851342388326557e-05,
+      "loss": 0.7043,
+      "step": 10000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4850765637165846e-05,
+      "loss": 0.4502,
+      "step": 10001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4850188886005134e-05,
+      "loss": 0.515,
+      "step": 10002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4849612134844422e-05,
+      "loss": 0.4026,
+      "step": 10003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.484903538368371e-05,
+      "loss": 0.4854,
+      "step": 10004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4848458632522999e-05,
+      "loss": 0.5564,
+      "step": 10005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4847881881362287e-05,
+      "loss": 0.589,
+      "step": 10006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4847305130201574e-05,
+      "loss": 0.4552,
+      "step": 10007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4846728379040865e-05,
+      "loss": 0.4997,
+      "step": 10008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4846151627880151e-05,
+      "loss": 0.45,
+      "step": 10009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4845574876719439e-05,
+      "loss": 0.5223,
+      "step": 10010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.484499812555873e-05,
+      "loss": 0.4096,
+      "step": 10011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4844421374398018e-05,
+      "loss": 0.4595,
+      "step": 10012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4843844623237304e-05,
+      "loss": 0.7203,
+      "step": 10013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4843267872076595e-05,
+      "loss": 0.5741,
+      "step": 10014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4842691120915883e-05,
+      "loss": 0.5428,
+      "step": 10015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.484211436975517e-05,
+      "loss": 0.5398,
+      "step": 10016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.484153761859446e-05,
+      "loss": 0.5153,
+      "step": 10017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4840960867433747e-05,
+      "loss": 0.5898,
+      "step": 10018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4840384116273035e-05,
+      "loss": 0.5043,
+      "step": 10019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4839807365112324e-05,
+      "loss": 0.5533,
+      "step": 10020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4839230613951612e-05,
+      "loss": 0.4837,
+      "step": 10021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.48386538627909e-05,
+      "loss": 0.5749,
+      "step": 10022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4838077111630189e-05,
+      "loss": 0.5025,
+      "step": 10023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4837500360469477e-05,
+      "loss": 0.5773,
+      "step": 10024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4836923609308764e-05,
+      "loss": 0.6627,
+      "step": 10025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4836346858148054e-05,
+      "loss": 0.6751,
+      "step": 10026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4835770106987342e-05,
+      "loss": 0.5555,
+      "step": 10027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.483519335582663e-05,
+      "loss": 0.6896,
+      "step": 10028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4834616604665919e-05,
+      "loss": 0.5774,
+      "step": 10029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4834039853505206e-05,
+      "loss": 0.5649,
+      "step": 10030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4833463102344494e-05,
+      "loss": 0.4219,
+      "step": 10031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4832886351183783e-05,
+      "loss": 0.5255,
+      "step": 10032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4832309600023071e-05,
+      "loss": 0.5526,
+      "step": 10033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4831732848862359e-05,
+      "loss": 0.5378,
+      "step": 10034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4831156097701648e-05,
+      "loss": 0.6439,
+      "step": 10035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4830579346540936e-05,
+      "loss": 0.5676,
+      "step": 10036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4830002595380223e-05,
+      "loss": 0.4795,
+      "step": 10037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4829425844219515e-05,
+      "loss": 0.5145,
+      "step": 10038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4828849093058802e-05,
+      "loss": 0.595,
+      "step": 10039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4828272341898088e-05,
+      "loss": 0.4915,
+      "step": 10040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4827695590737376e-05,
+      "loss": 0.3986,
+      "step": 10041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4827118839576667e-05,
+      "loss": 0.532,
+      "step": 10042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4826542088415955e-05,
+      "loss": 0.5939,
+      "step": 10043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.482596533725524e-05,
+      "loss": 0.477,
+      "step": 10044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4825388586094532e-05,
+      "loss": 0.5548,
+      "step": 10045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.482481183493382e-05,
+      "loss": 0.4903,
+      "step": 10046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4824235083773107e-05,
+      "loss": 0.5523,
+      "step": 10047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4823658332612396e-05,
+      "loss": 0.6696,
+      "step": 10048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4823081581451684e-05,
+      "loss": 0.6193,
+      "step": 10049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4822504830290972e-05,
+      "loss": 0.5635,
+      "step": 10050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4821928079130261e-05,
+      "loss": 0.5057,
+      "step": 10051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4821351327969549e-05,
+      "loss": 0.4259,
+      "step": 10052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4820774576808837e-05,
+      "loss": 0.5255,
+      "step": 10053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4820197825648126e-05,
+      "loss": 0.5147,
+      "step": 10054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4819621074487414e-05,
+      "loss": 0.6177,
+      "step": 10055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4819044323326701e-05,
+      "loss": 0.3736,
+      "step": 10056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.481846757216599e-05,
+      "loss": 0.4371,
+      "step": 10057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4817890821005278e-05,
+      "loss": 0.4261,
+      "step": 10058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4817314069844566e-05,
+      "loss": 0.554,
+      "step": 10059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4816737318683855e-05,
+      "loss": 0.5499,
+      "step": 10060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4816160567523143e-05,
+      "loss": 0.4984,
+      "step": 10061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.481558381636243e-05,
+      "loss": 0.4261,
+      "step": 10062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.481500706520172e-05,
+      "loss": 0.4695,
+      "step": 10063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4814430314041008e-05,
+      "loss": 0.481,
+      "step": 10064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4813853562880296e-05,
+      "loss": 0.5458,
+      "step": 10065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4813276811719585e-05,
+      "loss": 0.5577,
+      "step": 10066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4812700060558873e-05,
+      "loss": 0.6371,
+      "step": 10067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.481212330939816e-05,
+      "loss": 0.4697,
+      "step": 10068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4811546558237451e-05,
+      "loss": 0.5229,
+      "step": 10069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4810969807076739e-05,
+      "loss": 0.5002,
+      "step": 10070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4810393055916025e-05,
+      "loss": 0.6046,
+      "step": 10071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4809816304755316e-05,
+      "loss": 0.5629,
+      "step": 10072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4809239553594604e-05,
+      "loss": 0.4784,
+      "step": 10073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4808662802433892e-05,
+      "loss": 0.4199,
+      "step": 10074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4808086051273177e-05,
+      "loss": 0.595,
+      "step": 10075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4807509300112469e-05,
+      "loss": 0.4584,
+      "step": 10076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4806932548951756e-05,
+      "loss": 0.6418,
+      "step": 10077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4806355797791044e-05,
+      "loss": 0.5634,
+      "step": 10078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4805779046630333e-05,
+      "loss": 0.514,
+      "step": 10079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4805202295469621e-05,
+      "loss": 0.5781,
+      "step": 10080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4804625544308909e-05,
+      "loss": 0.5669,
+      "step": 10081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4804048793148198e-05,
+      "loss": 0.4426,
+      "step": 10082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4803472041987486e-05,
+      "loss": 0.4939,
+      "step": 10083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4802895290826773e-05,
+      "loss": 0.4637,
+      "step": 10084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4802318539666063e-05,
+      "loss": 0.6249,
+      "step": 10085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.480174178850535e-05,
+      "loss": 0.503,
+      "step": 10086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4801165037344638e-05,
+      "loss": 0.5542,
+      "step": 10087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4800588286183928e-05,
+      "loss": 0.5881,
+      "step": 10088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4800011535023215e-05,
+      "loss": 0.5553,
+      "step": 10089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4799434783862503e-05,
+      "loss": 0.454,
+      "step": 10090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4798858032701792e-05,
+      "loss": 0.5509,
+      "step": 10091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.479828128154108e-05,
+      "loss": 0.5011,
+      "step": 10092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4797704530380368e-05,
+      "loss": 0.4925,
+      "step": 10093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4797127779219657e-05,
+      "loss": 0.4272,
+      "step": 10094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4796551028058945e-05,
+      "loss": 0.5044,
+      "step": 10095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4795974276898232e-05,
+      "loss": 0.583,
+      "step": 10096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4795397525737522e-05,
+      "loss": 0.5171,
+      "step": 10097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.479482077457681e-05,
+      "loss": 0.4599,
+      "step": 10098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 1.4794244023416097e-05,
+      "loss": 0.4656,
+      "step": 10099
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4793667272255388e-05,
+      "loss": 0.464,
+      "step": 10100
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4793090521094676e-05,
+      "loss": 0.5276,
+      "step": 10101
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4792513769933962e-05,
+      "loss": 0.5348,
+      "step": 10102
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4791937018773253e-05,
+      "loss": 0.62,
+      "step": 10103
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.479136026761254e-05,
+      "loss": 0.5752,
+      "step": 10104
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4790783516451828e-05,
+      "loss": 0.5281,
+      "step": 10105
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4790206765291118e-05,
+      "loss": 0.5037,
+      "step": 10106
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4789630014130405e-05,
+      "loss": 0.3852,
+      "step": 10107
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4789053262969693e-05,
+      "loss": 0.4786,
+      "step": 10108
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.478847651180898e-05,
+      "loss": 0.3977,
+      "step": 10109
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.478789976064827e-05,
+      "loss": 0.5548,
+      "step": 10110
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4787323009487558e-05,
+      "loss": 0.5389,
+      "step": 10111
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4786746258326846e-05,
+      "loss": 0.5681,
+      "step": 10112
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4786169507166135e-05,
+      "loss": 0.5538,
+      "step": 10113
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4785592756005423e-05,
+      "loss": 0.3876,
+      "step": 10114
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.478501600484471e-05,
+      "loss": 0.4656,
+      "step": 10115
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4784439253684e-05,
+      "loss": 0.5334,
+      "step": 10116
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4783862502523287e-05,
+      "loss": 0.4706,
+      "step": 10117
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4783285751362575e-05,
+      "loss": 0.4951,
+      "step": 10118
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4782709000201864e-05,
+      "loss": 0.4788,
+      "step": 10119
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4782132249041152e-05,
+      "loss": 0.586,
+      "step": 10120
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.478155549788044e-05,
+      "loss": 0.4742,
+      "step": 10121
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.478097874671973e-05,
+      "loss": 0.4779,
+      "step": 10122
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4780401995559017e-05,
+      "loss": 0.5094,
+      "step": 10123
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4779825244398305e-05,
+      "loss": 0.5086,
+      "step": 10124
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4779248493237594e-05,
+      "loss": 0.5176,
+      "step": 10125
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4778671742076882e-05,
+      "loss": 0.5313,
+      "step": 10126
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.477809499091617e-05,
+      "loss": 0.5577,
+      "step": 10127
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4777518239755459e-05,
+      "loss": 0.583,
+      "step": 10128
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4776941488594746e-05,
+      "loss": 0.5164,
+      "step": 10129
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4776364737434034e-05,
+      "loss": 0.5588,
+      "step": 10130
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4775787986273325e-05,
+      "loss": 0.41,
+      "step": 10131
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4775211235112613e-05,
+      "loss": 0.456,
+      "step": 10132
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4774634483951899e-05,
+      "loss": 0.4732,
+      "step": 10133
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.477405773279119e-05,
+      "loss": 0.5902,
+      "step": 10134
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4773480981630478e-05,
+      "loss": 0.3769,
+      "step": 10135
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4772904230469765e-05,
+      "loss": 0.5023,
+      "step": 10136
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4772327479309055e-05,
+      "loss": 0.5982,
+      "step": 10137
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4771750728148342e-05,
+      "loss": 0.5518,
+      "step": 10138
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.477117397698763e-05,
+      "loss": 0.5294,
+      "step": 10139
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.477059722582692e-05,
+      "loss": 0.4878,
+      "step": 10140
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4770020474666207e-05,
+      "loss": 0.4079,
+      "step": 10141
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4769443723505495e-05,
+      "loss": 0.5102,
+      "step": 10142
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4768866972344782e-05,
+      "loss": 0.4805,
+      "step": 10143
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4768290221184072e-05,
+      "loss": 0.5003,
+      "step": 10144
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.476771347002336e-05,
+      "loss": 0.5568,
+      "step": 10145
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4767136718862647e-05,
+      "loss": 0.4579,
+      "step": 10146
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4766559967701937e-05,
+      "loss": 0.4719,
+      "step": 10147
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4765983216541224e-05,
+      "loss": 0.5318,
+      "step": 10148
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4765406465380512e-05,
+      "loss": 0.5372,
+      "step": 10149
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4764829714219801e-05,
+      "loss": 0.6373,
+      "step": 10150
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4764252963059089e-05,
+      "loss": 0.5912,
+      "step": 10151
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4763676211898377e-05,
+      "loss": 0.6219,
+      "step": 10152
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4763099460737666e-05,
+      "loss": 0.547,
+      "step": 10153
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4762522709576954e-05,
+      "loss": 0.4388,
+      "step": 10154
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4761945958416241e-05,
+      "loss": 0.5928,
+      "step": 10155
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.476136920725553e-05,
+      "loss": 0.5638,
+      "step": 10156
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4760792456094818e-05,
+      "loss": 0.5501,
+      "step": 10157
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4760215704934106e-05,
+      "loss": 0.4987,
+      "step": 10158
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4759638953773396e-05,
+      "loss": 0.5015,
+      "step": 10159
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4759062202612683e-05,
+      "loss": 0.5561,
+      "step": 10160
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4758485451451971e-05,
+      "loss": 0.5662,
+      "step": 10161
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4757908700291262e-05,
+      "loss": 0.5515,
+      "step": 10162
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.475733194913055e-05,
+      "loss": 0.5541,
+      "step": 10163
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4756755197969836e-05,
+      "loss": 0.4593,
+      "step": 10164
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4756178446809127e-05,
+      "loss": 0.4513,
+      "step": 10165
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4755601695648414e-05,
+      "loss": 0.5522,
+      "step": 10166
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4755024944487702e-05,
+      "loss": 0.4881,
+      "step": 10167
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4754448193326992e-05,
+      "loss": 0.6307,
+      "step": 10168
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.475387144216628e-05,
+      "loss": 0.3668,
+      "step": 10169
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4753294691005567e-05,
+      "loss": 0.5186,
+      "step": 10170
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4752717939844856e-05,
+      "loss": 0.6087,
+      "step": 10171
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4752141188684144e-05,
+      "loss": 0.6526,
+      "step": 10172
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4751564437523432e-05,
+      "loss": 0.4712,
+      "step": 10173
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4750987686362721e-05,
+      "loss": 0.4362,
+      "step": 10174
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4750410935202009e-05,
+      "loss": 0.5108,
+      "step": 10175
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4749834184041296e-05,
+      "loss": 0.6254,
+      "step": 10176
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4749257432880586e-05,
+      "loss": 0.4748,
+      "step": 10177
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4748680681719873e-05,
+      "loss": 0.6633,
+      "step": 10178
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4748103930559161e-05,
+      "loss": 0.5005,
+      "step": 10179
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4747527179398449e-05,
+      "loss": 0.5044,
+      "step": 10180
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4746950428237738e-05,
+      "loss": 0.5097,
+      "step": 10181
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4746373677077026e-05,
+      "loss": 0.4297,
+      "step": 10182
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4745796925916314e-05,
+      "loss": 0.5171,
+      "step": 10183
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4745220174755603e-05,
+      "loss": 0.4298,
+      "step": 10184
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.474464342359489e-05,
+      "loss": 0.5291,
+      "step": 10185
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4744066672434178e-05,
+      "loss": 0.5284,
+      "step": 10186
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4743489921273468e-05,
+      "loss": 0.5567,
+      "step": 10187
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4742913170112755e-05,
+      "loss": 0.4704,
+      "step": 10188
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4742336418952043e-05,
+      "loss": 0.5336,
+      "step": 10189
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4741759667791332e-05,
+      "loss": 0.5513,
+      "step": 10190
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.474118291663062e-05,
+      "loss": 0.503,
+      "step": 10191
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4740606165469908e-05,
+      "loss": 0.5155,
+      "step": 10192
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4740029414309199e-05,
+      "loss": 0.4395,
+      "step": 10193
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4739452663148487e-05,
+      "loss": 0.4807,
+      "step": 10194
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4738875911987773e-05,
+      "loss": 0.5193,
+      "step": 10195
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4738299160827064e-05,
+      "loss": 0.5591,
+      "step": 10196
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4737722409666351e-05,
+      "loss": 0.4277,
+      "step": 10197
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4737145658505639e-05,
+      "loss": 0.4457,
+      "step": 10198
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4736568907344928e-05,
+      "loss": 0.4642,
+      "step": 10199
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4735992156184216e-05,
+      "loss": 0.5888,
+      "step": 10200
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4735415405023504e-05,
+      "loss": 0.4692,
+      "step": 10201
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4734838653862793e-05,
+      "loss": 0.5062,
+      "step": 10202
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.473426190270208e-05,
+      "loss": 0.4587,
+      "step": 10203
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4733685151541368e-05,
+      "loss": 0.5393,
+      "step": 10204
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4733108400380658e-05,
+      "loss": 0.4654,
+      "step": 10205
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4732531649219946e-05,
+      "loss": 0.5312,
+      "step": 10206
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4731954898059233e-05,
+      "loss": 0.4976,
+      "step": 10207
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4731378146898523e-05,
+      "loss": 0.5398,
+      "step": 10208
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.473080139573781e-05,
+      "loss": 0.6074,
+      "step": 10209
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4730224644577098e-05,
+      "loss": 0.5509,
+      "step": 10210
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4729647893416387e-05,
+      "loss": 0.6053,
+      "step": 10211
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4729071142255675e-05,
+      "loss": 0.4781,
+      "step": 10212
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4728494391094963e-05,
+      "loss": 0.5625,
+      "step": 10213
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.472791763993425e-05,
+      "loss": 0.4112,
+      "step": 10214
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.472734088877354e-05,
+      "loss": 0.5284,
+      "step": 10215
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4726764137612827e-05,
+      "loss": 0.46,
+      "step": 10216
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4726187386452115e-05,
+      "loss": 0.3534,
+      "step": 10217
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4725610635291405e-05,
+      "loss": 0.5511,
+      "step": 10218
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4725033884130692e-05,
+      "loss": 0.6136,
+      "step": 10219
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.472445713296998e-05,
+      "loss": 0.5277,
+      "step": 10220
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.472388038180927e-05,
+      "loss": 0.4052,
+      "step": 10221
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4723303630648557e-05,
+      "loss": 0.5021,
+      "step": 10222
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4722726879487845e-05,
+      "loss": 0.4951,
+      "step": 10223
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4722150128327136e-05,
+      "loss": 0.4522,
+      "step": 10224
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4721573377166423e-05,
+      "loss": 0.497,
+      "step": 10225
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.472099662600571e-05,
+      "loss": 0.4356,
+      "step": 10226
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4720419874845e-05,
+      "loss": 0.4715,
+      "step": 10227
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4719843123684288e-05,
+      "loss": 0.6118,
+      "step": 10228
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4719266372523576e-05,
+      "loss": 0.4614,
+      "step": 10229
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4718689621362865e-05,
+      "loss": 0.6081,
+      "step": 10230
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4718112870202153e-05,
+      "loss": 0.4518,
+      "step": 10231
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.471753611904144e-05,
+      "loss": 0.4962,
+      "step": 10232
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.471695936788073e-05,
+      "loss": 0.4068,
+      "step": 10233
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4716382616720018e-05,
+      "loss": 0.521,
+      "step": 10234
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4715805865559305e-05,
+      "loss": 0.507,
+      "step": 10235
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4715229114398595e-05,
+      "loss": 0.408,
+      "step": 10236
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4714652363237882e-05,
+      "loss": 0.4459,
+      "step": 10237
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.471407561207717e-05,
+      "loss": 0.5685,
+      "step": 10238
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.471349886091646e-05,
+      "loss": 0.446,
+      "step": 10239
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4712922109755747e-05,
+      "loss": 0.4872,
+      "step": 10240
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4712345358595035e-05,
+      "loss": 0.5256,
+      "step": 10241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4711768607434324e-05,
+      "loss": 0.5814,
+      "step": 10242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4711191856273612e-05,
+      "loss": 0.5539,
+      "step": 10243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.47106151051129e-05,
+      "loss": 0.3828,
+      "step": 10244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4710038353952189e-05,
+      "loss": 0.4717,
+      "step": 10245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4709461602791477e-05,
+      "loss": 0.494,
+      "step": 10246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4708884851630764e-05,
+      "loss": 0.4521,
+      "step": 10247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4708308100470052e-05,
+      "loss": 0.483,
+      "step": 10248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4707731349309341e-05,
+      "loss": 0.5773,
+      "step": 10249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4707154598148629e-05,
+      "loss": 0.5049,
+      "step": 10250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4706577846987917e-05,
+      "loss": 0.5168,
+      "step": 10251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4706001095827208e-05,
+      "loss": 0.5231,
+      "step": 10252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4705424344666494e-05,
+      "loss": 0.5265,
+      "step": 10253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4704847593505782e-05,
+      "loss": 0.558,
+      "step": 10254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4704270842345073e-05,
+      "loss": 0.5504,
+      "step": 10255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.470369409118436e-05,
+      "loss": 0.4758,
+      "step": 10256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4703117340023646e-05,
+      "loss": 0.5244,
+      "step": 10257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4702540588862937e-05,
+      "loss": 0.6528,
+      "step": 10258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4701963837702225e-05,
+      "loss": 0.5024,
+      "step": 10259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4701387086541513e-05,
+      "loss": 0.5882,
+      "step": 10260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4700810335380802e-05,
+      "loss": 0.5916,
+      "step": 10261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.470023358422009e-05,
+      "loss": 0.5103,
+      "step": 10262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4699656833059377e-05,
+      "loss": 0.5107,
+      "step": 10263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4699080081898667e-05,
+      "loss": 0.6477,
+      "step": 10264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4698503330737955e-05,
+      "loss": 0.5189,
+      "step": 10265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4697926579577242e-05,
+      "loss": 0.4905,
+      "step": 10266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4697349828416532e-05,
+      "loss": 0.4412,
+      "step": 10267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.469677307725582e-05,
+      "loss": 0.5023,
+      "step": 10268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4696196326095107e-05,
+      "loss": 0.4191,
+      "step": 10269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4695619574934396e-05,
+      "loss": 0.6022,
+      "step": 10270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4695042823773684e-05,
+      "loss": 0.4931,
+      "step": 10271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4694466072612972e-05,
+      "loss": 0.5229,
+      "step": 10272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4693889321452261e-05,
+      "loss": 0.4651,
+      "step": 10273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4693312570291549e-05,
+      "loss": 0.5429,
+      "step": 10274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4692735819130836e-05,
+      "loss": 0.56,
+      "step": 10275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4692159067970126e-05,
+      "loss": 0.3335,
+      "step": 10276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4691582316809414e-05,
+      "loss": 0.5158,
+      "step": 10277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.4691005565648701e-05,
+      "loss": 0.316,
+      "step": 10278
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.469042881448799e-05,
+      "loss": 0.413,
+      "step": 10279
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4689852063327278e-05,
+      "loss": 0.3886,
+      "step": 10280
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4689275312166566e-05,
+      "loss": 0.5624,
+      "step": 10281
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4688698561005854e-05,
+      "loss": 0.4624,
+      "step": 10282
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4688121809845145e-05,
+      "loss": 0.4969,
+      "step": 10283
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.468754505868443e-05,
+      "loss": 0.5656,
+      "step": 10284
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4686968307523718e-05,
+      "loss": 0.5814,
+      "step": 10285
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.468639155636301e-05,
+      "loss": 0.4833,
+      "step": 10286
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4685814805202297e-05,
+      "loss": 0.5007,
+      "step": 10287
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4685238054041583e-05,
+      "loss": 0.5196,
+      "step": 10288
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4684661302880874e-05,
+      "loss": 0.5365,
+      "step": 10289
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4684084551720162e-05,
+      "loss": 0.513,
+      "step": 10290
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.468350780055945e-05,
+      "loss": 0.4923,
+      "step": 10291
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4682931049398739e-05,
+      "loss": 0.4671,
+      "step": 10292
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4682354298238027e-05,
+      "loss": 0.4891,
+      "step": 10293
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4681777547077314e-05,
+      "loss": 0.5015,
+      "step": 10294
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4681200795916604e-05,
+      "loss": 0.5761,
+      "step": 10295
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4680624044755891e-05,
+      "loss": 0.482,
+      "step": 10296
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4680047293595179e-05,
+      "loss": 0.5214,
+      "step": 10297
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4679470542434468e-05,
+      "loss": 0.4678,
+      "step": 10298
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4678893791273756e-05,
+      "loss": 0.5563,
+      "step": 10299
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4678317040113044e-05,
+      "loss": 0.4548,
+      "step": 10300
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4677740288952333e-05,
+      "loss": 0.5161,
+      "step": 10301
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4677163537791621e-05,
+      "loss": 0.5404,
+      "step": 10302
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4676586786630909e-05,
+      "loss": 0.4888,
+      "step": 10303
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4676010035470198e-05,
+      "loss": 0.4154,
+      "step": 10304
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4675433284309486e-05,
+      "loss": 0.4667,
+      "step": 10305
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4674856533148773e-05,
+      "loss": 0.5005,
+      "step": 10306
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4674279781988063e-05,
+      "loss": 0.5109,
+      "step": 10307
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.467370303082735e-05,
+      "loss": 0.5406,
+      "step": 10308
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4673126279666638e-05,
+      "loss": 0.5308,
+      "step": 10309
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4672549528505927e-05,
+      "loss": 0.4714,
+      "step": 10310
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4671972777345215e-05,
+      "loss": 0.6086,
+      "step": 10311
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4671396026184503e-05,
+      "loss": 0.6059,
+      "step": 10312
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4670819275023794e-05,
+      "loss": 0.5017,
+      "step": 10313
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4670242523863082e-05,
+      "loss": 0.4539,
+      "step": 10314
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4669665772702368e-05,
+      "loss": 0.5142,
+      "step": 10315
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4669089021541659e-05,
+      "loss": 0.4647,
+      "step": 10316
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4668512270380946e-05,
+      "loss": 0.6153,
+      "step": 10317
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4667935519220234e-05,
+      "loss": 0.477,
+      "step": 10318
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.466735876805952e-05,
+      "loss": 0.456,
+      "step": 10319
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4666782016898811e-05,
+      "loss": 0.5191,
+      "step": 10320
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4666205265738099e-05,
+      "loss": 0.6418,
+      "step": 10321
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4665628514577386e-05,
+      "loss": 0.5023,
+      "step": 10322
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4665051763416676e-05,
+      "loss": 0.5236,
+      "step": 10323
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4664475012255964e-05,
+      "loss": 0.616,
+      "step": 10324
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4663898261095251e-05,
+      "loss": 0.7291,
+      "step": 10325
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.466332150993454e-05,
+      "loss": 0.5441,
+      "step": 10326
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4662744758773828e-05,
+      "loss": 0.6248,
+      "step": 10327
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4662168007613116e-05,
+      "loss": 0.5598,
+      "step": 10328
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4661591256452405e-05,
+      "loss": 0.4934,
+      "step": 10329
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4661014505291693e-05,
+      "loss": 0.426,
+      "step": 10330
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.466043775413098e-05,
+      "loss": 0.511,
+      "step": 10331
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.465986100297027e-05,
+      "loss": 0.5481,
+      "step": 10332
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4659284251809558e-05,
+      "loss": 0.4528,
+      "step": 10333
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4658707500648845e-05,
+      "loss": 0.4493,
+      "step": 10334
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4658130749488135e-05,
+      "loss": 0.5872,
+      "step": 10335
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4657553998327423e-05,
+      "loss": 0.6036,
+      "step": 10336
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.465697724716671e-05,
+      "loss": 0.5729,
+      "step": 10337
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4656400496006e-05,
+      "loss": 0.5384,
+      "step": 10338
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4655823744845287e-05,
+      "loss": 0.4121,
+      "step": 10339
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4655246993684575e-05,
+      "loss": 0.4401,
+      "step": 10340
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4654670242523864e-05,
+      "loss": 0.6773,
+      "step": 10341
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4654093491363152e-05,
+      "loss": 0.4563,
+      "step": 10342
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.465351674020244e-05,
+      "loss": 0.7342,
+      "step": 10343
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.465293998904173e-05,
+      "loss": 0.5125,
+      "step": 10344
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4652363237881018e-05,
+      "loss": 0.5994,
+      "step": 10345
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4651786486720304e-05,
+      "loss": 0.3947,
+      "step": 10346
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4651209735559596e-05,
+      "loss": 0.5225,
+      "step": 10347
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4650632984398883e-05,
+      "loss": 0.664,
+      "step": 10348
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4650056233238171e-05,
+      "loss": 0.5114,
+      "step": 10349
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.464947948207746e-05,
+      "loss": 0.4579,
+      "step": 10350
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4648902730916748e-05,
+      "loss": 0.4679,
+      "step": 10351
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4648325979756036e-05,
+      "loss": 0.474,
+      "step": 10352
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4647749228595323e-05,
+      "loss": 0.5294,
+      "step": 10353
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4647172477434613e-05,
+      "loss": 0.481,
+      "step": 10354
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.46465957262739e-05,
+      "loss": 0.5274,
+      "step": 10355
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4646018975113188e-05,
+      "loss": 0.5165,
+      "step": 10356
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4645442223952477e-05,
+      "loss": 0.3523,
+      "step": 10357
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4644865472791765e-05,
+      "loss": 0.5776,
+      "step": 10358
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4644288721631053e-05,
+      "loss": 0.6725,
+      "step": 10359
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4643711970470342e-05,
+      "loss": 0.601,
+      "step": 10360
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.464313521930963e-05,
+      "loss": 0.5261,
+      "step": 10361
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4642558468148918e-05,
+      "loss": 0.5046,
+      "step": 10362
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4641981716988207e-05,
+      "loss": 0.6145,
+      "step": 10363
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4641404965827495e-05,
+      "loss": 0.5744,
+      "step": 10364
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4640828214666782e-05,
+      "loss": 0.5242,
+      "step": 10365
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4640251463506072e-05,
+      "loss": 0.5815,
+      "step": 10366
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.463967471234536e-05,
+      "loss": 0.4764,
+      "step": 10367
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4639097961184647e-05,
+      "loss": 0.541,
+      "step": 10368
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4638521210023936e-05,
+      "loss": 0.502,
+      "step": 10369
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4637944458863224e-05,
+      "loss": 0.6712,
+      "step": 10370
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4637367707702512e-05,
+      "loss": 0.4473,
+      "step": 10371
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4636790956541801e-05,
+      "loss": 0.4676,
+      "step": 10372
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4636214205381089e-05,
+      "loss": 0.5327,
+      "step": 10373
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4635637454220377e-05,
+      "loss": 0.5803,
+      "step": 10374
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4635060703059668e-05,
+      "loss": 0.5187,
+      "step": 10375
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4634483951898955e-05,
+      "loss": 0.4924,
+      "step": 10376
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4633907200738241e-05,
+      "loss": 0.5793,
+      "step": 10377
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4633330449577532e-05,
+      "loss": 0.5274,
+      "step": 10378
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.463275369841682e-05,
+      "loss": 0.5112,
+      "step": 10379
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4632176947256108e-05,
+      "loss": 0.6541,
+      "step": 10380
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4631600196095397e-05,
+      "loss": 0.4436,
+      "step": 10381
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4631023444934685e-05,
+      "loss": 0.4424,
+      "step": 10382
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4630446693773973e-05,
+      "loss": 0.5896,
+      "step": 10383
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4629869942613262e-05,
+      "loss": 0.5002,
+      "step": 10384
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.462929319145255e-05,
+      "loss": 0.5309,
+      "step": 10385
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4628716440291837e-05,
+      "loss": 0.5456,
+      "step": 10386
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4628139689131125e-05,
+      "loss": 0.5654,
+      "step": 10387
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4627562937970414e-05,
+      "loss": 0.5133,
+      "step": 10388
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4626986186809702e-05,
+      "loss": 0.5647,
+      "step": 10389
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.462640943564899e-05,
+      "loss": 0.5255,
+      "step": 10390
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4625832684488279e-05,
+      "loss": 0.5147,
+      "step": 10391
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4625255933327567e-05,
+      "loss": 0.4779,
+      "step": 10392
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4624679182166854e-05,
+      "loss": 0.5383,
+      "step": 10393
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4624102431006144e-05,
+      "loss": 0.474,
+      "step": 10394
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4623525679845432e-05,
+      "loss": 0.5811,
+      "step": 10395
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.462294892868472e-05,
+      "loss": 0.6262,
+      "step": 10396
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4622372177524009e-05,
+      "loss": 0.5418,
+      "step": 10397
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4621795426363296e-05,
+      "loss": 0.5249,
+      "step": 10398
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4621218675202584e-05,
+      "loss": 0.5149,
+      "step": 10399
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4620641924041873e-05,
+      "loss": 0.6391,
+      "step": 10400
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4620065172881161e-05,
+      "loss": 0.4213,
+      "step": 10401
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4619488421720449e-05,
+      "loss": 0.4864,
+      "step": 10402
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4618911670559738e-05,
+      "loss": 0.4804,
+      "step": 10403
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4618334919399026e-05,
+      "loss": 0.5073,
+      "step": 10404
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4617758168238313e-05,
+      "loss": 0.5674,
+      "step": 10405
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4617181417077605e-05,
+      "loss": 0.4662,
+      "step": 10406
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4616604665916892e-05,
+      "loss": 0.5777,
+      "step": 10407
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4616027914756178e-05,
+      "loss": 0.5455,
+      "step": 10408
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.461545116359547e-05,
+      "loss": 0.4873,
+      "step": 10409
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4614874412434757e-05,
+      "loss": 0.5199,
+      "step": 10410
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4614297661274045e-05,
+      "loss": 0.529,
+      "step": 10411
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4613720910113334e-05,
+      "loss": 0.6437,
+      "step": 10412
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4613144158952622e-05,
+      "loss": 0.4612,
+      "step": 10413
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.461256740779191e-05,
+      "loss": 0.5572,
+      "step": 10414
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4611990656631199e-05,
+      "loss": 0.4376,
+      "step": 10415
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4611413905470486e-05,
+      "loss": 0.4452,
+      "step": 10416
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4610837154309774e-05,
+      "loss": 0.5278,
+      "step": 10417
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4610260403149064e-05,
+      "loss": 0.4462,
+      "step": 10418
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4609683651988351e-05,
+      "loss": 0.5291,
+      "step": 10419
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4609106900827639e-05,
+      "loss": 0.443,
+      "step": 10420
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4608530149666928e-05,
+      "loss": 0.4572,
+      "step": 10421
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4607953398506216e-05,
+      "loss": 0.5037,
+      "step": 10422
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4607376647345504e-05,
+      "loss": 0.5062,
+      "step": 10423
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4606799896184791e-05,
+      "loss": 0.5392,
+      "step": 10424
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.460622314502408e-05,
+      "loss": 0.6328,
+      "step": 10425
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4605646393863368e-05,
+      "loss": 0.4855,
+      "step": 10426
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4605069642702656e-05,
+      "loss": 0.5148,
+      "step": 10427
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4604492891541945e-05,
+      "loss": 0.4081,
+      "step": 10428
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4603916140381233e-05,
+      "loss": 0.4544,
+      "step": 10429
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4603339389220521e-05,
+      "loss": 0.4508,
+      "step": 10430
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.460276263805981e-05,
+      "loss": 0.4831,
+      "step": 10431
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4602185886899098e-05,
+      "loss": 0.544,
+      "step": 10432
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4601609135738386e-05,
+      "loss": 0.4462,
+      "step": 10433
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4601032384577675e-05,
+      "loss": 0.4989,
+      "step": 10434
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4600455633416963e-05,
+      "loss": 0.5662,
+      "step": 10435
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.459987888225625e-05,
+      "loss": 0.5086,
+      "step": 10436
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4599302131095541e-05,
+      "loss": 0.5703,
+      "step": 10437
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4598725379934829e-05,
+      "loss": 0.4634,
+      "step": 10438
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4598148628774115e-05,
+      "loss": 0.581,
+      "step": 10439
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4597571877613406e-05,
+      "loss": 0.5031,
+      "step": 10440
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4596995126452694e-05,
+      "loss": 0.5075,
+      "step": 10441
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4596418375291982e-05,
+      "loss": 0.4973,
+      "step": 10442
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4595841624131271e-05,
+      "loss": 0.5905,
+      "step": 10443
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4595264872970559e-05,
+      "loss": 0.5081,
+      "step": 10444
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4594688121809846e-05,
+      "loss": 0.5408,
+      "step": 10445
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4594111370649136e-05,
+      "loss": 0.5101,
+      "step": 10446
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4593534619488423e-05,
+      "loss": 0.4913,
+      "step": 10447
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4592957868327711e-05,
+      "loss": 0.5337,
+      "step": 10448
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4592381117167e-05,
+      "loss": 0.4821,
+      "step": 10449
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4591804366006288e-05,
+      "loss": 0.5314,
+      "step": 10450
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4591227614845576e-05,
+      "loss": 0.5597,
+      "step": 10451
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4590650863684865e-05,
+      "loss": 0.5648,
+      "step": 10452
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4590074112524153e-05,
+      "loss": 0.4926,
+      "step": 10453
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.458949736136344e-05,
+      "loss": 0.5677,
+      "step": 10454
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.458892061020273e-05,
+      "loss": 0.5386,
+      "step": 10455
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 1.4588343859042018e-05,
+      "loss": 0.4994,
+      "step": 10456
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4587767107881305e-05,
+      "loss": 0.5079,
+      "step": 10457
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4587190356720593e-05,
+      "loss": 0.5374,
+      "step": 10458
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4586613605559882e-05,
+      "loss": 0.5303,
+      "step": 10459
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.458603685439917e-05,
+      "loss": 0.4219,
+      "step": 10460
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4585460103238458e-05,
+      "loss": 0.46,
+      "step": 10461
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4584883352077747e-05,
+      "loss": 0.5838,
+      "step": 10462
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4584306600917035e-05,
+      "loss": 0.5017,
+      "step": 10463
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4583729849756322e-05,
+      "loss": 0.5122,
+      "step": 10464
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4583153098595612e-05,
+      "loss": 0.4449,
+      "step": 10465
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.45825763474349e-05,
+      "loss": 0.4894,
+      "step": 10466
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4581999596274187e-05,
+      "loss": 0.5361,
+      "step": 10467
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4581422845113478e-05,
+      "loss": 0.4667,
+      "step": 10468
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4580846093952766e-05,
+      "loss": 0.3886,
+      "step": 10469
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4580269342792052e-05,
+      "loss": 0.425,
+      "step": 10470
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4579692591631343e-05,
+      "loss": 0.4756,
+      "step": 10471
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.457911584047063e-05,
+      "loss": 0.4027,
+      "step": 10472
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4578539089309918e-05,
+      "loss": 0.48,
+      "step": 10473
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4577962338149208e-05,
+      "loss": 0.5929,
+      "step": 10474
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4577385586988495e-05,
+      "loss": 0.4182,
+      "step": 10475
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4576808835827783e-05,
+      "loss": 0.3868,
+      "step": 10476
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4576232084667073e-05,
+      "loss": 0.6535,
+      "step": 10477
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.457565533350636e-05,
+      "loss": 0.5693,
+      "step": 10478
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4575078582345648e-05,
+      "loss": 0.5713,
+      "step": 10479
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4574501831184937e-05,
+      "loss": 0.4305,
+      "step": 10480
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4573925080024225e-05,
+      "loss": 0.4471,
+      "step": 10481
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4573348328863513e-05,
+      "loss": 0.4424,
+      "step": 10482
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4572771577702802e-05,
+      "loss": 0.4302,
+      "step": 10483
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.457219482654209e-05,
+      "loss": 0.5025,
+      "step": 10484
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4571618075381377e-05,
+      "loss": 0.6201,
+      "step": 10485
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4571041324220667e-05,
+      "loss": 0.3945,
+      "step": 10486
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4570464573059954e-05,
+      "loss": 0.4964,
+      "step": 10487
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4569887821899242e-05,
+      "loss": 0.5257,
+      "step": 10488
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4569311070738532e-05,
+      "loss": 0.4857,
+      "step": 10489
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.456873431957782e-05,
+      "loss": 0.5544,
+      "step": 10490
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4568157568417107e-05,
+      "loss": 0.4414,
+      "step": 10491
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4567580817256395e-05,
+      "loss": 0.5744,
+      "step": 10492
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4567004066095684e-05,
+      "loss": 0.4116,
+      "step": 10493
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4566427314934972e-05,
+      "loss": 0.5939,
+      "step": 10494
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.456585056377426e-05,
+      "loss": 0.4132,
+      "step": 10495
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4565273812613549e-05,
+      "loss": 0.4588,
+      "step": 10496
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4564697061452836e-05,
+      "loss": 0.4652,
+      "step": 10497
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4564120310292124e-05,
+      "loss": 0.5451,
+      "step": 10498
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4563543559131415e-05,
+      "loss": 0.4434,
+      "step": 10499
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4562966807970703e-05,
+      "loss": 0.5763,
+      "step": 10500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4562390056809989e-05,
+      "loss": 0.5204,
+      "step": 10501
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.456181330564928e-05,
+      "loss": 0.6148,
+      "step": 10502
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4561236554488568e-05,
+      "loss": 0.519,
+      "step": 10503
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4560659803327855e-05,
+      "loss": 0.5589,
+      "step": 10504
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4560083052167145e-05,
+      "loss": 0.4994,
+      "step": 10505
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4559506301006432e-05,
+      "loss": 0.3947,
+      "step": 10506
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.455892954984572e-05,
+      "loss": 0.4981,
+      "step": 10507
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.455835279868501e-05,
+      "loss": 0.4331,
+      "step": 10508
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4557776047524297e-05,
+      "loss": 0.6388,
+      "step": 10509
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4557199296363585e-05,
+      "loss": 0.6059,
+      "step": 10510
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4556622545202874e-05,
+      "loss": 0.4442,
+      "step": 10511
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4556045794042162e-05,
+      "loss": 0.5244,
+      "step": 10512
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.455546904288145e-05,
+      "loss": 0.604,
+      "step": 10513
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4554892291720739e-05,
+      "loss": 0.461,
+      "step": 10514
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4554315540560027e-05,
+      "loss": 0.4898,
+      "step": 10515
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4553738789399314e-05,
+      "loss": 0.475,
+      "step": 10516
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4553162038238604e-05,
+      "loss": 0.5307,
+      "step": 10517
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4552585287077891e-05,
+      "loss": 0.4923,
+      "step": 10518
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4552008535917179e-05,
+      "loss": 0.5674,
+      "step": 10519
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4551431784756468e-05,
+      "loss": 0.5037,
+      "step": 10520
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4550855033595756e-05,
+      "loss": 0.5106,
+      "step": 10521
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4550278282435044e-05,
+      "loss": 0.4572,
+      "step": 10522
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4549701531274333e-05,
+      "loss": 0.46,
+      "step": 10523
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4549124780113621e-05,
+      "loss": 0.4556,
+      "step": 10524
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4548548028952909e-05,
+      "loss": 0.5251,
+      "step": 10525
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4547971277792196e-05,
+      "loss": 0.4196,
+      "step": 10526
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4547394526631486e-05,
+      "loss": 0.7173,
+      "step": 10527
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4546817775470773e-05,
+      "loss": 0.472,
+      "step": 10528
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4546241024310061e-05,
+      "loss": 0.5897,
+      "step": 10529
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4545664273149352e-05,
+      "loss": 0.5543,
+      "step": 10530
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.454508752198864e-05,
+      "loss": 0.3733,
+      "step": 10531
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4544510770827926e-05,
+      "loss": 0.5182,
+      "step": 10532
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4543934019667217e-05,
+      "loss": 0.446,
+      "step": 10533
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4543357268506504e-05,
+      "loss": 0.5795,
+      "step": 10534
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4542780517345792e-05,
+      "loss": 0.5425,
+      "step": 10535
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4542203766185082e-05,
+      "loss": 0.5745,
+      "step": 10536
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.454162701502437e-05,
+      "loss": 0.4629,
+      "step": 10537
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4541050263863657e-05,
+      "loss": 0.6467,
+      "step": 10538
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4540473512702946e-05,
+      "loss": 0.5368,
+      "step": 10539
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4539896761542234e-05,
+      "loss": 0.6836,
+      "step": 10540
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4539320010381522e-05,
+      "loss": 0.4904,
+      "step": 10541
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4538743259220811e-05,
+      "loss": 0.6361,
+      "step": 10542
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4538166508060099e-05,
+      "loss": 0.5055,
+      "step": 10543
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4537589756899386e-05,
+      "loss": 0.4653,
+      "step": 10544
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4537013005738676e-05,
+      "loss": 0.4244,
+      "step": 10545
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4536436254577963e-05,
+      "loss": 0.5551,
+      "step": 10546
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4535859503417251e-05,
+      "loss": 0.5573,
+      "step": 10547
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.453528275225654e-05,
+      "loss": 0.4993,
+      "step": 10548
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4534706001095828e-05,
+      "loss": 0.4292,
+      "step": 10549
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4534129249935116e-05,
+      "loss": 0.583,
+      "step": 10550
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4533552498774405e-05,
+      "loss": 0.4836,
+      "step": 10551
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4532975747613693e-05,
+      "loss": 0.5334,
+      "step": 10552
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.453239899645298e-05,
+      "loss": 0.4687,
+      "step": 10553
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.453182224529227e-05,
+      "loss": 0.5466,
+      "step": 10554
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4531245494131558e-05,
+      "loss": 0.7376,
+      "step": 10555
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4530668742970845e-05,
+      "loss": 0.5121,
+      "step": 10556
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4530091991810136e-05,
+      "loss": 0.4323,
+      "step": 10557
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4529515240649422e-05,
+      "loss": 0.602,
+      "step": 10558
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.452893848948871e-05,
+      "loss": 0.5092,
+      "step": 10559
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4528361738328001e-05,
+      "loss": 0.569,
+      "step": 10560
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4527784987167289e-05,
+      "loss": 0.3573,
+      "step": 10561
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4527208236006577e-05,
+      "loss": 0.5148,
+      "step": 10562
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4526631484845863e-05,
+      "loss": 0.4393,
+      "step": 10563
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4526054733685154e-05,
+      "loss": 0.5524,
+      "step": 10564
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4525477982524441e-05,
+      "loss": 0.5871,
+      "step": 10565
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4524901231363729e-05,
+      "loss": 0.5294,
+      "step": 10566
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4524324480203018e-05,
+      "loss": 0.61,
+      "step": 10567
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4523747729042306e-05,
+      "loss": 0.5323,
+      "step": 10568
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4523170977881594e-05,
+      "loss": 0.6158,
+      "step": 10569
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4522594226720883e-05,
+      "loss": 0.4571,
+      "step": 10570
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4522017475560171e-05,
+      "loss": 0.5583,
+      "step": 10571
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4521440724399459e-05,
+      "loss": 0.4548,
+      "step": 10572
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4520863973238748e-05,
+      "loss": 0.7047,
+      "step": 10573
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4520287222078036e-05,
+      "loss": 0.5059,
+      "step": 10574
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4519710470917323e-05,
+      "loss": 0.4992,
+      "step": 10575
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4519133719756613e-05,
+      "loss": 0.4587,
+      "step": 10576
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.45185569685959e-05,
+      "loss": 0.515,
+      "step": 10577
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4517980217435188e-05,
+      "loss": 0.5961,
+      "step": 10578
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4517403466274477e-05,
+      "loss": 0.5682,
+      "step": 10579
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4516826715113765e-05,
+      "loss": 0.6477,
+      "step": 10580
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4516249963953053e-05,
+      "loss": 0.4474,
+      "step": 10581
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4515673212792342e-05,
+      "loss": 0.5945,
+      "step": 10582
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.451509646163163e-05,
+      "loss": 0.6051,
+      "step": 10583
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4514519710470918e-05,
+      "loss": 0.4474,
+      "step": 10584
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4513942959310207e-05,
+      "loss": 0.5262,
+      "step": 10585
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4513366208149495e-05,
+      "loss": 0.4644,
+      "step": 10586
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4512789456988782e-05,
+      "loss": 0.5731,
+      "step": 10587
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4512212705828073e-05,
+      "loss": 0.4793,
+      "step": 10588
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.451163595466736e-05,
+      "loss": 0.5398,
+      "step": 10589
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4511059203506647e-05,
+      "loss": 0.562,
+      "step": 10590
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4510482452345938e-05,
+      "loss": 0.5276,
+      "step": 10591
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4509905701185226e-05,
+      "loss": 0.5018,
+      "step": 10592
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4509328950024513e-05,
+      "loss": 0.4723,
+      "step": 10593
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4508752198863803e-05,
+      "loss": 0.5976,
+      "step": 10594
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.450817544770309e-05,
+      "loss": 0.5282,
+      "step": 10595
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4507598696542378e-05,
+      "loss": 0.5238,
+      "step": 10596
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4507021945381666e-05,
+      "loss": 0.5933,
+      "step": 10597
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4506445194220955e-05,
+      "loss": 0.6104,
+      "step": 10598
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4505868443060243e-05,
+      "loss": 0.5857,
+      "step": 10599
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.450529169189953e-05,
+      "loss": 0.4622,
+      "step": 10600
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.450471494073882e-05,
+      "loss": 0.4933,
+      "step": 10601
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4504138189578108e-05,
+      "loss": 0.5579,
+      "step": 10602
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4503561438417395e-05,
+      "loss": 0.4147,
+      "step": 10603
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4502984687256685e-05,
+      "loss": 0.5279,
+      "step": 10604
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4502407936095972e-05,
+      "loss": 0.4696,
+      "step": 10605
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.450183118493526e-05,
+      "loss": 0.6528,
+      "step": 10606
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.450125443377455e-05,
+      "loss": 0.4368,
+      "step": 10607
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4500677682613837e-05,
+      "loss": 0.4368,
+      "step": 10608
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4500100931453125e-05,
+      "loss": 0.4414,
+      "step": 10609
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4499524180292414e-05,
+      "loss": 0.4335,
+      "step": 10610
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4498947429131702e-05,
+      "loss": 0.4735,
+      "step": 10611
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.449837067797099e-05,
+      "loss": 0.4961,
+      "step": 10612
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4497793926810279e-05,
+      "loss": 0.6276,
+      "step": 10613
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4497217175649567e-05,
+      "loss": 0.4199,
+      "step": 10614
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4496640424488854e-05,
+      "loss": 0.4456,
+      "step": 10615
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4496063673328144e-05,
+      "loss": 0.5981,
+      "step": 10616
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4495486922167431e-05,
+      "loss": 0.5671,
+      "step": 10617
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.449491017100672e-05,
+      "loss": 0.4501,
+      "step": 10618
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.449433341984601e-05,
+      "loss": 0.5975,
+      "step": 10619
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4493756668685298e-05,
+      "loss": 0.5612,
+      "step": 10620
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4493179917524584e-05,
+      "loss": 0.5226,
+      "step": 10621
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4492603166363875e-05,
+      "loss": 0.4285,
+      "step": 10622
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4492026415203163e-05,
+      "loss": 0.5991,
+      "step": 10623
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.449144966404245e-05,
+      "loss": 0.5283,
+      "step": 10624
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.449087291288174e-05,
+      "loss": 0.7284,
+      "step": 10625
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4490296161721027e-05,
+      "loss": 0.5479,
+      "step": 10626
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4489719410560315e-05,
+      "loss": 0.6068,
+      "step": 10627
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4489142659399604e-05,
+      "loss": 0.5882,
+      "step": 10628
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4488565908238892e-05,
+      "loss": 0.484,
+      "step": 10629
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.448798915707818e-05,
+      "loss": 0.586,
+      "step": 10630
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4487412405917468e-05,
+      "loss": 0.5536,
+      "step": 10631
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4486835654756757e-05,
+      "loss": 0.5204,
+      "step": 10632
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4486258903596045e-05,
+      "loss": 0.5612,
+      "step": 10633
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4485682152435332e-05,
+      "loss": 0.5156,
+      "step": 10634
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 1.4485105401274622e-05,
+      "loss": 0.6611,
+      "step": 10635
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.448452865011391e-05,
+      "loss": 0.5812,
+      "step": 10636
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4483951898953197e-05,
+      "loss": 0.4987,
+      "step": 10637
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4483375147792486e-05,
+      "loss": 0.5525,
+      "step": 10638
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4482798396631774e-05,
+      "loss": 0.5829,
+      "step": 10639
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4482221645471062e-05,
+      "loss": 0.5683,
+      "step": 10640
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4481644894310351e-05,
+      "loss": 0.4438,
+      "step": 10641
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4481068143149639e-05,
+      "loss": 0.4388,
+      "step": 10642
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4480491391988927e-05,
+      "loss": 0.5436,
+      "step": 10643
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4479914640828216e-05,
+      "loss": 0.4423,
+      "step": 10644
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4479337889667504e-05,
+      "loss": 0.5647,
+      "step": 10645
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4478761138506791e-05,
+      "loss": 0.6191,
+      "step": 10646
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.447818438734608e-05,
+      "loss": 0.4948,
+      "step": 10647
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4477607636185368e-05,
+      "loss": 0.4183,
+      "step": 10648
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4477030885024656e-05,
+      "loss": 0.447,
+      "step": 10649
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4476454133863947e-05,
+      "loss": 0.6195,
+      "step": 10650
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4475877382703235e-05,
+      "loss": 0.4951,
+      "step": 10651
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.447530063154252e-05,
+      "loss": 0.5343,
+      "step": 10652
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4474723880381812e-05,
+      "loss": 0.5269,
+      "step": 10653
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.44741471292211e-05,
+      "loss": 0.4747,
+      "step": 10654
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4473570378060387e-05,
+      "loss": 0.5783,
+      "step": 10655
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4472993626899677e-05,
+      "loss": 0.4655,
+      "step": 10656
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4472416875738964e-05,
+      "loss": 0.5114,
+      "step": 10657
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4471840124578252e-05,
+      "loss": 0.5358,
+      "step": 10658
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4471263373417541e-05,
+      "loss": 0.5474,
+      "step": 10659
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4470686622256829e-05,
+      "loss": 0.4692,
+      "step": 10660
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4470109871096117e-05,
+      "loss": 0.5468,
+      "step": 10661
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4469533119935406e-05,
+      "loss": 0.5358,
+      "step": 10662
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4468956368774694e-05,
+      "loss": 0.4084,
+      "step": 10663
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4468379617613981e-05,
+      "loss": 0.5483,
+      "step": 10664
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.446780286645327e-05,
+      "loss": 0.583,
+      "step": 10665
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4467226115292559e-05,
+      "loss": 0.5641,
+      "step": 10666
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4466649364131846e-05,
+      "loss": 0.4238,
+      "step": 10667
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4466072612971134e-05,
+      "loss": 0.5166,
+      "step": 10668
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4465495861810423e-05,
+      "loss": 0.4349,
+      "step": 10669
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4464919110649711e-05,
+      "loss": 0.4758,
+      "step": 10670
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4464342359488999e-05,
+      "loss": 0.6051,
+      "step": 10671
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4463765608328288e-05,
+      "loss": 0.6541,
+      "step": 10672
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4463188857167576e-05,
+      "loss": 0.5152,
+      "step": 10673
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4462612106006863e-05,
+      "loss": 0.5593,
+      "step": 10674
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4462035354846153e-05,
+      "loss": 0.4324,
+      "step": 10675
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.446145860368544e-05,
+      "loss": 0.4763,
+      "step": 10676
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4460881852524728e-05,
+      "loss": 0.7122,
+      "step": 10677
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4460305101364018e-05,
+      "loss": 0.4809,
+      "step": 10678
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4459728350203305e-05,
+      "loss": 0.4738,
+      "step": 10679
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4459151599042593e-05,
+      "loss": 0.5828,
+      "step": 10680
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4458574847881884e-05,
+      "loss": 0.4986,
+      "step": 10681
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4457998096721172e-05,
+      "loss": 0.5478,
+      "step": 10682
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4457421345560458e-05,
+      "loss": 0.5691,
+      "step": 10683
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4456844594399749e-05,
+      "loss": 0.5535,
+      "step": 10684
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4456267843239036e-05,
+      "loss": 0.5442,
+      "step": 10685
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4455691092078324e-05,
+      "loss": 0.4822,
+      "step": 10686
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4455114340917613e-05,
+      "loss": 0.5957,
+      "step": 10687
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4454537589756901e-05,
+      "loss": 0.3577,
+      "step": 10688
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4453960838596189e-05,
+      "loss": 0.5577,
+      "step": 10689
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4453384087435478e-05,
+      "loss": 0.5021,
+      "step": 10690
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4452807336274766e-05,
+      "loss": 0.5189,
+      "step": 10691
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4452230585114054e-05,
+      "loss": 0.5538,
+      "step": 10692
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4451653833953343e-05,
+      "loss": 0.5311,
+      "step": 10693
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.445107708279263e-05,
+      "loss": 0.5278,
+      "step": 10694
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4450500331631918e-05,
+      "loss": 0.5124,
+      "step": 10695
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4449923580471208e-05,
+      "loss": 0.4779,
+      "step": 10696
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4449346829310495e-05,
+      "loss": 0.5415,
+      "step": 10697
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4448770078149783e-05,
+      "loss": 0.5269,
+      "step": 10698
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4448193326989072e-05,
+      "loss": 0.6123,
+      "step": 10699
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.444761657582836e-05,
+      "loss": 0.5205,
+      "step": 10700
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4447039824667648e-05,
+      "loss": 0.6053,
+      "step": 10701
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4446463073506936e-05,
+      "loss": 0.3215,
+      "step": 10702
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4445886322346225e-05,
+      "loss": 0.5201,
+      "step": 10703
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4445309571185513e-05,
+      "loss": 0.5104,
+      "step": 10704
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.44447328200248e-05,
+      "loss": 0.5048,
+      "step": 10705
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.444415606886409e-05,
+      "loss": 0.5193,
+      "step": 10706
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4443579317703377e-05,
+      "loss": 0.5192,
+      "step": 10707
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4443002566542665e-05,
+      "loss": 0.4307,
+      "step": 10708
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4442425815381954e-05,
+      "loss": 0.4058,
+      "step": 10709
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4441849064221242e-05,
+      "loss": 0.4355,
+      "step": 10710
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.444127231306053e-05,
+      "loss": 0.6244,
+      "step": 10711
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4440695561899821e-05,
+      "loss": 0.5022,
+      "step": 10712
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4440118810739109e-05,
+      "loss": 0.5255,
+      "step": 10713
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4439542059578395e-05,
+      "loss": 0.4791,
+      "step": 10714
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4438965308417686e-05,
+      "loss": 0.5146,
+      "step": 10715
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4438388557256973e-05,
+      "loss": 0.3701,
+      "step": 10716
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4437811806096261e-05,
+      "loss": 0.6917,
+      "step": 10717
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.443723505493555e-05,
+      "loss": 0.5453,
+      "step": 10718
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4436658303774838e-05,
+      "loss": 0.4432,
+      "step": 10719
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4436081552614126e-05,
+      "loss": 0.5862,
+      "step": 10720
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4435504801453415e-05,
+      "loss": 0.6074,
+      "step": 10721
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4434928050292703e-05,
+      "loss": 0.4759,
+      "step": 10722
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.443435129913199e-05,
+      "loss": 0.4196,
+      "step": 10723
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.443377454797128e-05,
+      "loss": 0.5887,
+      "step": 10724
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4433197796810568e-05,
+      "loss": 0.5176,
+      "step": 10725
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4432621045649855e-05,
+      "loss": 0.5588,
+      "step": 10726
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4432044294489145e-05,
+      "loss": 0.4243,
+      "step": 10727
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4431467543328432e-05,
+      "loss": 0.5148,
+      "step": 10728
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.443089079216772e-05,
+      "loss": 0.4752,
+      "step": 10729
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.443031404100701e-05,
+      "loss": 0.547,
+      "step": 10730
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4429737289846297e-05,
+      "loss": 0.5806,
+      "step": 10731
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4429160538685585e-05,
+      "loss": 0.505,
+      "step": 10732
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4428583787524874e-05,
+      "loss": 0.499,
+      "step": 10733
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4428007036364162e-05,
+      "loss": 0.5178,
+      "step": 10734
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.442743028520345e-05,
+      "loss": 0.4788,
+      "step": 10735
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4426853534042737e-05,
+      "loss": 0.4798,
+      "step": 10736
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4426276782882027e-05,
+      "loss": 0.3679,
+      "step": 10737
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4425700031721314e-05,
+      "loss": 0.478,
+      "step": 10738
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4425123280560602e-05,
+      "loss": 0.411,
+      "step": 10739
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4424546529399891e-05,
+      "loss": 0.482,
+      "step": 10740
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4423969778239179e-05,
+      "loss": 0.5143,
+      "step": 10741
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4423393027078467e-05,
+      "loss": 0.5651,
+      "step": 10742
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4422816275917758e-05,
+      "loss": 0.459,
+      "step": 10743
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4422239524757045e-05,
+      "loss": 0.4334,
+      "step": 10744
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4421662773596331e-05,
+      "loss": 0.5114,
+      "step": 10745
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4421086022435622e-05,
+      "loss": 0.4041,
+      "step": 10746
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.442050927127491e-05,
+      "loss": 0.4779,
+      "step": 10747
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4419932520114198e-05,
+      "loss": 0.5123,
+      "step": 10748
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4419355768953487e-05,
+      "loss": 0.5875,
+      "step": 10749
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4418779017792775e-05,
+      "loss": 0.3323,
+      "step": 10750
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4418202266632063e-05,
+      "loss": 0.5779,
+      "step": 10751
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4417625515471352e-05,
+      "loss": 0.712,
+      "step": 10752
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.441704876431064e-05,
+      "loss": 0.4922,
+      "step": 10753
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4416472013149927e-05,
+      "loss": 0.5067,
+      "step": 10754
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4415895261989217e-05,
+      "loss": 0.5252,
+      "step": 10755
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4415318510828504e-05,
+      "loss": 0.5012,
+      "step": 10756
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4414741759667792e-05,
+      "loss": 0.4749,
+      "step": 10757
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4414165008507081e-05,
+      "loss": 0.5237,
+      "step": 10758
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.441358825734637e-05,
+      "loss": 0.4627,
+      "step": 10759
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4413011506185657e-05,
+      "loss": 0.4367,
+      "step": 10760
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4412434755024946e-05,
+      "loss": 0.4525,
+      "step": 10761
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4411858003864234e-05,
+      "loss": 0.4516,
+      "step": 10762
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4411281252703522e-05,
+      "loss": 0.517,
+      "step": 10763
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4410704501542811e-05,
+      "loss": 0.515,
+      "step": 10764
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4410127750382099e-05,
+      "loss": 0.4432,
+      "step": 10765
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4409550999221386e-05,
+      "loss": 0.669,
+      "step": 10766
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4408974248060676e-05,
+      "loss": 0.5562,
+      "step": 10767
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4408397496899963e-05,
+      "loss": 0.5525,
+      "step": 10768
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4407820745739251e-05,
+      "loss": 0.4235,
+      "step": 10769
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4407243994578539e-05,
+      "loss": 0.5414,
+      "step": 10770
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4406667243417828e-05,
+      "loss": 0.622,
+      "step": 10771
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4406090492257116e-05,
+      "loss": 0.5637,
+      "step": 10772
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4405513741096404e-05,
+      "loss": 0.4354,
+      "step": 10773
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4404936989935695e-05,
+      "loss": 0.4766,
+      "step": 10774
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4404360238774982e-05,
+      "loss": 0.6017,
+      "step": 10775
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4403783487614268e-05,
+      "loss": 0.5512,
+      "step": 10776
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.440320673645356e-05,
+      "loss": 0.4384,
+      "step": 10777
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4402629985292847e-05,
+      "loss": 0.4998,
+      "step": 10778
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4402053234132135e-05,
+      "loss": 0.4503,
+      "step": 10779
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4401476482971424e-05,
+      "loss": 0.5641,
+      "step": 10780
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4400899731810712e-05,
+      "loss": 0.4721,
+      "step": 10781
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.440032298065e-05,
+      "loss": 0.5466,
+      "step": 10782
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4399746229489289e-05,
+      "loss": 0.6578,
+      "step": 10783
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4399169478328577e-05,
+      "loss": 0.6324,
+      "step": 10784
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4398592727167864e-05,
+      "loss": 0.5243,
+      "step": 10785
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4398015976007154e-05,
+      "loss": 0.6007,
+      "step": 10786
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4397439224846441e-05,
+      "loss": 0.495,
+      "step": 10787
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4396862473685729e-05,
+      "loss": 0.5835,
+      "step": 10788
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4396285722525018e-05,
+      "loss": 0.505,
+      "step": 10789
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4395708971364306e-05,
+      "loss": 0.5896,
+      "step": 10790
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4395132220203594e-05,
+      "loss": 0.4947,
+      "step": 10791
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4394555469042883e-05,
+      "loss": 0.5891,
+      "step": 10792
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.439397871788217e-05,
+      "loss": 0.4604,
+      "step": 10793
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4393401966721458e-05,
+      "loss": 0.5127,
+      "step": 10794
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4392825215560748e-05,
+      "loss": 0.4962,
+      "step": 10795
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4392248464400036e-05,
+      "loss": 0.4169,
+      "step": 10796
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4391671713239323e-05,
+      "loss": 0.4431,
+      "step": 10797
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4391094962078613e-05,
+      "loss": 0.5649,
+      "step": 10798
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.43905182109179e-05,
+      "loss": 0.5753,
+      "step": 10799
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4389941459757188e-05,
+      "loss": 0.523,
+      "step": 10800
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4389364708596479e-05,
+      "loss": 0.5653,
+      "step": 10801
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4388787957435765e-05,
+      "loss": 0.4062,
+      "step": 10802
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4388211206275053e-05,
+      "loss": 0.4784,
+      "step": 10803
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4387634455114344e-05,
+      "loss": 0.4698,
+      "step": 10804
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4387057703953631e-05,
+      "loss": 0.5348,
+      "step": 10805
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.438648095279292e-05,
+      "loss": 0.5855,
+      "step": 10806
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4385904201632205e-05,
+      "loss": 0.5785,
+      "step": 10807
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4385327450471496e-05,
+      "loss": 0.5503,
+      "step": 10808
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4384750699310784e-05,
+      "loss": 0.5604,
+      "step": 10809
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4384173948150072e-05,
+      "loss": 0.5038,
+      "step": 10810
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4383597196989361e-05,
+      "loss": 0.4721,
+      "step": 10811
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4383020445828649e-05,
+      "loss": 0.513,
+      "step": 10812
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4382443694667936e-05,
+      "loss": 0.4989,
+      "step": 10813
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.4381866943507226e-05,
+      "loss": 0.4581,
+      "step": 10814
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4381290192346513e-05,
+      "loss": 0.4987,
+      "step": 10815
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4380713441185801e-05,
+      "loss": 0.5476,
+      "step": 10816
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.438013669002509e-05,
+      "loss": 0.6298,
+      "step": 10817
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4379559938864378e-05,
+      "loss": 0.6348,
+      "step": 10818
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4378983187703666e-05,
+      "loss": 0.4907,
+      "step": 10819
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4378406436542955e-05,
+      "loss": 0.4637,
+      "step": 10820
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4377829685382243e-05,
+      "loss": 0.4876,
+      "step": 10821
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.437725293422153e-05,
+      "loss": 0.5324,
+      "step": 10822
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.437667618306082e-05,
+      "loss": 0.5293,
+      "step": 10823
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4376099431900108e-05,
+      "loss": 0.4752,
+      "step": 10824
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4375522680739395e-05,
+      "loss": 0.491,
+      "step": 10825
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4374945929578685e-05,
+      "loss": 0.4454,
+      "step": 10826
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4374369178417972e-05,
+      "loss": 0.4806,
+      "step": 10827
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.437379242725726e-05,
+      "loss": 0.6485,
+      "step": 10828
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.437321567609655e-05,
+      "loss": 0.4149,
+      "step": 10829
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4372638924935837e-05,
+      "loss": 0.7086,
+      "step": 10830
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4372062173775125e-05,
+      "loss": 0.4152,
+      "step": 10831
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4371485422614416e-05,
+      "loss": 0.4872,
+      "step": 10832
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4370908671453702e-05,
+      "loss": 0.5971,
+      "step": 10833
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.437033192029299e-05,
+      "loss": 0.4809,
+      "step": 10834
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.436975516913228e-05,
+      "loss": 0.6724,
+      "step": 10835
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4369178417971568e-05,
+      "loss": 0.5552,
+      "step": 10836
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4368601666810856e-05,
+      "loss": 0.4709,
+      "step": 10837
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4368024915650145e-05,
+      "loss": 0.4321,
+      "step": 10838
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4367448164489433e-05,
+      "loss": 0.4295,
+      "step": 10839
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.436687141332872e-05,
+      "loss": 0.5177,
+      "step": 10840
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4366294662168008e-05,
+      "loss": 0.4269,
+      "step": 10841
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4365717911007298e-05,
+      "loss": 0.448,
+      "step": 10842
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4365141159846586e-05,
+      "loss": 0.407,
+      "step": 10843
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4364564408685873e-05,
+      "loss": 0.5295,
+      "step": 10844
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4363987657525163e-05,
+      "loss": 0.5487,
+      "step": 10845
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.436341090636445e-05,
+      "loss": 0.4304,
+      "step": 10846
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4362834155203738e-05,
+      "loss": 0.571,
+      "step": 10847
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4362257404043027e-05,
+      "loss": 0.4902,
+      "step": 10848
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4361680652882315e-05,
+      "loss": 0.5942,
+      "step": 10849
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4361103901721603e-05,
+      "loss": 0.5391,
+      "step": 10850
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4360527150560892e-05,
+      "loss": 0.5705,
+      "step": 10851
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.435995039940018e-05,
+      "loss": 0.4772,
+      "step": 10852
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4359373648239467e-05,
+      "loss": 0.4892,
+      "step": 10853
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4358796897078757e-05,
+      "loss": 0.5557,
+      "step": 10854
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4358220145918045e-05,
+      "loss": 0.5503,
+      "step": 10855
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4357643394757332e-05,
+      "loss": 0.5194,
+      "step": 10856
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4357066643596622e-05,
+      "loss": 0.4728,
+      "step": 10857
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.435648989243591e-05,
+      "loss": 0.4995,
+      "step": 10858
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4355913141275197e-05,
+      "loss": 0.4957,
+      "step": 10859
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4355336390114486e-05,
+      "loss": 0.5002,
+      "step": 10860
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4354759638953774e-05,
+      "loss": 0.508,
+      "step": 10861
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4354182887793062e-05,
+      "loss": 0.4009,
+      "step": 10862
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4353606136632353e-05,
+      "loss": 0.5974,
+      "step": 10863
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4353029385471639e-05,
+      "loss": 0.4906,
+      "step": 10864
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4352452634310926e-05,
+      "loss": 0.5661,
+      "step": 10865
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4351875883150218e-05,
+      "loss": 0.3951,
+      "step": 10866
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4351299131989505e-05,
+      "loss": 0.5348,
+      "step": 10867
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4350722380828793e-05,
+      "loss": 0.6186,
+      "step": 10868
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4350145629668082e-05,
+      "loss": 0.329,
+      "step": 10869
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.434956887850737e-05,
+      "loss": 0.4775,
+      "step": 10870
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4348992127346658e-05,
+      "loss": 0.5214,
+      "step": 10871
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4348415376185947e-05,
+      "loss": 0.3894,
+      "step": 10872
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4347838625025235e-05,
+      "loss": 0.5198,
+      "step": 10873
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4347261873864522e-05,
+      "loss": 0.5169,
+      "step": 10874
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.434668512270381e-05,
+      "loss": 0.3418,
+      "step": 10875
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.43461083715431e-05,
+      "loss": 0.3317,
+      "step": 10876
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4345531620382387e-05,
+      "loss": 0.5993,
+      "step": 10877
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4344954869221675e-05,
+      "loss": 0.5363,
+      "step": 10878
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4344378118060964e-05,
+      "loss": 0.562,
+      "step": 10879
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4343801366900252e-05,
+      "loss": 0.3937,
+      "step": 10880
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.434322461573954e-05,
+      "loss": 0.5046,
+      "step": 10881
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4342647864578829e-05,
+      "loss": 0.5616,
+      "step": 10882
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4342071113418117e-05,
+      "loss": 0.5348,
+      "step": 10883
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4341494362257404e-05,
+      "loss": 0.5109,
+      "step": 10884
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4340917611096694e-05,
+      "loss": 0.5364,
+      "step": 10885
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4340340859935981e-05,
+      "loss": 0.657,
+      "step": 10886
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4339764108775269e-05,
+      "loss": 0.5009,
+      "step": 10887
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4339187357614558e-05,
+      "loss": 0.4913,
+      "step": 10888
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4338610606453846e-05,
+      "loss": 0.6016,
+      "step": 10889
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4338033855293134e-05,
+      "loss": 0.595,
+      "step": 10890
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4337457104132423e-05,
+      "loss": 0.4822,
+      "step": 10891
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4336880352971711e-05,
+      "loss": 0.4543,
+      "step": 10892
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4336303601810999e-05,
+      "loss": 0.5046,
+      "step": 10893
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.433572685065029e-05,
+      "loss": 0.3614,
+      "step": 10894
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4335150099489576e-05,
+      "loss": 0.5609,
+      "step": 10895
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4334573348328863e-05,
+      "loss": 0.568,
+      "step": 10896
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4333996597168154e-05,
+      "loss": 0.4653,
+      "step": 10897
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4333419846007442e-05,
+      "loss": 0.4821,
+      "step": 10898
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.433284309484673e-05,
+      "loss": 0.4244,
+      "step": 10899
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4332266343686019e-05,
+      "loss": 0.5007,
+      "step": 10900
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4331689592525307e-05,
+      "loss": 0.4405,
+      "step": 10901
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4331112841364595e-05,
+      "loss": 0.4731,
+      "step": 10902
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4330536090203884e-05,
+      "loss": 0.5178,
+      "step": 10903
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4329959339043172e-05,
+      "loss": 0.4772,
+      "step": 10904
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.432938258788246e-05,
+      "loss": 0.4671,
+      "step": 10905
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4328805836721749e-05,
+      "loss": 0.5498,
+      "step": 10906
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4328229085561036e-05,
+      "loss": 0.5216,
+      "step": 10907
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4327652334400324e-05,
+      "loss": 0.5885,
+      "step": 10908
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4327075583239612e-05,
+      "loss": 0.4104,
+      "step": 10909
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4326498832078901e-05,
+      "loss": 0.4567,
+      "step": 10910
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4325922080918189e-05,
+      "loss": 0.4652,
+      "step": 10911
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4325345329757476e-05,
+      "loss": 0.4142,
+      "step": 10912
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4324768578596766e-05,
+      "loss": 0.4888,
+      "step": 10913
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4324191827436054e-05,
+      "loss": 0.4745,
+      "step": 10914
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4323615076275341e-05,
+      "loss": 0.4774,
+      "step": 10915
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.432303832511463e-05,
+      "loss": 0.6226,
+      "step": 10916
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4322461573953918e-05,
+      "loss": 0.7,
+      "step": 10917
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4321884822793206e-05,
+      "loss": 0.5236,
+      "step": 10918
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4321308071632495e-05,
+      "loss": 0.531,
+      "step": 10919
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4320731320471783e-05,
+      "loss": 0.4301,
+      "step": 10920
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.432015456931107e-05,
+      "loss": 0.4824,
+      "step": 10921
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.431957781815036e-05,
+      "loss": 0.4909,
+      "step": 10922
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4319001066989648e-05,
+      "loss": 0.5802,
+      "step": 10923
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4318424315828935e-05,
+      "loss": 0.5447,
+      "step": 10924
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4317847564668227e-05,
+      "loss": 0.6127,
+      "step": 10925
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4317270813507513e-05,
+      "loss": 0.5266,
+      "step": 10926
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.43166940623468e-05,
+      "loss": 0.5023,
+      "step": 10927
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4316117311186091e-05,
+      "loss": 0.5773,
+      "step": 10928
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4315540560025379e-05,
+      "loss": 0.4935,
+      "step": 10929
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4314963808864667e-05,
+      "loss": 0.5002,
+      "step": 10930
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4314387057703956e-05,
+      "loss": 0.5406,
+      "step": 10931
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4313810306543244e-05,
+      "loss": 0.4376,
+      "step": 10932
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4313233555382531e-05,
+      "loss": 0.4446,
+      "step": 10933
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.431265680422182e-05,
+      "loss": 0.5101,
+      "step": 10934
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4312080053061108e-05,
+      "loss": 0.5059,
+      "step": 10935
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4311503301900396e-05,
+      "loss": 0.4858,
+      "step": 10936
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4310926550739686e-05,
+      "loss": 0.6058,
+      "step": 10937
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4310349799578973e-05,
+      "loss": 0.4222,
+      "step": 10938
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4309773048418261e-05,
+      "loss": 0.5263,
+      "step": 10939
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.430919629725755e-05,
+      "loss": 0.4454,
+      "step": 10940
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4308619546096838e-05,
+      "loss": 0.58,
+      "step": 10941
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4308042794936126e-05,
+      "loss": 0.483,
+      "step": 10942
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4307466043775415e-05,
+      "loss": 0.5237,
+      "step": 10943
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4306889292614703e-05,
+      "loss": 0.5327,
+      "step": 10944
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.430631254145399e-05,
+      "loss": 0.7252,
+      "step": 10945
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4305735790293278e-05,
+      "loss": 0.5494,
+      "step": 10946
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4305159039132567e-05,
+      "loss": 0.5449,
+      "step": 10947
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4304582287971855e-05,
+      "loss": 0.5389,
+      "step": 10948
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4304005536811143e-05,
+      "loss": 0.4881,
+      "step": 10949
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4303428785650432e-05,
+      "loss": 0.5449,
+      "step": 10950
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.430285203448972e-05,
+      "loss": 0.396,
+      "step": 10951
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4302275283329008e-05,
+      "loss": 0.4385,
+      "step": 10952
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4301698532168297e-05,
+      "loss": 0.4985,
+      "step": 10953
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4301121781007585e-05,
+      "loss": 0.5608,
+      "step": 10954
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4300545029846872e-05,
+      "loss": 0.5542,
+      "step": 10955
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4299968278686163e-05,
+      "loss": 0.6168,
+      "step": 10956
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.429939152752545e-05,
+      "loss": 0.4765,
+      "step": 10957
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4298814776364737e-05,
+      "loss": 0.4367,
+      "step": 10958
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4298238025204028e-05,
+      "loss": 0.4179,
+      "step": 10959
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4297661274043316e-05,
+      "loss": 0.4536,
+      "step": 10960
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4297084522882604e-05,
+      "loss": 0.3146,
+      "step": 10961
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4296507771721893e-05,
+      "loss": 0.3959,
+      "step": 10962
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.429593102056118e-05,
+      "loss": 0.5002,
+      "step": 10963
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4295354269400468e-05,
+      "loss": 0.5919,
+      "step": 10964
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4294777518239758e-05,
+      "loss": 0.4196,
+      "step": 10965
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4294200767079045e-05,
+      "loss": 0.4565,
+      "step": 10966
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4293624015918333e-05,
+      "loss": 0.5291,
+      "step": 10967
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4293047264757622e-05,
+      "loss": 0.6862,
+      "step": 10968
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.429247051359691e-05,
+      "loss": 0.5839,
+      "step": 10969
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4291893762436198e-05,
+      "loss": 0.4548,
+      "step": 10970
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4291317011275487e-05,
+      "loss": 0.4053,
+      "step": 10971
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4290740260114775e-05,
+      "loss": 0.5449,
+      "step": 10972
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4290163508954063e-05,
+      "loss": 0.5564,
+      "step": 10973
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4289586757793352e-05,
+      "loss": 0.4701,
+      "step": 10974
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.428901000663264e-05,
+      "loss": 0.5947,
+      "step": 10975
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4288433255471927e-05,
+      "loss": 0.4599,
+      "step": 10976
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4287856504311217e-05,
+      "loss": 0.523,
+      "step": 10977
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4287279753150504e-05,
+      "loss": 0.4449,
+      "step": 10978
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4286703001989792e-05,
+      "loss": 0.4907,
+      "step": 10979
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.428612625082908e-05,
+      "loss": 0.6761,
+      "step": 10980
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4285549499668369e-05,
+      "loss": 0.6,
+      "step": 10981
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4284972748507657e-05,
+      "loss": 0.5602,
+      "step": 10982
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4284395997346944e-05,
+      "loss": 0.5434,
+      "step": 10983
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4283819246186234e-05,
+      "loss": 0.5584,
+      "step": 10984
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4283242495025522e-05,
+      "loss": 0.6019,
+      "step": 10985
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.428266574386481e-05,
+      "loss": 0.4311,
+      "step": 10986
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.42820889927041e-05,
+      "loss": 0.4578,
+      "step": 10987
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4281512241543388e-05,
+      "loss": 0.455,
+      "step": 10988
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4280935490382674e-05,
+      "loss": 0.5448,
+      "step": 10989
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4280358739221965e-05,
+      "loss": 0.6506,
+      "step": 10990
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.4279781988061253e-05,
+      "loss": 0.528,
+      "step": 10991
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.427920523690054e-05,
+      "loss": 0.4837,
+      "step": 10992
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 1.427862848573983e-05,
+      "loss": 0.6462,
+      "step": 10993
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4278051734579117e-05,
+      "loss": 0.4868,
+      "step": 10994
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4277474983418405e-05,
+      "loss": 0.4898,
+      "step": 10995
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4276898232257695e-05,
+      "loss": 0.5748,
+      "step": 10996
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4276321481096982e-05,
+      "loss": 0.5689,
+      "step": 10997
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.427574472993627e-05,
+      "loss": 0.573,
+      "step": 10998
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.427516797877556e-05,
+      "loss": 0.53,
+      "step": 10999
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4274591227614847e-05,
+      "loss": 0.4945,
+      "step": 11000
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4274014476454135e-05,
+      "loss": 0.4678,
+      "step": 11001
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4273437725293424e-05,
+      "loss": 0.4946,
+      "step": 11002
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4272860974132712e-05,
+      "loss": 0.413,
+      "step": 11003
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4272284222972e-05,
+      "loss": 0.4415,
+      "step": 11004
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4271707471811289e-05,
+      "loss": 0.4746,
+      "step": 11005
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4271130720650576e-05,
+      "loss": 0.447,
+      "step": 11006
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4270553969489864e-05,
+      "loss": 0.4786,
+      "step": 11007
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4269977218329154e-05,
+      "loss": 0.498,
+      "step": 11008
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4269400467168441e-05,
+      "loss": 0.4526,
+      "step": 11009
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4268823716007729e-05,
+      "loss": 0.4627,
+      "step": 11010
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4268246964847018e-05,
+      "loss": 0.5681,
+      "step": 11011
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4267670213686306e-05,
+      "loss": 0.325,
+      "step": 11012
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4267093462525594e-05,
+      "loss": 0.4773,
+      "step": 11013
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4266516711364881e-05,
+      "loss": 0.4142,
+      "step": 11014
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.426593996020417e-05,
+      "loss": 0.6459,
+      "step": 11015
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4265363209043458e-05,
+      "loss": 0.5432,
+      "step": 11016
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4264786457882746e-05,
+      "loss": 0.5806,
+      "step": 11017
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4264209706722037e-05,
+      "loss": 0.5892,
+      "step": 11018
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4263632955561325e-05,
+      "loss": 0.531,
+      "step": 11019
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.426305620440061e-05,
+      "loss": 0.4627,
+      "step": 11020
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4262479453239902e-05,
+      "loss": 0.5349,
+      "step": 11021
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.426190270207919e-05,
+      "loss": 0.5228,
+      "step": 11022
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4261325950918477e-05,
+      "loss": 0.647,
+      "step": 11023
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4260749199757767e-05,
+      "loss": 0.4443,
+      "step": 11024
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4260172448597054e-05,
+      "loss": 0.5954,
+      "step": 11025
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4259595697436342e-05,
+      "loss": 0.4952,
+      "step": 11026
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4259018946275631e-05,
+      "loss": 0.5327,
+      "step": 11027
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4258442195114919e-05,
+      "loss": 0.5113,
+      "step": 11028
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4257865443954207e-05,
+      "loss": 0.5995,
+      "step": 11029
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4257288692793496e-05,
+      "loss": 0.5188,
+      "step": 11030
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4256711941632784e-05,
+      "loss": 0.4687,
+      "step": 11031
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4256135190472072e-05,
+      "loss": 0.5382,
+      "step": 11032
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4255558439311361e-05,
+      "loss": 0.3605,
+      "step": 11033
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4254981688150649e-05,
+      "loss": 0.5142,
+      "step": 11034
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4254404936989936e-05,
+      "loss": 0.6744,
+      "step": 11035
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4253828185829226e-05,
+      "loss": 0.5078,
+      "step": 11036
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4253251434668513e-05,
+      "loss": 0.5709,
+      "step": 11037
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4252674683507801e-05,
+      "loss": 0.6495,
+      "step": 11038
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.425209793234709e-05,
+      "loss": 0.6408,
+      "step": 11039
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4251521181186378e-05,
+      "loss": 0.4478,
+      "step": 11040
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4250944430025666e-05,
+      "loss": 0.5091,
+      "step": 11041
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4250367678864955e-05,
+      "loss": 0.6258,
+      "step": 11042
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4249790927704243e-05,
+      "loss": 0.4917,
+      "step": 11043
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.424921417654353e-05,
+      "loss": 0.4092,
+      "step": 11044
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4248637425382822e-05,
+      "loss": 0.4778,
+      "step": 11045
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4248060674222108e-05,
+      "loss": 0.3242,
+      "step": 11046
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4247483923061395e-05,
+      "loss": 0.5008,
+      "step": 11047
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4246907171900683e-05,
+      "loss": 0.6136,
+      "step": 11048
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4246330420739974e-05,
+      "loss": 0.475,
+      "step": 11049
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4245753669579262e-05,
+      "loss": 0.5811,
+      "step": 11050
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4245176918418548e-05,
+      "loss": 0.4217,
+      "step": 11051
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4244600167257839e-05,
+      "loss": 0.5364,
+      "step": 11052
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4244023416097126e-05,
+      "loss": 0.5904,
+      "step": 11053
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4243446664936414e-05,
+      "loss": 0.4962,
+      "step": 11054
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4242869913775704e-05,
+      "loss": 0.5587,
+      "step": 11055
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4242293162614991e-05,
+      "loss": 0.5093,
+      "step": 11056
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4241716411454279e-05,
+      "loss": 0.5753,
+      "step": 11057
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4241139660293568e-05,
+      "loss": 0.5653,
+      "step": 11058
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4240562909132856e-05,
+      "loss": 0.6895,
+      "step": 11059
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4239986157972144e-05,
+      "loss": 0.6482,
+      "step": 11060
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4239409406811433e-05,
+      "loss": 0.6429,
+      "step": 11061
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.423883265565072e-05,
+      "loss": 0.4538,
+      "step": 11062
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4238255904490008e-05,
+      "loss": 0.5995,
+      "step": 11063
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4237679153329298e-05,
+      "loss": 0.6424,
+      "step": 11064
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4237102402168585e-05,
+      "loss": 0.5665,
+      "step": 11065
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4236525651007873e-05,
+      "loss": 0.4518,
+      "step": 11066
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4235948899847163e-05,
+      "loss": 0.5831,
+      "step": 11067
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.423537214868645e-05,
+      "loss": 0.4888,
+      "step": 11068
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4234795397525738e-05,
+      "loss": 0.4867,
+      "step": 11069
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4234218646365027e-05,
+      "loss": 0.5042,
+      "step": 11070
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4233641895204315e-05,
+      "loss": 0.5275,
+      "step": 11071
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4233065144043603e-05,
+      "loss": 0.5674,
+      "step": 11072
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4232488392882892e-05,
+      "loss": 0.5874,
+      "step": 11073
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.423191164172218e-05,
+      "loss": 0.5407,
+      "step": 11074
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4231334890561467e-05,
+      "loss": 0.5446,
+      "step": 11075
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4230758139400758e-05,
+      "loss": 0.4701,
+      "step": 11076
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4230181388240044e-05,
+      "loss": 0.5294,
+      "step": 11077
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4229604637079332e-05,
+      "loss": 0.6189,
+      "step": 11078
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4229027885918623e-05,
+      "loss": 0.5479,
+      "step": 11079
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4228451134757911e-05,
+      "loss": 0.495,
+      "step": 11080
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4227874383597199e-05,
+      "loss": 0.5273,
+      "step": 11081
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4227297632436488e-05,
+      "loss": 0.4859,
+      "step": 11082
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4226720881275776e-05,
+      "loss": 0.4856,
+      "step": 11083
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4226144130115063e-05,
+      "loss": 0.5341,
+      "step": 11084
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4225567378954351e-05,
+      "loss": 0.5483,
+      "step": 11085
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.422499062779364e-05,
+      "loss": 0.5433,
+      "step": 11086
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4224413876632928e-05,
+      "loss": 0.5654,
+      "step": 11087
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4223837125472216e-05,
+      "loss": 0.5519,
+      "step": 11088
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4223260374311505e-05,
+      "loss": 0.5748,
+      "step": 11089
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4222683623150793e-05,
+      "loss": 0.5467,
+      "step": 11090
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.422210687199008e-05,
+      "loss": 0.5786,
+      "step": 11091
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.422153012082937e-05,
+      "loss": 0.5651,
+      "step": 11092
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4220953369668658e-05,
+      "loss": 0.5371,
+      "step": 11093
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4220376618507945e-05,
+      "loss": 0.4562,
+      "step": 11094
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4219799867347235e-05,
+      "loss": 0.5189,
+      "step": 11095
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4219223116186522e-05,
+      "loss": 0.6625,
+      "step": 11096
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.421864636502581e-05,
+      "loss": 0.5824,
+      "step": 11097
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.42180696138651e-05,
+      "loss": 0.5963,
+      "step": 11098
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4217492862704387e-05,
+      "loss": 0.4526,
+      "step": 11099
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4216916111543675e-05,
+      "loss": 0.5846,
+      "step": 11100
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4216339360382964e-05,
+      "loss": 0.5802,
+      "step": 11101
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4215762609222252e-05,
+      "loss": 0.4213,
+      "step": 11102
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.421518585806154e-05,
+      "loss": 0.5055,
+      "step": 11103
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4214609106900829e-05,
+      "loss": 0.5072,
+      "step": 11104
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4214032355740117e-05,
+      "loss": 0.5357,
+      "step": 11105
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4213455604579404e-05,
+      "loss": 0.659,
+      "step": 11106
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4212878853418695e-05,
+      "loss": 0.5327,
+      "step": 11107
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4212302102257981e-05,
+      "loss": 0.5257,
+      "step": 11108
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4211725351097269e-05,
+      "loss": 0.4643,
+      "step": 11109
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.421114859993656e-05,
+      "loss": 0.5035,
+      "step": 11110
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4210571848775848e-05,
+      "loss": 0.5586,
+      "step": 11111
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4209995097615135e-05,
+      "loss": 0.5058,
+      "step": 11112
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4209418346454425e-05,
+      "loss": 0.4108,
+      "step": 11113
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4208841595293713e-05,
+      "loss": 0.6183,
+      "step": 11114
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4208264844133e-05,
+      "loss": 0.5208,
+      "step": 11115
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.420768809297229e-05,
+      "loss": 0.6017,
+      "step": 11116
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4207111341811577e-05,
+      "loss": 0.5743,
+      "step": 11117
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4206534590650865e-05,
+      "loss": 0.4949,
+      "step": 11118
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4205957839490153e-05,
+      "loss": 0.4004,
+      "step": 11119
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4205381088329442e-05,
+      "loss": 0.4278,
+      "step": 11120
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.420480433716873e-05,
+      "loss": 0.4906,
+      "step": 11121
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4204227586008017e-05,
+      "loss": 0.5867,
+      "step": 11122
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4203650834847307e-05,
+      "loss": 0.5896,
+      "step": 11123
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4203074083686594e-05,
+      "loss": 0.5013,
+      "step": 11124
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4202497332525882e-05,
+      "loss": 0.5453,
+      "step": 11125
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4201920581365172e-05,
+      "loss": 0.6142,
+      "step": 11126
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.420134383020446e-05,
+      "loss": 0.6161,
+      "step": 11127
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4200767079043747e-05,
+      "loss": 0.5548,
+      "step": 11128
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4200190327883036e-05,
+      "loss": 0.5293,
+      "step": 11129
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4199613576722324e-05,
+      "loss": 0.6204,
+      "step": 11130
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4199036825561612e-05,
+      "loss": 0.5327,
+      "step": 11131
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4198460074400901e-05,
+      "loss": 0.5188,
+      "step": 11132
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4197883323240189e-05,
+      "loss": 0.417,
+      "step": 11133
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4197306572079476e-05,
+      "loss": 0.4615,
+      "step": 11134
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4196729820918766e-05,
+      "loss": 0.4486,
+      "step": 11135
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4196153069758053e-05,
+      "loss": 0.4796,
+      "step": 11136
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4195576318597341e-05,
+      "loss": 0.5126,
+      "step": 11137
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4194999567436632e-05,
+      "loss": 0.5976,
+      "step": 11138
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4194422816275918e-05,
+      "loss": 0.5016,
+      "step": 11139
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4193846065115206e-05,
+      "loss": 0.5243,
+      "step": 11140
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4193269313954497e-05,
+      "loss": 0.5534,
+      "step": 11141
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4192692562793785e-05,
+      "loss": 0.5181,
+      "step": 11142
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4192115811633072e-05,
+      "loss": 0.5023,
+      "step": 11143
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4191539060472362e-05,
+      "loss": 0.5524,
+      "step": 11144
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.419096230931165e-05,
+      "loss": 0.5442,
+      "step": 11145
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4190385558150937e-05,
+      "loss": 0.562,
+      "step": 11146
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4189808806990226e-05,
+      "loss": 0.5821,
+      "step": 11147
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4189232055829514e-05,
+      "loss": 0.58,
+      "step": 11148
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4188655304668802e-05,
+      "loss": 0.5333,
+      "step": 11149
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4188078553508091e-05,
+      "loss": 0.574,
+      "step": 11150
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4187501802347379e-05,
+      "loss": 0.4724,
+      "step": 11151
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4186925051186667e-05,
+      "loss": 0.5931,
+      "step": 11152
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4186348300025954e-05,
+      "loss": 0.4618,
+      "step": 11153
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4185771548865244e-05,
+      "loss": 0.5246,
+      "step": 11154
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4185194797704531e-05,
+      "loss": 0.5317,
+      "step": 11155
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4184618046543819e-05,
+      "loss": 0.4967,
+      "step": 11156
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4184041295383108e-05,
+      "loss": 0.5813,
+      "step": 11157
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4183464544222396e-05,
+      "loss": 0.4399,
+      "step": 11158
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4182887793061684e-05,
+      "loss": 0.541,
+      "step": 11159
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4182311041900973e-05,
+      "loss": 0.5094,
+      "step": 11160
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.418173429074026e-05,
+      "loss": 0.4132,
+      "step": 11161
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4181157539579549e-05,
+      "loss": 0.4526,
+      "step": 11162
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4180580788418838e-05,
+      "loss": 0.567,
+      "step": 11163
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4180004037258126e-05,
+      "loss": 0.6573,
+      "step": 11164
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4179427286097413e-05,
+      "loss": 0.4233,
+      "step": 11165
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4178850534936703e-05,
+      "loss": 0.5213,
+      "step": 11166
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.417827378377599e-05,
+      "loss": 0.6267,
+      "step": 11167
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4177697032615278e-05,
+      "loss": 0.4499,
+      "step": 11168
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4177120281454569e-05,
+      "loss": 0.5421,
+      "step": 11169
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4176543530293855e-05,
+      "loss": 0.3685,
+      "step": 11170
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 1.4175966779133143e-05,
+      "loss": 0.5088,
+      "step": 11171
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4175390027972434e-05,
+      "loss": 0.5449,
+      "step": 11172
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4174813276811722e-05,
+      "loss": 0.4685,
+      "step": 11173
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.417423652565101e-05,
+      "loss": 0.5511,
+      "step": 11174
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4173659774490299e-05,
+      "loss": 0.6755,
+      "step": 11175
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4173083023329586e-05,
+      "loss": 0.5787,
+      "step": 11176
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4172506272168874e-05,
+      "loss": 0.5232,
+      "step": 11177
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4171929521008163e-05,
+      "loss": 0.5486,
+      "step": 11178
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4171352769847451e-05,
+      "loss": 0.5696,
+      "step": 11179
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4170776018686739e-05,
+      "loss": 0.4999,
+      "step": 11180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4170199267526028e-05,
+      "loss": 0.662,
+      "step": 11181
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4169622516365316e-05,
+      "loss": 0.4597,
+      "step": 11182
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4169045765204603e-05,
+      "loss": 0.4363,
+      "step": 11183
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4168469014043893e-05,
+      "loss": 0.5082,
+      "step": 11184
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.416789226288318e-05,
+      "loss": 0.5292,
+      "step": 11185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4167315511722468e-05,
+      "loss": 0.5093,
+      "step": 11186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4166738760561758e-05,
+      "loss": 0.5292,
+      "step": 11187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4166162009401045e-05,
+      "loss": 0.4648,
+      "step": 11188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4165585258240333e-05,
+      "loss": 0.6067,
+      "step": 11189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.416500850707962e-05,
+      "loss": 0.701,
+      "step": 11190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.416443175591891e-05,
+      "loss": 0.5732,
+      "step": 11191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4163855004758198e-05,
+      "loss": 0.5337,
+      "step": 11192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4163278253597485e-05,
+      "loss": 0.5033,
+      "step": 11193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4162701502436775e-05,
+      "loss": 0.4303,
+      "step": 11194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4162124751276062e-05,
+      "loss": 0.5184,
+      "step": 11195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.416154800011535e-05,
+      "loss": 0.6216,
+      "step": 11196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.416097124895464e-05,
+      "loss": 0.4805,
+      "step": 11197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4160394497793927e-05,
+      "loss": 0.5719,
+      "step": 11198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4159817746633215e-05,
+      "loss": 0.5455,
+      "step": 11199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4159240995472506e-05,
+      "loss": 0.4412,
+      "step": 11200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4158664244311792e-05,
+      "loss": 0.3962,
+      "step": 11201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.415808749315108e-05,
+      "loss": 0.4778,
+      "step": 11202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.415751074199037e-05,
+      "loss": 0.5257,
+      "step": 11203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4156933990829658e-05,
+      "loss": 0.5132,
+      "step": 11204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4156357239668946e-05,
+      "loss": 0.4561,
+      "step": 11205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4155780488508235e-05,
+      "loss": 0.3589,
+      "step": 11206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4155203737347523e-05,
+      "loss": 0.551,
+      "step": 11207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.415462698618681e-05,
+      "loss": 0.5366,
+      "step": 11208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.41540502350261e-05,
+      "loss": 0.4413,
+      "step": 11209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4153473483865388e-05,
+      "loss": 0.6161,
+      "step": 11210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4152896732704676e-05,
+      "loss": 0.6191,
+      "step": 11211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4152319981543965e-05,
+      "loss": 0.5292,
+      "step": 11212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4151743230383253e-05,
+      "loss": 0.4008,
+      "step": 11213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.415116647922254e-05,
+      "loss": 0.5709,
+      "step": 11214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.415058972806183e-05,
+      "loss": 0.5848,
+      "step": 11215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4150012976901117e-05,
+      "loss": 0.5098,
+      "step": 11216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4149436225740405e-05,
+      "loss": 0.6483,
+      "step": 11217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4148859474579694e-05,
+      "loss": 0.613,
+      "step": 11218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4148282723418982e-05,
+      "loss": 0.5122,
+      "step": 11219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.414770597225827e-05,
+      "loss": 0.4592,
+      "step": 11220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.414712922109756e-05,
+      "loss": 0.5334,
+      "step": 11221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4146552469936847e-05,
+      "loss": 0.4295,
+      "step": 11222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4145975718776135e-05,
+      "loss": 0.5675,
+      "step": 11223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4145398967615422e-05,
+      "loss": 0.4546,
+      "step": 11224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4144822216454712e-05,
+      "loss": 0.613,
+      "step": 11225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4144245465294e-05,
+      "loss": 0.5378,
+      "step": 11226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4143668714133287e-05,
+      "loss": 0.4564,
+      "step": 11227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4143091962972576e-05,
+      "loss": 0.5275,
+      "step": 11228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4142515211811864e-05,
+      "loss": 0.4567,
+      "step": 11229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4141938460651152e-05,
+      "loss": 0.6259,
+      "step": 11230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4141361709490443e-05,
+      "loss": 0.5526,
+      "step": 11231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4140784958329729e-05,
+      "loss": 0.5252,
+      "step": 11232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4140208207169016e-05,
+      "loss": 0.4552,
+      "step": 11233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4139631456008308e-05,
+      "loss": 0.5775,
+      "step": 11234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4139054704847595e-05,
+      "loss": 0.4792,
+      "step": 11235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4138477953686883e-05,
+      "loss": 0.5923,
+      "step": 11236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4137901202526172e-05,
+      "loss": 0.4578,
+      "step": 11237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.413732445136546e-05,
+      "loss": 0.5332,
+      "step": 11238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4136747700204748e-05,
+      "loss": 0.4644,
+      "step": 11239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4136170949044037e-05,
+      "loss": 0.5881,
+      "step": 11240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4135594197883325e-05,
+      "loss": 0.5937,
+      "step": 11241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4135017446722612e-05,
+      "loss": 0.6414,
+      "step": 11242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4134440695561902e-05,
+      "loss": 0.5428,
+      "step": 11243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.413386394440119e-05,
+      "loss": 0.5456,
+      "step": 11244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4133287193240477e-05,
+      "loss": 0.5433,
+      "step": 11245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4132710442079767e-05,
+      "loss": 0.5005,
+      "step": 11246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4132133690919054e-05,
+      "loss": 0.4919,
+      "step": 11247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4131556939758342e-05,
+      "loss": 0.6419,
+      "step": 11248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4130980188597631e-05,
+      "loss": 0.5098,
+      "step": 11249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4130403437436919e-05,
+      "loss": 0.4444,
+      "step": 11250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4129826686276207e-05,
+      "loss": 0.4236,
+      "step": 11251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4129249935115496e-05,
+      "loss": 0.5473,
+      "step": 11252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4128673183954784e-05,
+      "loss": 0.5298,
+      "step": 11253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4128096432794071e-05,
+      "loss": 0.5536,
+      "step": 11254
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.412751968163336e-05,
+      "loss": 0.3981,
+      "step": 11255
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4126942930472649e-05,
+      "loss": 0.6375,
+      "step": 11256
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4126366179311936e-05,
+      "loss": 0.5798,
+      "step": 11257
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4125789428151224e-05,
+      "loss": 0.5256,
+      "step": 11258
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4125212676990513e-05,
+      "loss": 0.4581,
+      "step": 11259
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4124635925829801e-05,
+      "loss": 0.3604,
+      "step": 11260
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4124059174669089e-05,
+      "loss": 0.4501,
+      "step": 11261
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.412348242350838e-05,
+      "loss": 0.4765,
+      "step": 11262
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4122905672347666e-05,
+      "loss": 0.4343,
+      "step": 11263
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4122328921186953e-05,
+      "loss": 0.5583,
+      "step": 11264
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4121752170026244e-05,
+      "loss": 0.5371,
+      "step": 11265
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4121175418865532e-05,
+      "loss": 0.5385,
+      "step": 11266
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.412059866770482e-05,
+      "loss": 0.6,
+      "step": 11267
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.412002191654411e-05,
+      "loss": 0.6353,
+      "step": 11268
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4119445165383397e-05,
+      "loss": 0.5286,
+      "step": 11269
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4118868414222685e-05,
+      "loss": 0.5414,
+      "step": 11270
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4118291663061974e-05,
+      "loss": 0.6341,
+      "step": 11271
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4117714911901262e-05,
+      "loss": 0.6091,
+      "step": 11272
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.411713816074055e-05,
+      "loss": 0.4524,
+      "step": 11273
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4116561409579839e-05,
+      "loss": 0.7308,
+      "step": 11274
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4115984658419126e-05,
+      "loss": 0.5672,
+      "step": 11275
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4115407907258414e-05,
+      "loss": 0.5673,
+      "step": 11276
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4114831156097703e-05,
+      "loss": 0.4325,
+      "step": 11277
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4114254404936991e-05,
+      "loss": 0.623,
+      "step": 11278
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4113677653776279e-05,
+      "loss": 0.5811,
+      "step": 11279
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4113100902615568e-05,
+      "loss": 0.6066,
+      "step": 11280
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4112524151454856e-05,
+      "loss": 0.4824,
+      "step": 11281
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4111947400294144e-05,
+      "loss": 0.4129,
+      "step": 11282
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4111370649133433e-05,
+      "loss": 0.5355,
+      "step": 11283
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.411079389797272e-05,
+      "loss": 0.5252,
+      "step": 11284
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4110217146812008e-05,
+      "loss": 0.443,
+      "step": 11285
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4109640395651298e-05,
+      "loss": 0.4919,
+      "step": 11286
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4109063644490585e-05,
+      "loss": 0.3715,
+      "step": 11287
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4108486893329873e-05,
+      "loss": 0.4727,
+      "step": 11288
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4107910142169164e-05,
+      "loss": 0.4551,
+      "step": 11289
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.410733339100845e-05,
+      "loss": 0.657,
+      "step": 11290
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4106756639847738e-05,
+      "loss": 0.537,
+      "step": 11291
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4106179888687025e-05,
+      "loss": 0.4244,
+      "step": 11292
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4105603137526317e-05,
+      "loss": 0.506,
+      "step": 11293
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4105026386365603e-05,
+      "loss": 0.3808,
+      "step": 11294
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.410444963520489e-05,
+      "loss": 0.5488,
+      "step": 11295
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4103872884044181e-05,
+      "loss": 0.5211,
+      "step": 11296
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4103296132883469e-05,
+      "loss": 0.5501,
+      "step": 11297
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4102719381722757e-05,
+      "loss": 0.5631,
+      "step": 11298
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4102142630562046e-05,
+      "loss": 0.4666,
+      "step": 11299
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4101565879401334e-05,
+      "loss": 0.5264,
+      "step": 11300
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4100989128240621e-05,
+      "loss": 0.4656,
+      "step": 11301
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.410041237707991e-05,
+      "loss": 0.4758,
+      "step": 11302
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4099835625919198e-05,
+      "loss": 0.4198,
+      "step": 11303
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4099258874758486e-05,
+      "loss": 0.4995,
+      "step": 11304
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4098682123597776e-05,
+      "loss": 0.4251,
+      "step": 11305
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4098105372437063e-05,
+      "loss": 0.5631,
+      "step": 11306
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4097528621276351e-05,
+      "loss": 0.5923,
+      "step": 11307
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.409695187011564e-05,
+      "loss": 0.587,
+      "step": 11308
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4096375118954928e-05,
+      "loss": 0.4225,
+      "step": 11309
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4095798367794216e-05,
+      "loss": 0.4096,
+      "step": 11310
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4095221616633505e-05,
+      "loss": 0.5088,
+      "step": 11311
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4094644865472793e-05,
+      "loss": 0.5907,
+      "step": 11312
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.409406811431208e-05,
+      "loss": 0.4322,
+      "step": 11313
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.409349136315137e-05,
+      "loss": 0.5882,
+      "step": 11314
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4092914611990657e-05,
+      "loss": 0.5713,
+      "step": 11315
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4092337860829945e-05,
+      "loss": 0.5274,
+      "step": 11316
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4091761109669235e-05,
+      "loss": 0.4715,
+      "step": 11317
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4091184358508522e-05,
+      "loss": 0.5619,
+      "step": 11318
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.409060760734781e-05,
+      "loss": 0.398,
+      "step": 11319
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4090030856187101e-05,
+      "loss": 0.4879,
+      "step": 11320
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4089454105026387e-05,
+      "loss": 0.5501,
+      "step": 11321
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4088877353865675e-05,
+      "loss": 0.5184,
+      "step": 11322
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4088300602704966e-05,
+      "loss": 0.6495,
+      "step": 11323
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4087723851544253e-05,
+      "loss": 0.4858,
+      "step": 11324
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.408714710038354e-05,
+      "loss": 0.3458,
+      "step": 11325
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.408657034922283e-05,
+      "loss": 0.5761,
+      "step": 11326
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4085993598062118e-05,
+      "loss": 0.5808,
+      "step": 11327
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4085416846901406e-05,
+      "loss": 0.5021,
+      "step": 11328
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4084840095740694e-05,
+      "loss": 0.6367,
+      "step": 11329
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4084263344579983e-05,
+      "loss": 0.6684,
+      "step": 11330
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.408368659341927e-05,
+      "loss": 0.7196,
+      "step": 11331
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4083109842258558e-05,
+      "loss": 0.5385,
+      "step": 11332
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4082533091097848e-05,
+      "loss": 0.4318,
+      "step": 11333
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4081956339937135e-05,
+      "loss": 0.4716,
+      "step": 11334
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4081379588776423e-05,
+      "loss": 0.5192,
+      "step": 11335
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4080802837615712e-05,
+      "loss": 0.5708,
+      "step": 11336
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4080226086455e-05,
+      "loss": 0.5037,
+      "step": 11337
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4079649335294288e-05,
+      "loss": 0.5274,
+      "step": 11338
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4079072584133577e-05,
+      "loss": 0.3849,
+      "step": 11339
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4078495832972865e-05,
+      "loss": 0.5311,
+      "step": 11340
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4077919081812153e-05,
+      "loss": 0.5809,
+      "step": 11341
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4077342330651442e-05,
+      "loss": 0.5405,
+      "step": 11342
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.407676557949073e-05,
+      "loss": 0.5652,
+      "step": 11343
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4076188828330017e-05,
+      "loss": 0.4972,
+      "step": 11344
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4075612077169307e-05,
+      "loss": 0.6767,
+      "step": 11345
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4075035326008594e-05,
+      "loss": 0.3764,
+      "step": 11346
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4074458574847882e-05,
+      "loss": 0.5913,
+      "step": 11347
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4073881823687171e-05,
+      "loss": 0.4593,
+      "step": 11348
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4073305072526459e-05,
+      "loss": 0.4995,
+      "step": 11349
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 1.4072728321365747e-05,
+      "loss": 0.3488,
+      "step": 11350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4072151570205038e-05,
+      "loss": 0.4584,
+      "step": 11351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4071574819044324e-05,
+      "loss": 0.4586,
+      "step": 11352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4070998067883612e-05,
+      "loss": 0.5516,
+      "step": 11353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4070421316722903e-05,
+      "loss": 0.3809,
+      "step": 11354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.406984456556219e-05,
+      "loss": 0.4781,
+      "step": 11355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4069267814401478e-05,
+      "loss": 0.5197,
+      "step": 11356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4068691063240767e-05,
+      "loss": 0.5733,
+      "step": 11357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4068114312080055e-05,
+      "loss": 0.4643,
+      "step": 11358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4067537560919343e-05,
+      "loss": 0.6098,
+      "step": 11359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4066960809758632e-05,
+      "loss": 0.4728,
+      "step": 11360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.406638405859792e-05,
+      "loss": 0.5036,
+      "step": 11361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4065807307437207e-05,
+      "loss": 0.5658,
+      "step": 11362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4065230556276495e-05,
+      "loss": 0.4698,
+      "step": 11363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4064653805115785e-05,
+      "loss": 0.6617,
+      "step": 11364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4064077053955072e-05,
+      "loss": 0.428,
+      "step": 11365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.406350030279436e-05,
+      "loss": 0.443,
+      "step": 11366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.406292355163365e-05,
+      "loss": 0.5117,
+      "step": 11367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4062346800472937e-05,
+      "loss": 0.5628,
+      "step": 11368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4061770049312225e-05,
+      "loss": 0.5899,
+      "step": 11369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4061193298151514e-05,
+      "loss": 0.5627,
+      "step": 11370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4060616546990802e-05,
+      "loss": 0.4274,
+      "step": 11371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.406003979583009e-05,
+      "loss": 0.6018,
+      "step": 11372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4059463044669379e-05,
+      "loss": 0.5359,
+      "step": 11373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4058886293508666e-05,
+      "loss": 0.3023,
+      "step": 11374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4058309542347954e-05,
+      "loss": 0.4294,
+      "step": 11375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4057732791187244e-05,
+      "loss": 0.493,
+      "step": 11376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4057156040026531e-05,
+      "loss": 0.4063,
+      "step": 11377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4056579288865819e-05,
+      "loss": 0.5273,
+      "step": 11378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4056002537705108e-05,
+      "loss": 0.4143,
+      "step": 11379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4055425786544396e-05,
+      "loss": 0.4932,
+      "step": 11380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4054849035383684e-05,
+      "loss": 0.4974,
+      "step": 11381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4054272284222975e-05,
+      "loss": 0.5352,
+      "step": 11382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.405369553306226e-05,
+      "loss": 0.4977,
+      "step": 11383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4053118781901548e-05,
+      "loss": 0.5523,
+      "step": 11384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.405254203074084e-05,
+      "loss": 0.548,
+      "step": 11385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4051965279580127e-05,
+      "loss": 0.4987,
+      "step": 11386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4051388528419415e-05,
+      "loss": 0.4906,
+      "step": 11387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4050811777258704e-05,
+      "loss": 0.3498,
+      "step": 11388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4050235026097992e-05,
+      "loss": 0.5306,
+      "step": 11389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.404965827493728e-05,
+      "loss": 0.5942,
+      "step": 11390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4049081523776569e-05,
+      "loss": 0.47,
+      "step": 11391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4048504772615857e-05,
+      "loss": 0.4708,
+      "step": 11392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4047928021455144e-05,
+      "loss": 0.5314,
+      "step": 11393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4047351270294434e-05,
+      "loss": 0.4425,
+      "step": 11394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4046774519133721e-05,
+      "loss": 0.4012,
+      "step": 11395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4046197767973009e-05,
+      "loss": 0.4844,
+      "step": 11396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4045621016812297e-05,
+      "loss": 0.4896,
+      "step": 11397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4045044265651586e-05,
+      "loss": 0.4633,
+      "step": 11398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4044467514490874e-05,
+      "loss": 0.4144,
+      "step": 11399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4043890763330162e-05,
+      "loss": 0.5031,
+      "step": 11400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4043314012169451e-05,
+      "loss": 0.6611,
+      "step": 11401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4042737261008739e-05,
+      "loss": 0.5241,
+      "step": 11402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4042160509848026e-05,
+      "loss": 0.5543,
+      "step": 11403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4041583758687316e-05,
+      "loss": 0.5838,
+      "step": 11404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4041007007526603e-05,
+      "loss": 0.5369,
+      "step": 11405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4040430256365891e-05,
+      "loss": 0.5168,
+      "step": 11406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.403985350520518e-05,
+      "loss": 0.4447,
+      "step": 11407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4039276754044468e-05,
+      "loss": 0.4279,
+      "step": 11408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4038700002883756e-05,
+      "loss": 0.5636,
+      "step": 11409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4038123251723045e-05,
+      "loss": 0.6764,
+      "step": 11410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4037546500562333e-05,
+      "loss": 0.4983,
+      "step": 11411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.403696974940162e-05,
+      "loss": 0.5929,
+      "step": 11412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4036392998240912e-05,
+      "loss": 0.4361,
+      "step": 11413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4035816247080198e-05,
+      "loss": 0.5496,
+      "step": 11414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4035239495919485e-05,
+      "loss": 0.5638,
+      "step": 11415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4034662744758776e-05,
+      "loss": 0.5338,
+      "step": 11416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4034085993598064e-05,
+      "loss": 0.5138,
+      "step": 11417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4033509242437352e-05,
+      "loss": 0.5477,
+      "step": 11418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4032932491276641e-05,
+      "loss": 0.5557,
+      "step": 11419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4032355740115929e-05,
+      "loss": 0.5628,
+      "step": 11420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4031778988955216e-05,
+      "loss": 0.5829,
+      "step": 11421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4031202237794506e-05,
+      "loss": 0.5824,
+      "step": 11422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4030625486633794e-05,
+      "loss": 0.5091,
+      "step": 11423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4030048735473081e-05,
+      "loss": 0.4604,
+      "step": 11424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.402947198431237e-05,
+      "loss": 0.4941,
+      "step": 11425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4028895233151658e-05,
+      "loss": 0.5408,
+      "step": 11426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4028318481990946e-05,
+      "loss": 0.5958,
+      "step": 11427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4027741730830235e-05,
+      "loss": 0.5662,
+      "step": 11428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4027164979669523e-05,
+      "loss": 0.5853,
+      "step": 11429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.402658822850881e-05,
+      "loss": 0.4582,
+      "step": 11430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4026011477348098e-05,
+      "loss": 0.5382,
+      "step": 11431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4025434726187388e-05,
+      "loss": 0.5431,
+      "step": 11432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4024857975026675e-05,
+      "loss": 0.682,
+      "step": 11433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4024281223865963e-05,
+      "loss": 0.4645,
+      "step": 11434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4023704472705253e-05,
+      "loss": 0.5935,
+      "step": 11435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.402312772154454e-05,
+      "loss": 0.4622,
+      "step": 11436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4022550970383828e-05,
+      "loss": 0.5941,
+      "step": 11437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4021974219223117e-05,
+      "loss": 0.4322,
+      "step": 11438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4021397468062405e-05,
+      "loss": 0.4879,
+      "step": 11439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4020820716901693e-05,
+      "loss": 0.4827,
+      "step": 11440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4020243965740982e-05,
+      "loss": 0.4623,
+      "step": 11441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.401966721458027e-05,
+      "loss": 0.447,
+      "step": 11442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4019090463419557e-05,
+      "loss": 0.5512,
+      "step": 11443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4018513712258848e-05,
+      "loss": 0.531,
+      "step": 11444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4017936961098134e-05,
+      "loss": 0.5976,
+      "step": 11445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4017360209937422e-05,
+      "loss": 0.3773,
+      "step": 11446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4016783458776713e-05,
+      "loss": 0.5286,
+      "step": 11447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4016206707616001e-05,
+      "loss": 0.4584,
+      "step": 11448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4015629956455289e-05,
+      "loss": 0.5447,
+      "step": 11449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4015053205294578e-05,
+      "loss": 0.5487,
+      "step": 11450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4014476454133866e-05,
+      "loss": 0.4414,
+      "step": 11451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4013899702973153e-05,
+      "loss": 0.5924,
+      "step": 11452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4013322951812443e-05,
+      "loss": 0.4973,
+      "step": 11453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.401274620065173e-05,
+      "loss": 0.4421,
+      "step": 11454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4012169449491018e-05,
+      "loss": 0.4782,
+      "step": 11455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4011592698330307e-05,
+      "loss": 0.4224,
+      "step": 11456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4011015947169595e-05,
+      "loss": 0.4825,
+      "step": 11457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4010439196008883e-05,
+      "loss": 0.4941,
+      "step": 11458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4009862444848172e-05,
+      "loss": 0.6099,
+      "step": 11459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.400928569368746e-05,
+      "loss": 0.4892,
+      "step": 11460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4008708942526748e-05,
+      "loss": 0.4486,
+      "step": 11461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4008132191366037e-05,
+      "loss": 0.3389,
+      "step": 11462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4007555440205325e-05,
+      "loss": 0.5198,
+      "step": 11463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4006978689044612e-05,
+      "loss": 0.4731,
+      "step": 11464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4006401937883902e-05,
+      "loss": 0.4945,
+      "step": 11465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.400582518672319e-05,
+      "loss": 0.5515,
+      "step": 11466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4005248435562477e-05,
+      "loss": 0.5399,
+      "step": 11467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4004671684401765e-05,
+      "loss": 0.5201,
+      "step": 11468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4004094933241054e-05,
+      "loss": 0.5235,
+      "step": 11469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4003518182080342e-05,
+      "loss": 0.5761,
+      "step": 11470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.400294143091963e-05,
+      "loss": 0.5285,
+      "step": 11471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4002364679758919e-05,
+      "loss": 0.4322,
+      "step": 11472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4001787928598207e-05,
+      "loss": 0.4696,
+      "step": 11473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4001211177437494e-05,
+      "loss": 0.5348,
+      "step": 11474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4000634426276785e-05,
+      "loss": 0.5718,
+      "step": 11475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.4000057675116071e-05,
+      "loss": 0.3834,
+      "step": 11476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3999480923955359e-05,
+      "loss": 0.5796,
+      "step": 11477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.399890417279465e-05,
+      "loss": 0.5345,
+      "step": 11478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3998327421633938e-05,
+      "loss": 0.5016,
+      "step": 11479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3997750670473225e-05,
+      "loss": 0.4989,
+      "step": 11480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3997173919312515e-05,
+      "loss": 0.4695,
+      "step": 11481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3996597168151803e-05,
+      "loss": 0.5035,
+      "step": 11482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.399602041699109e-05,
+      "loss": 0.3962,
+      "step": 11483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.399544366583038e-05,
+      "loss": 0.4168,
+      "step": 11484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3994866914669667e-05,
+      "loss": 0.4592,
+      "step": 11485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3994290163508955e-05,
+      "loss": 0.3967,
+      "step": 11486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3993713412348244e-05,
+      "loss": 0.5656,
+      "step": 11487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3993136661187532e-05,
+      "loss": 0.5637,
+      "step": 11488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.399255991002682e-05,
+      "loss": 0.4825,
+      "step": 11489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3991983158866109e-05,
+      "loss": 0.559,
+      "step": 11490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3991406407705397e-05,
+      "loss": 0.4152,
+      "step": 11491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3990829656544684e-05,
+      "loss": 0.5309,
+      "step": 11492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3990252905383974e-05,
+      "loss": 0.5544,
+      "step": 11493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3989676154223262e-05,
+      "loss": 0.3451,
+      "step": 11494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.398909940306255e-05,
+      "loss": 0.5325,
+      "step": 11495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3988522651901839e-05,
+      "loss": 0.5629,
+      "step": 11496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3987945900741126e-05,
+      "loss": 0.4103,
+      "step": 11497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3987369149580414e-05,
+      "loss": 0.5167,
+      "step": 11498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3986792398419703e-05,
+      "loss": 0.4891,
+      "step": 11499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3986215647258991e-05,
+      "loss": 0.4141,
+      "step": 11500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3985638896098279e-05,
+      "loss": 0.5904,
+      "step": 11501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3985062144937566e-05,
+      "loss": 0.3727,
+      "step": 11502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3984485393776856e-05,
+      "loss": 0.598,
+      "step": 11503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3983908642616143e-05,
+      "loss": 0.4547,
+      "step": 11504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3983331891455431e-05,
+      "loss": 0.5306,
+      "step": 11505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3982755140294722e-05,
+      "loss": 0.5762,
+      "step": 11506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3982178389134008e-05,
+      "loss": 0.4781,
+      "step": 11507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3981601637973296e-05,
+      "loss": 0.4942,
+      "step": 11508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3981024886812587e-05,
+      "loss": 0.5577,
+      "step": 11509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3980448135651875e-05,
+      "loss": 0.5805,
+      "step": 11510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3979871384491162e-05,
+      "loss": 0.4841,
+      "step": 11511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3979294633330452e-05,
+      "loss": 0.4686,
+      "step": 11512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.397871788216974e-05,
+      "loss": 0.5274,
+      "step": 11513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3978141131009027e-05,
+      "loss": 0.4566,
+      "step": 11514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3977564379848316e-05,
+      "loss": 0.5676,
+      "step": 11515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3976987628687604e-05,
+      "loss": 0.6515,
+      "step": 11516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3976410877526892e-05,
+      "loss": 0.5109,
+      "step": 11517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3975834126366181e-05,
+      "loss": 0.451,
+      "step": 11518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3975257375205469e-05,
+      "loss": 0.6188,
+      "step": 11519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3974680624044757e-05,
+      "loss": 0.4289,
+      "step": 11520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3974103872884046e-05,
+      "loss": 0.439,
+      "step": 11521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3973527121723334e-05,
+      "loss": 0.6213,
+      "step": 11522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3972950370562621e-05,
+      "loss": 0.3994,
+      "step": 11523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.397237361940191e-05,
+      "loss": 0.4562,
+      "step": 11524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3971796868241198e-05,
+      "loss": 0.4778,
+      "step": 11525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3971220117080486e-05,
+      "loss": 0.5875,
+      "step": 11526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3970643365919775e-05,
+      "loss": 0.5344,
+      "step": 11527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3970066614759063e-05,
+      "loss": 0.4277,
+      "step": 11528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 1.3969489863598351e-05,
+      "loss": 0.4137,
+      "step": 11529
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.396891311243764e-05,
+      "loss": 0.6304,
+      "step": 11530
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3968336361276928e-05,
+      "loss": 0.6602,
+      "step": 11531
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3967759610116216e-05,
+      "loss": 0.5293,
+      "step": 11532
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3967182858955507e-05,
+      "loss": 0.4364,
+      "step": 11533
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3966606107794793e-05,
+      "loss": 0.647,
+      "step": 11534
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.396602935663408e-05,
+      "loss": 0.5365,
+      "step": 11535
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3965452605473368e-05,
+      "loss": 0.4599,
+      "step": 11536
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3964875854312659e-05,
+      "loss": 0.4851,
+      "step": 11537
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3964299103151945e-05,
+      "loss": 0.6325,
+      "step": 11538
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3963722351991233e-05,
+      "loss": 0.4955,
+      "step": 11539
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3963145600830524e-05,
+      "loss": 0.5381,
+      "step": 11540
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3962568849669812e-05,
+      "loss": 0.5012,
+      "step": 11541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.39619920985091e-05,
+      "loss": 0.4625,
+      "step": 11542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3961415347348389e-05,
+      "loss": 0.471,
+      "step": 11543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3960838596187676e-05,
+      "loss": 0.3894,
+      "step": 11544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3960261845026964e-05,
+      "loss": 0.5503,
+      "step": 11545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3959685093866253e-05,
+      "loss": 0.4784,
+      "step": 11546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3959108342705541e-05,
+      "loss": 0.5161,
+      "step": 11547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3958531591544829e-05,
+      "loss": 0.5607,
+      "step": 11548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3957954840384118e-05,
+      "loss": 0.5882,
+      "step": 11549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3957378089223406e-05,
+      "loss": 0.5553,
+      "step": 11550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3956801338062693e-05,
+      "loss": 0.4094,
+      "step": 11551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3956224586901983e-05,
+      "loss": 0.4521,
+      "step": 11552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.395564783574127e-05,
+      "loss": 0.5782,
+      "step": 11553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3955071084580558e-05,
+      "loss": 0.5315,
+      "step": 11554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3954494333419848e-05,
+      "loss": 0.4364,
+      "step": 11555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3953917582259135e-05,
+      "loss": 0.531,
+      "step": 11556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3953340831098423e-05,
+      "loss": 0.3733,
+      "step": 11557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3952764079937712e-05,
+      "loss": 0.4806,
+      "step": 11558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3952187328777e-05,
+      "loss": 0.4964,
+      "step": 11559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3951610577616288e-05,
+      "loss": 0.5508,
+      "step": 11560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3951033826455577e-05,
+      "loss": 0.417,
+      "step": 11561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3950457075294865e-05,
+      "loss": 0.5124,
+      "step": 11562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3949880324134152e-05,
+      "loss": 0.5369,
+      "step": 11563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3949303572973444e-05,
+      "loss": 0.4954,
+      "step": 11564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.394872682181273e-05,
+      "loss": 0.5803,
+      "step": 11565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3948150070652017e-05,
+      "loss": 0.524,
+      "step": 11566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3947573319491308e-05,
+      "loss": 0.5182,
+      "step": 11567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3946996568330596e-05,
+      "loss": 0.3903,
+      "step": 11568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3946419817169882e-05,
+      "loss": 0.5506,
+      "step": 11569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3945843066009173e-05,
+      "loss": 0.6769,
+      "step": 11570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.394526631484846e-05,
+      "loss": 0.4157,
+      "step": 11571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3944689563687748e-05,
+      "loss": 0.5101,
+      "step": 11572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3944112812527036e-05,
+      "loss": 0.49,
+      "step": 11573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3943536061366325e-05,
+      "loss": 0.429,
+      "step": 11574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3942959310205613e-05,
+      "loss": 0.3638,
+      "step": 11575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3942382559044901e-05,
+      "loss": 0.4805,
+      "step": 11576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.394180580788419e-05,
+      "loss": 0.5035,
+      "step": 11577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3941229056723478e-05,
+      "loss": 0.5888,
+      "step": 11578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3940652305562766e-05,
+      "loss": 0.518,
+      "step": 11579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3940075554402055e-05,
+      "loss": 0.5851,
+      "step": 11580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3939498803241343e-05,
+      "loss": 0.6278,
+      "step": 11581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393892205208063e-05,
+      "loss": 0.362,
+      "step": 11582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393834530091992e-05,
+      "loss": 0.5025,
+      "step": 11583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3937768549759207e-05,
+      "loss": 0.3301,
+      "step": 11584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3937191798598495e-05,
+      "loss": 0.526,
+      "step": 11585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3936615047437784e-05,
+      "loss": 0.4887,
+      "step": 11586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3936038296277072e-05,
+      "loss": 0.5328,
+      "step": 11587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393546154511636e-05,
+      "loss": 0.4739,
+      "step": 11588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393488479395565e-05,
+      "loss": 0.539,
+      "step": 11589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3934308042794937e-05,
+      "loss": 0.644,
+      "step": 11590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3933731291634225e-05,
+      "loss": 0.5267,
+      "step": 11591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3933154540473514e-05,
+      "loss": 0.4718,
+      "step": 11592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3932577789312802e-05,
+      "loss": 0.6788,
+      "step": 11593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393200103815209e-05,
+      "loss": 0.5745,
+      "step": 11594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.393142428699138e-05,
+      "loss": 0.5946,
+      "step": 11595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3930847535830666e-05,
+      "loss": 0.5419,
+      "step": 11596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3930270784669954e-05,
+      "loss": 0.475,
+      "step": 11597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3929694033509245e-05,
+      "loss": 0.539,
+      "step": 11598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3929117282348533e-05,
+      "loss": 0.595,
+      "step": 11599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3928540531187819e-05,
+      "loss": 0.5035,
+      "step": 11600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.392796378002711e-05,
+      "loss": 0.4842,
+      "step": 11601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3927387028866398e-05,
+      "loss": 0.5034,
+      "step": 11602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3926810277705685e-05,
+      "loss": 0.4062,
+      "step": 11603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3926233526544975e-05,
+      "loss": 0.4653,
+      "step": 11604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3925656775384262e-05,
+      "loss": 0.53,
+      "step": 11605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.392508002422355e-05,
+      "loss": 0.6002,
+      "step": 11606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3924503273062838e-05,
+      "loss": 0.4972,
+      "step": 11607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3923926521902127e-05,
+      "loss": 0.5007,
+      "step": 11608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3923349770741415e-05,
+      "loss": 0.3975,
+      "step": 11609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3922773019580702e-05,
+      "loss": 0.503,
+      "step": 11610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3922196268419992e-05,
+      "loss": 0.6214,
+      "step": 11611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.392161951725928e-05,
+      "loss": 0.4709,
+      "step": 11612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3921042766098567e-05,
+      "loss": 0.5805,
+      "step": 11613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3920466014937857e-05,
+      "loss": 0.5016,
+      "step": 11614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3919889263777144e-05,
+      "loss": 0.5238,
+      "step": 11615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3919312512616432e-05,
+      "loss": 0.5232,
+      "step": 11616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3918735761455721e-05,
+      "loss": 0.4977,
+      "step": 11617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3918159010295009e-05,
+      "loss": 0.5142,
+      "step": 11618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3917582259134297e-05,
+      "loss": 0.6452,
+      "step": 11619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3917005507973586e-05,
+      "loss": 0.522,
+      "step": 11620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3916428756812874e-05,
+      "loss": 0.4254,
+      "step": 11621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3915852005652161e-05,
+      "loss": 0.5571,
+      "step": 11622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3915275254491451e-05,
+      "loss": 0.583,
+      "step": 11623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3914698503330739e-05,
+      "loss": 0.5434,
+      "step": 11624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3914121752170026e-05,
+      "loss": 0.5292,
+      "step": 11625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3913545001009317e-05,
+      "loss": 0.4845,
+      "step": 11626
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3912968249848603e-05,
+      "loss": 0.4972,
+      "step": 11627
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3912391498687891e-05,
+      "loss": 0.4372,
+      "step": 11628
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3911814747527182e-05,
+      "loss": 0.4302,
+      "step": 11629
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.391123799636647e-05,
+      "loss": 0.5669,
+      "step": 11630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3910661245205756e-05,
+      "loss": 0.4691,
+      "step": 11631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3910084494045047e-05,
+      "loss": 0.5148,
+      "step": 11632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3909507742884334e-05,
+      "loss": 0.5868,
+      "step": 11633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3908930991723622e-05,
+      "loss": 0.5623,
+      "step": 11634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3908354240562912e-05,
+      "loss": 0.5083,
+      "step": 11635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.39077774894022e-05,
+      "loss": 0.5742,
+      "step": 11636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3907200738241487e-05,
+      "loss": 0.5374,
+      "step": 11637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3906623987080776e-05,
+      "loss": 0.4606,
+      "step": 11638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3906047235920064e-05,
+      "loss": 0.5614,
+      "step": 11639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3905470484759352e-05,
+      "loss": 0.5168,
+      "step": 11640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.390489373359864e-05,
+      "loss": 0.5511,
+      "step": 11641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3904316982437929e-05,
+      "loss": 0.69,
+      "step": 11642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3903740231277216e-05,
+      "loss": 0.5138,
+      "step": 11643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3903163480116504e-05,
+      "loss": 0.5028,
+      "step": 11644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3902586728955793e-05,
+      "loss": 0.3977,
+      "step": 11645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3902009977795081e-05,
+      "loss": 0.5018,
+      "step": 11646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3901433226634369e-05,
+      "loss": 0.4869,
+      "step": 11647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3900856475473658e-05,
+      "loss": 0.4858,
+      "step": 11648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3900279724312946e-05,
+      "loss": 0.5173,
+      "step": 11649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3899702973152234e-05,
+      "loss": 0.5309,
+      "step": 11650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3899126221991523e-05,
+      "loss": 0.5885,
+      "step": 11651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.389854947083081e-05,
+      "loss": 0.4518,
+      "step": 11652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3897972719670098e-05,
+      "loss": 0.5294,
+      "step": 11653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3897395968509388e-05,
+      "loss": 0.5096,
+      "step": 11654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3896819217348675e-05,
+      "loss": 0.5648,
+      "step": 11655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3896242466187963e-05,
+      "loss": 0.5694,
+      "step": 11656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3895665715027254e-05,
+      "loss": 0.5001,
+      "step": 11657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.389508896386654e-05,
+      "loss": 0.5635,
+      "step": 11658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3894512212705828e-05,
+      "loss": 0.6093,
+      "step": 11659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3893935461545119e-05,
+      "loss": 0.5527,
+      "step": 11660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3893358710384407e-05,
+      "loss": 0.5415,
+      "step": 11661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3892781959223693e-05,
+      "loss": 0.431,
+      "step": 11662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3892205208062984e-05,
+      "loss": 0.4667,
+      "step": 11663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3891628456902271e-05,
+      "loss": 0.4591,
+      "step": 11664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3891051705741559e-05,
+      "loss": 0.4599,
+      "step": 11665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3890474954580848e-05,
+      "loss": 0.5181,
+      "step": 11666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3889898203420136e-05,
+      "loss": 0.5384,
+      "step": 11667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3889321452259424e-05,
+      "loss": 0.4074,
+      "step": 11668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3888744701098713e-05,
+      "loss": 0.5221,
+      "step": 11669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3888167949938001e-05,
+      "loss": 0.508,
+      "step": 11670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3887591198777289e-05,
+      "loss": 0.4578,
+      "step": 11671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3887014447616578e-05,
+      "loss": 0.5535,
+      "step": 11672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3886437696455866e-05,
+      "loss": 0.4938,
+      "step": 11673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3885860945295153e-05,
+      "loss": 0.4297,
+      "step": 11674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3885284194134441e-05,
+      "loss": 0.5602,
+      "step": 11675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.388470744297373e-05,
+      "loss": 0.425,
+      "step": 11676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3884130691813018e-05,
+      "loss": 0.4498,
+      "step": 11677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3883553940652306e-05,
+      "loss": 0.4954,
+      "step": 11678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3882977189491595e-05,
+      "loss": 0.3953,
+      "step": 11679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3882400438330883e-05,
+      "loss": 0.4929,
+      "step": 11680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.388182368717017e-05,
+      "loss": 0.5363,
+      "step": 11681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.388124693600946e-05,
+      "loss": 0.5522,
+      "step": 11682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3880670184848748e-05,
+      "loss": 0.4346,
+      "step": 11683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3880093433688035e-05,
+      "loss": 0.416,
+      "step": 11684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3879516682527325e-05,
+      "loss": 0.5424,
+      "step": 11685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3878939931366612e-05,
+      "loss": 0.3962,
+      "step": 11686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.38783631802059e-05,
+      "loss": 0.5166,
+      "step": 11687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3877786429045191e-05,
+      "loss": 0.6052,
+      "step": 11688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3877209677884477e-05,
+      "loss": 0.52,
+      "step": 11689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3876632926723765e-05,
+      "loss": 0.518,
+      "step": 11690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3876056175563056e-05,
+      "loss": 0.4969,
+      "step": 11691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3875479424402343e-05,
+      "loss": 0.5762,
+      "step": 11692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.387490267324163e-05,
+      "loss": 0.4636,
+      "step": 11693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.387432592208092e-05,
+      "loss": 0.4134,
+      "step": 11694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3873749170920208e-05,
+      "loss": 0.5534,
+      "step": 11695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3873172419759496e-05,
+      "loss": 0.4954,
+      "step": 11696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3872595668598785e-05,
+      "loss": 0.5535,
+      "step": 11697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3872018917438073e-05,
+      "loss": 0.3779,
+      "step": 11698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.387144216627736e-05,
+      "loss": 0.4579,
+      "step": 11699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.387086541511665e-05,
+      "loss": 0.4025,
+      "step": 11700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3870288663955938e-05,
+      "loss": 0.5363,
+      "step": 11701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3869711912795225e-05,
+      "loss": 0.4407,
+      "step": 11702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3869135161634515e-05,
+      "loss": 0.5323,
+      "step": 11703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3868558410473802e-05,
+      "loss": 0.5013,
+      "step": 11704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.386798165931309e-05,
+      "loss": 0.503,
+      "step": 11705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.386740490815238e-05,
+      "loss": 0.3903,
+      "step": 11706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3866828156991667e-05,
+      "loss": 0.5068,
+      "step": 11707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.3866251405830955e-05,
+      "loss": 0.5833,
+      "step": 11708
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3865674654670244e-05,
+      "loss": 0.4282,
+      "step": 11709
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3865097903509532e-05,
+      "loss": 0.3703,
+      "step": 11710
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.386452115234882e-05,
+      "loss": 0.4733,
+      "step": 11711
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3863944401188107e-05,
+      "loss": 0.5988,
+      "step": 11712
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3863367650027397e-05,
+      "loss": 0.5157,
+      "step": 11713
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3862790898866684e-05,
+      "loss": 0.5845,
+      "step": 11714
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3862214147705972e-05,
+      "loss": 0.4373,
+      "step": 11715
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3861637396545261e-05,
+      "loss": 0.3683,
+      "step": 11716
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.386106064538455e-05,
+      "loss": 0.4651,
+      "step": 11717
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3860483894223837e-05,
+      "loss": 0.5773,
+      "step": 11718
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3859907143063128e-05,
+      "loss": 0.4295,
+      "step": 11719
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3859330391902414e-05,
+      "loss": 0.4748,
+      "step": 11720
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3858753640741702e-05,
+      "loss": 0.4836,
+      "step": 11721
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3858176889580993e-05,
+      "loss": 0.4798,
+      "step": 11722
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.385760013842028e-05,
+      "loss": 0.4317,
+      "step": 11723
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3857023387259568e-05,
+      "loss": 0.57,
+      "step": 11724
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3856446636098857e-05,
+      "loss": 0.469,
+      "step": 11725
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3855869884938145e-05,
+      "loss": 0.6163,
+      "step": 11726
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3855293133777433e-05,
+      "loss": 0.5476,
+      "step": 11727
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3854716382616722e-05,
+      "loss": 0.4798,
+      "step": 11728
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.385413963145601e-05,
+      "loss": 0.6083,
+      "step": 11729
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3853562880295298e-05,
+      "loss": 0.6786,
+      "step": 11730
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3852986129134587e-05,
+      "loss": 0.5056,
+      "step": 11731
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3852409377973875e-05,
+      "loss": 0.5531,
+      "step": 11732
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3851832626813162e-05,
+      "loss": 0.5446,
+      "step": 11733
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3851255875652452e-05,
+      "loss": 0.618,
+      "step": 11734
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.385067912449174e-05,
+      "loss": 0.536,
+      "step": 11735
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3850102373331027e-05,
+      "loss": 0.5219,
+      "step": 11736
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3849525622170316e-05,
+      "loss": 0.4904,
+      "step": 11737
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3848948871009604e-05,
+      "loss": 0.6385,
+      "step": 11738
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3848372119848892e-05,
+      "loss": 0.4777,
+      "step": 11739
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3847795368688181e-05,
+      "loss": 0.4996,
+      "step": 11740
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3847218617527469e-05,
+      "loss": 0.542,
+      "step": 11741
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3846641866366757e-05,
+      "loss": 0.5273,
+      "step": 11742
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3846065115206046e-05,
+      "loss": 0.5436,
+      "step": 11743
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3845488364045334e-05,
+      "loss": 0.4598,
+      "step": 11744
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3844911612884621e-05,
+      "loss": 0.5204,
+      "step": 11745
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3844334861723909e-05,
+      "loss": 0.5531,
+      "step": 11746
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3843758110563198e-05,
+      "loss": 0.573,
+      "step": 11747
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3843181359402486e-05,
+      "loss": 0.3434,
+      "step": 11748
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3842604608241774e-05,
+      "loss": 0.5014,
+      "step": 11749
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3842027857081065e-05,
+      "loss": 0.5026,
+      "step": 11750
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.384145110592035e-05,
+      "loss": 0.631,
+      "step": 11751
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3840874354759638e-05,
+      "loss": 0.4896,
+      "step": 11752
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.384029760359893e-05,
+      "loss": 0.5057,
+      "step": 11753
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3839720852438217e-05,
+      "loss": 0.4499,
+      "step": 11754
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3839144101277505e-05,
+      "loss": 0.4366,
+      "step": 11755
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3838567350116794e-05,
+      "loss": 0.5243,
+      "step": 11756
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3837990598956082e-05,
+      "loss": 0.5631,
+      "step": 11757
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.383741384779537e-05,
+      "loss": 0.5508,
+      "step": 11758
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3836837096634659e-05,
+      "loss": 0.5453,
+      "step": 11759
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3836260345473947e-05,
+      "loss": 0.5243,
+      "step": 11760
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3835683594313234e-05,
+      "loss": 0.5436,
+      "step": 11761
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3835106843152524e-05,
+      "loss": 0.4996,
+      "step": 11762
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3834530091991811e-05,
+      "loss": 0.4466,
+      "step": 11763
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.38339533408311e-05,
+      "loss": 0.4969,
+      "step": 11764
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3833376589670389e-05,
+      "loss": 0.4413,
+      "step": 11765
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3832799838509676e-05,
+      "loss": 0.4575,
+      "step": 11766
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3832223087348964e-05,
+      "loss": 0.543,
+      "step": 11767
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3831646336188253e-05,
+      "loss": 0.5229,
+      "step": 11768
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3831069585027541e-05,
+      "loss": 0.5203,
+      "step": 11769
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3830492833866829e-05,
+      "loss": 0.5146,
+      "step": 11770
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3829916082706118e-05,
+      "loss": 0.66,
+      "step": 11771
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3829339331545406e-05,
+      "loss": 0.5118,
+      "step": 11772
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3828762580384693e-05,
+      "loss": 0.4459,
+      "step": 11773
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3828185829223983e-05,
+      "loss": 0.4168,
+      "step": 11774
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.382760907806327e-05,
+      "loss": 0.3744,
+      "step": 11775
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3827032326902558e-05,
+      "loss": 0.5571,
+      "step": 11776
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3826455575741848e-05,
+      "loss": 0.5165,
+      "step": 11777
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3825878824581135e-05,
+      "loss": 0.4926,
+      "step": 11778
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3825302073420423e-05,
+      "loss": 0.541,
+      "step": 11779
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.382472532225971e-05,
+      "loss": 0.4091,
+      "step": 11780
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3824148571099002e-05,
+      "loss": 0.4327,
+      "step": 11781
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3823571819938288e-05,
+      "loss": 0.496,
+      "step": 11782
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3822995068777575e-05,
+      "loss": 0.6213,
+      "step": 11783
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3822418317616866e-05,
+      "loss": 0.4458,
+      "step": 11784
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3821841566456154e-05,
+      "loss": 0.4702,
+      "step": 11785
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3821264815295442e-05,
+      "loss": 0.5499,
+      "step": 11786
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3820688064134731e-05,
+      "loss": 0.6311,
+      "step": 11787
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3820111312974019e-05,
+      "loss": 0.5446,
+      "step": 11788
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3819534561813307e-05,
+      "loss": 0.4934,
+      "step": 11789
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3818957810652596e-05,
+      "loss": 0.5989,
+      "step": 11790
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3818381059491884e-05,
+      "loss": 0.5466,
+      "step": 11791
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3817804308331171e-05,
+      "loss": 0.5509,
+      "step": 11792
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.381722755717046e-05,
+      "loss": 0.3994,
+      "step": 11793
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3816650806009748e-05,
+      "loss": 0.6852,
+      "step": 11794
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3816074054849036e-05,
+      "loss": 0.5072,
+      "step": 11795
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3815497303688325e-05,
+      "loss": 0.444,
+      "step": 11796
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3814920552527613e-05,
+      "loss": 0.4397,
+      "step": 11797
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.38143438013669e-05,
+      "loss": 0.3928,
+      "step": 11798
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.381376705020619e-05,
+      "loss": 0.6286,
+      "step": 11799
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3813190299045478e-05,
+      "loss": 0.4576,
+      "step": 11800
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3812613547884766e-05,
+      "loss": 0.6108,
+      "step": 11801
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3812036796724055e-05,
+      "loss": 0.4898,
+      "step": 11802
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3811460045563343e-05,
+      "loss": 0.5725,
+      "step": 11803
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.381088329440263e-05,
+      "loss": 0.4742,
+      "step": 11804
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.381030654324192e-05,
+      "loss": 0.3799,
+      "step": 11805
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3809729792081207e-05,
+      "loss": 0.4345,
+      "step": 11806
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3809153040920495e-05,
+      "loss": 0.5062,
+      "step": 11807
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3808576289759786e-05,
+      "loss": 0.6301,
+      "step": 11808
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3807999538599072e-05,
+      "loss": 0.437,
+      "step": 11809
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.380742278743836e-05,
+      "loss": 0.5959,
+      "step": 11810
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3806846036277651e-05,
+      "loss": 0.4816,
+      "step": 11811
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3806269285116939e-05,
+      "loss": 0.5241,
+      "step": 11812
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3805692533956225e-05,
+      "loss": 0.5489,
+      "step": 11813
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3805115782795512e-05,
+      "loss": 0.4771,
+      "step": 11814
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3804539031634803e-05,
+      "loss": 0.5667,
+      "step": 11815
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3803962280474091e-05,
+      "loss": 0.6067,
+      "step": 11816
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3803385529313379e-05,
+      "loss": 0.488,
+      "step": 11817
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3802808778152668e-05,
+      "loss": 0.637,
+      "step": 11818
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3802232026991956e-05,
+      "loss": 0.7199,
+      "step": 11819
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3801655275831243e-05,
+      "loss": 0.6039,
+      "step": 11820
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3801078524670533e-05,
+      "loss": 0.4756,
+      "step": 11821
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.380050177350982e-05,
+      "loss": 0.377,
+      "step": 11822
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3799925022349108e-05,
+      "loss": 0.4713,
+      "step": 11823
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3799348271188398e-05,
+      "loss": 0.4039,
+      "step": 11824
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3798771520027685e-05,
+      "loss": 0.451,
+      "step": 11825
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3798194768866973e-05,
+      "loss": 0.6812,
+      "step": 11826
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3797618017706262e-05,
+      "loss": 0.4282,
+      "step": 11827
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.379704126654555e-05,
+      "loss": 0.4511,
+      "step": 11828
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3796464515384838e-05,
+      "loss": 0.5707,
+      "step": 11829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3795887764224127e-05,
+      "loss": 0.6167,
+      "step": 11830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3795311013063415e-05,
+      "loss": 0.5224,
+      "step": 11831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3794734261902702e-05,
+      "loss": 0.5469,
+      "step": 11832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3794157510741992e-05,
+      "loss": 0.3579,
+      "step": 11833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.379358075958128e-05,
+      "loss": 0.4907,
+      "step": 11834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3793004008420567e-05,
+      "loss": 0.4222,
+      "step": 11835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3792427257259857e-05,
+      "loss": 0.3613,
+      "step": 11836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3791850506099144e-05,
+      "loss": 0.5864,
+      "step": 11837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3791273754938432e-05,
+      "loss": 0.6047,
+      "step": 11838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3790697003777723e-05,
+      "loss": 0.4664,
+      "step": 11839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3790120252617009e-05,
+      "loss": 0.4771,
+      "step": 11840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3789543501456297e-05,
+      "loss": 0.4521,
+      "step": 11841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3788966750295588e-05,
+      "loss": 0.4876,
+      "step": 11842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3788389999134875e-05,
+      "loss": 0.4377,
+      "step": 11843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3787813247974161e-05,
+      "loss": 0.471,
+      "step": 11844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3787236496813452e-05,
+      "loss": 0.4235,
+      "step": 11845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.378665974565274e-05,
+      "loss": 0.5887,
+      "step": 11846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3786082994492028e-05,
+      "loss": 0.5504,
+      "step": 11847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3785506243331317e-05,
+      "loss": 0.5471,
+      "step": 11848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3784929492170605e-05,
+      "loss": 0.5449,
+      "step": 11849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3784352741009893e-05,
+      "loss": 0.4808,
+      "step": 11850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.378377598984918e-05,
+      "loss": 0.5765,
+      "step": 11851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.378319923868847e-05,
+      "loss": 0.5543,
+      "step": 11852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3782622487527757e-05,
+      "loss": 0.5401,
+      "step": 11853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3782045736367045e-05,
+      "loss": 0.564,
+      "step": 11854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3781468985206334e-05,
+      "loss": 0.5872,
+      "step": 11855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3780892234045622e-05,
+      "loss": 0.6828,
+      "step": 11856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.378031548288491e-05,
+      "loss": 0.5234,
+      "step": 11857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.37797387317242e-05,
+      "loss": 0.6323,
+      "step": 11858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3779161980563487e-05,
+      "loss": 0.55,
+      "step": 11859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3778585229402775e-05,
+      "loss": 0.5826,
+      "step": 11860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3778008478242064e-05,
+      "loss": 0.5107,
+      "step": 11861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3777431727081352e-05,
+      "loss": 0.496,
+      "step": 11862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.377685497592064e-05,
+      "loss": 0.6432,
+      "step": 11863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3776278224759929e-05,
+      "loss": 0.448,
+      "step": 11864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3775701473599216e-05,
+      "loss": 0.5757,
+      "step": 11865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3775124722438504e-05,
+      "loss": 0.4466,
+      "step": 11866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3774547971277793e-05,
+      "loss": 0.3828,
+      "step": 11867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3773971220117081e-05,
+      "loss": 0.4634,
+      "step": 11868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3773394468956369e-05,
+      "loss": 0.4769,
+      "step": 11869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.377281771779566e-05,
+      "loss": 0.4944,
+      "step": 11870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3772240966634946e-05,
+      "loss": 0.3382,
+      "step": 11871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3771664215474234e-05,
+      "loss": 0.4375,
+      "step": 11872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3771087464313525e-05,
+      "loss": 0.6653,
+      "step": 11873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3770510713152812e-05,
+      "loss": 0.4526,
+      "step": 11874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3769933961992098e-05,
+      "loss": 0.5205,
+      "step": 11875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.376935721083139e-05,
+      "loss": 0.5334,
+      "step": 11876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3768780459670677e-05,
+      "loss": 0.4711,
+      "step": 11877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3768203708509965e-05,
+      "loss": 0.453,
+      "step": 11878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3767626957349254e-05,
+      "loss": 0.5696,
+      "step": 11879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3767050206188542e-05,
+      "loss": 0.5518,
+      "step": 11880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.376647345502783e-05,
+      "loss": 0.6104,
+      "step": 11881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3765896703867119e-05,
+      "loss": 0.4685,
+      "step": 11882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3765319952706407e-05,
+      "loss": 0.5875,
+      "step": 11883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3764743201545694e-05,
+      "loss": 0.5995,
+      "step": 11884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3764166450384982e-05,
+      "loss": 0.5217,
+      "step": 11885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 1.3763589699224271e-05,
+      "loss": 0.6196,
+      "step": 11886
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3763012948063559e-05,
+      "loss": 0.526,
+      "step": 11887
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3762436196902847e-05,
+      "loss": 0.5407,
+      "step": 11888
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3761859445742136e-05,
+      "loss": 0.5356,
+      "step": 11889
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3761282694581424e-05,
+      "loss": 0.5016,
+      "step": 11890
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3760705943420711e-05,
+      "loss": 0.5688,
+      "step": 11891
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.376012919226e-05,
+      "loss": 0.4727,
+      "step": 11892
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3759552441099288e-05,
+      "loss": 0.5421,
+      "step": 11893
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3758975689938576e-05,
+      "loss": 0.5091,
+      "step": 11894
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3758398938777866e-05,
+      "loss": 0.5212,
+      "step": 11895
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3757822187617153e-05,
+      "loss": 0.4603,
+      "step": 11896
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3757245436456441e-05,
+      "loss": 0.4366,
+      "step": 11897
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.375666868529573e-05,
+      "loss": 0.5158,
+      "step": 11898
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3756091934135018e-05,
+      "loss": 0.5436,
+      "step": 11899
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3755515182974306e-05,
+      "loss": 0.4635,
+      "step": 11900
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3754938431813597e-05,
+      "loss": 0.4962,
+      "step": 11901
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3754361680652883e-05,
+      "loss": 0.4962,
+      "step": 11902
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.375378492949217e-05,
+      "loss": 0.4787,
+      "step": 11903
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3753208178331461e-05,
+      "loss": 0.5483,
+      "step": 11904
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.375263142717075e-05,
+      "loss": 0.5131,
+      "step": 11905
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3752054676010035e-05,
+      "loss": 0.3616,
+      "step": 11906
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3751477924849326e-05,
+      "loss": 0.5976,
+      "step": 11907
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3750901173688614e-05,
+      "loss": 0.6948,
+      "step": 11908
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3750324422527902e-05,
+      "loss": 0.5102,
+      "step": 11909
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3749747671367191e-05,
+      "loss": 0.4766,
+      "step": 11910
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3749170920206479e-05,
+      "loss": 0.5074,
+      "step": 11911
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3748594169045766e-05,
+      "loss": 0.5975,
+      "step": 11912
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3748017417885056e-05,
+      "loss": 0.5243,
+      "step": 11913
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3747440666724343e-05,
+      "loss": 0.5331,
+      "step": 11914
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3746863915563631e-05,
+      "loss": 0.5324,
+      "step": 11915
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.374628716440292e-05,
+      "loss": 0.5304,
+      "step": 11916
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3745710413242208e-05,
+      "loss": 0.5615,
+      "step": 11917
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3745133662081496e-05,
+      "loss": 0.3757,
+      "step": 11918
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3744556910920784e-05,
+      "loss": 0.5877,
+      "step": 11919
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3743980159760073e-05,
+      "loss": 0.4901,
+      "step": 11920
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.374340340859936e-05,
+      "loss": 0.4839,
+      "step": 11921
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3742826657438648e-05,
+      "loss": 0.524,
+      "step": 11922
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3742249906277938e-05,
+      "loss": 0.453,
+      "step": 11923
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3741673155117225e-05,
+      "loss": 0.5919,
+      "step": 11924
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3741096403956513e-05,
+      "loss": 0.3922,
+      "step": 11925
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3740519652795802e-05,
+      "loss": 0.6244,
+      "step": 11926
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.373994290163509e-05,
+      "loss": 0.5567,
+      "step": 11927
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3739366150474378e-05,
+      "loss": 0.4438,
+      "step": 11928
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3738789399313667e-05,
+      "loss": 0.5335,
+      "step": 11929
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3738212648152955e-05,
+      "loss": 0.5138,
+      "step": 11930
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3737635896992243e-05,
+      "loss": 0.5487,
+      "step": 11931
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3737059145831534e-05,
+      "loss": 0.4078,
+      "step": 11932
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.373648239467082e-05,
+      "loss": 0.585,
+      "step": 11933
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3735905643510107e-05,
+      "loss": 0.4049,
+      "step": 11934
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3735328892349398e-05,
+      "loss": 0.6241,
+      "step": 11935
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3734752141188686e-05,
+      "loss": 0.6231,
+      "step": 11936
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3734175390027972e-05,
+      "loss": 0.349,
+      "step": 11937
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3733598638867263e-05,
+      "loss": 0.5409,
+      "step": 11938
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.373302188770655e-05,
+      "loss": 0.4417,
+      "step": 11939
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3732445136545838e-05,
+      "loss": 0.5265,
+      "step": 11940
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3731868385385128e-05,
+      "loss": 0.5361,
+      "step": 11941
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3731291634224416e-05,
+      "loss": 0.4565,
+      "step": 11942
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3730714883063703e-05,
+      "loss": 0.4998,
+      "step": 11943
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3730138131902993e-05,
+      "loss": 0.4606,
+      "step": 11944
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.372956138074228e-05,
+      "loss": 0.5474,
+      "step": 11945
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3728984629581568e-05,
+      "loss": 0.4924,
+      "step": 11946
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3728407878420857e-05,
+      "loss": 0.3295,
+      "step": 11947
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3727831127260145e-05,
+      "loss": 0.5057,
+      "step": 11948
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3727254376099433e-05,
+      "loss": 0.5261,
+      "step": 11949
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3726677624938722e-05,
+      "loss": 0.4679,
+      "step": 11950
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.372610087377801e-05,
+      "loss": 0.4515,
+      "step": 11951
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3725524122617297e-05,
+      "loss": 0.3515,
+      "step": 11952
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3724947371456587e-05,
+      "loss": 0.5695,
+      "step": 11953
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3724370620295875e-05,
+      "loss": 0.414,
+      "step": 11954
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3723793869135162e-05,
+      "loss": 0.4193,
+      "step": 11955
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.372321711797445e-05,
+      "loss": 0.601,
+      "step": 11956
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.372264036681374e-05,
+      "loss": 0.5223,
+      "step": 11957
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3722063615653027e-05,
+      "loss": 0.5652,
+      "step": 11958
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3721486864492315e-05,
+      "loss": 0.5144,
+      "step": 11959
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3720910113331604e-05,
+      "loss": 0.4492,
+      "step": 11960
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3720333362170892e-05,
+      "loss": 0.5116,
+      "step": 11961
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.371975661101018e-05,
+      "loss": 0.5231,
+      "step": 11962
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.371917985984947e-05,
+      "loss": 0.4931,
+      "step": 11963
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3718603108688756e-05,
+      "loss": 0.5403,
+      "step": 11964
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3718026357528044e-05,
+      "loss": 0.5347,
+      "step": 11965
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3717449606367335e-05,
+      "loss": 0.4244,
+      "step": 11966
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3716872855206623e-05,
+      "loss": 0.6255,
+      "step": 11967
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3716296104045909e-05,
+      "loss": 0.4251,
+      "step": 11968
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.37157193528852e-05,
+      "loss": 0.6901,
+      "step": 11969
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3715142601724488e-05,
+      "loss": 0.5262,
+      "step": 11970
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3714565850563775e-05,
+      "loss": 0.4969,
+      "step": 11971
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3713989099403065e-05,
+      "loss": 0.5556,
+      "step": 11972
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3713412348242352e-05,
+      "loss": 0.5072,
+      "step": 11973
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.371283559708164e-05,
+      "loss": 0.5145,
+      "step": 11974
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.371225884592093e-05,
+      "loss": 0.4629,
+      "step": 11975
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3711682094760217e-05,
+      "loss": 0.5123,
+      "step": 11976
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3711105343599505e-05,
+      "loss": 0.5609,
+      "step": 11977
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3710528592438794e-05,
+      "loss": 0.3983,
+      "step": 11978
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3709951841278082e-05,
+      "loss": 0.4977,
+      "step": 11979
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.370937509011737e-05,
+      "loss": 0.4881,
+      "step": 11980
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3708798338956659e-05,
+      "loss": 0.4684,
+      "step": 11981
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3708221587795947e-05,
+      "loss": 0.4717,
+      "step": 11982
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3707644836635234e-05,
+      "loss": 0.4982,
+      "step": 11983
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3707068085474524e-05,
+      "loss": 0.4793,
+      "step": 11984
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3706491334313811e-05,
+      "loss": 0.5653,
+      "step": 11985
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3705914583153099e-05,
+      "loss": 0.5664,
+      "step": 11986
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3705337831992388e-05,
+      "loss": 0.3977,
+      "step": 11987
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3704761080831676e-05,
+      "loss": 0.4472,
+      "step": 11988
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3704184329670964e-05,
+      "loss": 0.4033,
+      "step": 11989
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3703607578510252e-05,
+      "loss": 0.4435,
+      "step": 11990
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3703030827349541e-05,
+      "loss": 0.428,
+      "step": 11991
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3702454076188829e-05,
+      "loss": 0.5929,
+      "step": 11992
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3701877325028116e-05,
+      "loss": 0.5521,
+      "step": 11993
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3701300573867407e-05,
+      "loss": 0.4942,
+      "step": 11994
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3700723822706693e-05,
+      "loss": 0.4664,
+      "step": 11995
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3700147071545981e-05,
+      "loss": 0.4998,
+      "step": 11996
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3699570320385272e-05,
+      "loss": 0.4012,
+      "step": 11997
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.369899356922456e-05,
+      "loss": 0.4902,
+      "step": 11998
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3698416818063846e-05,
+      "loss": 0.5187,
+      "step": 11999
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3697840066903137e-05,
+      "loss": 0.4998,
+      "step": 12000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3697263315742425e-05,
+      "loss": 0.4985,
+      "step": 12001
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3696686564581712e-05,
+      "loss": 0.6267,
+      "step": 12002
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3696109813421002e-05,
+      "loss": 0.471,
+      "step": 12003
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.369553306226029e-05,
+      "loss": 0.5241,
+      "step": 12004
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3694956311099577e-05,
+      "loss": 0.5143,
+      "step": 12005
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3694379559938866e-05,
+      "loss": 0.4948,
+      "step": 12006
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3693802808778154e-05,
+      "loss": 0.5225,
+      "step": 12007
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3693226057617442e-05,
+      "loss": 0.4754,
+      "step": 12008
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3692649306456731e-05,
+      "loss": 0.4845,
+      "step": 12009
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3692072555296019e-05,
+      "loss": 0.5427,
+      "step": 12010
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3691495804135306e-05,
+      "loss": 0.5184,
+      "step": 12011
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3690919052974596e-05,
+      "loss": 0.567,
+      "step": 12012
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3690342301813884e-05,
+      "loss": 0.4669,
+      "step": 12013
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3689765550653171e-05,
+      "loss": 0.6158,
+      "step": 12014
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.368918879949246e-05,
+      "loss": 0.5114,
+      "step": 12015
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3688612048331748e-05,
+      "loss": 0.4642,
+      "step": 12016
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3688035297171036e-05,
+      "loss": 0.4208,
+      "step": 12017
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3687458546010325e-05,
+      "loss": 0.4747,
+      "step": 12018
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3686881794849613e-05,
+      "loss": 0.5388,
+      "step": 12019
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.36863050436889e-05,
+      "loss": 0.5595,
+      "step": 12020
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.368572829252819e-05,
+      "loss": 0.6735,
+      "step": 12021
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3685151541367478e-05,
+      "loss": 0.5241,
+      "step": 12022
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3684574790206765e-05,
+      "loss": 0.4845,
+      "step": 12023
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3683998039046053e-05,
+      "loss": 0.5599,
+      "step": 12024
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3683421287885344e-05,
+      "loss": 0.4726,
+      "step": 12025
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.368284453672463e-05,
+      "loss": 0.42,
+      "step": 12026
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3682267785563918e-05,
+      "loss": 0.4904,
+      "step": 12027
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3681691034403209e-05,
+      "loss": 0.5825,
+      "step": 12028
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3681114283242497e-05,
+      "loss": 0.5448,
+      "step": 12029
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3680537532081783e-05,
+      "loss": 0.4869,
+      "step": 12030
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3679960780921074e-05,
+      "loss": 0.4787,
+      "step": 12031
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3679384029760361e-05,
+      "loss": 0.519,
+      "step": 12032
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3678807278599649e-05,
+      "loss": 0.4985,
+      "step": 12033
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3678230527438938e-05,
+      "loss": 0.5438,
+      "step": 12034
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3677653776278226e-05,
+      "loss": 0.5048,
+      "step": 12035
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3677077025117514e-05,
+      "loss": 0.4867,
+      "step": 12036
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3676500273956803e-05,
+      "loss": 0.4818,
+      "step": 12037
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3675923522796091e-05,
+      "loss": 0.5435,
+      "step": 12038
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3675346771635379e-05,
+      "loss": 0.5584,
+      "step": 12039
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3674770020474668e-05,
+      "loss": 0.5109,
+      "step": 12040
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3674193269313956e-05,
+      "loss": 0.495,
+      "step": 12041
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3673616518153243e-05,
+      "loss": 0.5136,
+      "step": 12042
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3673039766992533e-05,
+      "loss": 0.4504,
+      "step": 12043
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.367246301583182e-05,
+      "loss": 0.5496,
+      "step": 12044
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3671886264671108e-05,
+      "loss": 0.4695,
+      "step": 12045
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3671309513510397e-05,
+      "loss": 0.5283,
+      "step": 12046
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3670732762349685e-05,
+      "loss": 0.4876,
+      "step": 12047
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3670156011188973e-05,
+      "loss": 0.4217,
+      "step": 12048
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3669579260028262e-05,
+      "loss": 0.5185,
+      "step": 12049
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.366900250886755e-05,
+      "loss": 0.4467,
+      "step": 12050
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3668425757706838e-05,
+      "loss": 0.5404,
+      "step": 12051
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3667849006546127e-05,
+      "loss": 0.451,
+      "step": 12052
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3667272255385415e-05,
+      "loss": 0.5991,
+      "step": 12053
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3666695504224702e-05,
+      "loss": 0.5376,
+      "step": 12054
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3666118753063993e-05,
+      "loss": 0.5445,
+      "step": 12055
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3665542001903281e-05,
+      "loss": 0.5457,
+      "step": 12056
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3664965250742567e-05,
+      "loss": 0.5116,
+      "step": 12057
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3664388499581855e-05,
+      "loss": 0.5003,
+      "step": 12058
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3663811748421146e-05,
+      "loss": 0.5152,
+      "step": 12059
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3663234997260434e-05,
+      "loss": 0.5428,
+      "step": 12060
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.366265824609972e-05,
+      "loss": 0.5937,
+      "step": 12061
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.366208149493901e-05,
+      "loss": 0.4198,
+      "step": 12062
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3661504743778298e-05,
+      "loss": 0.4604,
+      "step": 12063
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3660927992617586e-05,
+      "loss": 0.6497,
+      "step": 12064
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 1.3660351241456875e-05,
+      "loss": 0.481,
+      "step": 12065
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3659774490296163e-05,
+      "loss": 0.4633,
+      "step": 12066
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.365919773913545e-05,
+      "loss": 0.4968,
+      "step": 12067
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.365862098797474e-05,
+      "loss": 0.5123,
+      "step": 12068
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3658044236814028e-05,
+      "loss": 0.5145,
+      "step": 12069
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3657467485653315e-05,
+      "loss": 0.6168,
+      "step": 12070
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3656890734492605e-05,
+      "loss": 0.5357,
+      "step": 12071
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3656313983331893e-05,
+      "loss": 0.4732,
+      "step": 12072
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.365573723217118e-05,
+      "loss": 0.5481,
+      "step": 12073
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.365516048101047e-05,
+      "loss": 0.5023,
+      "step": 12074
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3654583729849757e-05,
+      "loss": 0.6242,
+      "step": 12075
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3654006978689045e-05,
+      "loss": 0.4184,
+      "step": 12076
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3653430227528334e-05,
+      "loss": 0.4907,
+      "step": 12077
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3652853476367622e-05,
+      "loss": 0.4138,
+      "step": 12078
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.365227672520691e-05,
+      "loss": 0.4894,
+      "step": 12079
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3651699974046199e-05,
+      "loss": 0.5039,
+      "step": 12080
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3651123222885487e-05,
+      "loss": 0.4285,
+      "step": 12081
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3650546471724774e-05,
+      "loss": 0.5939,
+      "step": 12082
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3649969720564064e-05,
+      "loss": 0.5065,
+      "step": 12083
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3649392969403352e-05,
+      "loss": 0.4756,
+      "step": 12084
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.364881621824264e-05,
+      "loss": 0.6075,
+      "step": 12085
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.364823946708193e-05,
+      "loss": 0.6465,
+      "step": 12086
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3647662715921218e-05,
+      "loss": 0.3937,
+      "step": 12087
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3647085964760504e-05,
+      "loss": 0.3987,
+      "step": 12088
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3646509213599795e-05,
+      "loss": 0.4683,
+      "step": 12089
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3645932462439083e-05,
+      "loss": 0.5663,
+      "step": 12090
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.364535571127837e-05,
+      "loss": 0.387,
+      "step": 12091
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.364477896011766e-05,
+      "loss": 0.4995,
+      "step": 12092
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3644202208956947e-05,
+      "loss": 0.4483,
+      "step": 12093
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3643625457796235e-05,
+      "loss": 0.4452,
+      "step": 12094
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3643048706635523e-05,
+      "loss": 0.5151,
+      "step": 12095
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3642471955474812e-05,
+      "loss": 0.4354,
+      "step": 12096
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.36418952043141e-05,
+      "loss": 0.6118,
+      "step": 12097
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3641318453153388e-05,
+      "loss": 0.5939,
+      "step": 12098
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3640741701992677e-05,
+      "loss": 0.6057,
+      "step": 12099
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3640164950831965e-05,
+      "loss": 0.4971,
+      "step": 12100
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3639588199671252e-05,
+      "loss": 0.456,
+      "step": 12101
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3639011448510542e-05,
+      "loss": 0.45,
+      "step": 12102
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.363843469734983e-05,
+      "loss": 0.5221,
+      "step": 12103
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3637857946189117e-05,
+      "loss": 0.435,
+      "step": 12104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3637281195028406e-05,
+      "loss": 0.6791,
+      "step": 12105
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3636704443867694e-05,
+      "loss": 0.4489,
+      "step": 12106
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3636127692706982e-05,
+      "loss": 0.6307,
+      "step": 12107
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3635550941546271e-05,
+      "loss": 0.501,
+      "step": 12108
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3634974190385559e-05,
+      "loss": 0.5794,
+      "step": 12109
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3634397439224847e-05,
+      "loss": 0.5395,
+      "step": 12110
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3633820688064136e-05,
+      "loss": 0.4948,
+      "step": 12111
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3633243936903424e-05,
+      "loss": 0.63,
+      "step": 12112
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3632667185742711e-05,
+      "loss": 0.5226,
+      "step": 12113
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3632090434582e-05,
+      "loss": 0.4851,
+      "step": 12114
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3631513683421288e-05,
+      "loss": 0.4564,
+      "step": 12115
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3630936932260576e-05,
+      "loss": 0.6748,
+      "step": 12116
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3630360181099867e-05,
+      "loss": 0.5086,
+      "step": 12117
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3629783429939155e-05,
+      "loss": 0.438,
+      "step": 12118
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.362920667877844e-05,
+      "loss": 0.397,
+      "step": 12119
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3628629927617732e-05,
+      "loss": 0.4793,
+      "step": 12120
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.362805317645702e-05,
+      "loss": 0.5234,
+      "step": 12121
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3627476425296307e-05,
+      "loss": 0.5651,
+      "step": 12122
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3626899674135597e-05,
+      "loss": 0.4039,
+      "step": 12123
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3626322922974884e-05,
+      "loss": 0.3357,
+      "step": 12124
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3625746171814172e-05,
+      "loss": 0.5137,
+      "step": 12125
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3625169420653461e-05,
+      "loss": 0.5403,
+      "step": 12126
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3624592669492749e-05,
+      "loss": 0.3586,
+      "step": 12127
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3624015918332037e-05,
+      "loss": 0.4177,
+      "step": 12128
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3623439167171324e-05,
+      "loss": 0.5463,
+      "step": 12129
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3622862416010614e-05,
+      "loss": 0.5245,
+      "step": 12130
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3622285664849902e-05,
+      "loss": 0.5466,
+      "step": 12131
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.362170891368919e-05,
+      "loss": 0.5977,
+      "step": 12132
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3621132162528479e-05,
+      "loss": 0.5732,
+      "step": 12133
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3620555411367766e-05,
+      "loss": 0.4782,
+      "step": 12134
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3619978660207054e-05,
+      "loss": 0.3898,
+      "step": 12135
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3619401909046343e-05,
+      "loss": 0.4955,
+      "step": 12136
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3618825157885631e-05,
+      "loss": 0.471,
+      "step": 12137
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3618248406724919e-05,
+      "loss": 0.4976,
+      "step": 12138
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3617671655564208e-05,
+      "loss": 0.5057,
+      "step": 12139
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3617094904403496e-05,
+      "loss": 0.5348,
+      "step": 12140
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3616518153242783e-05,
+      "loss": 0.4627,
+      "step": 12141
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3615941402082073e-05,
+      "loss": 0.4797,
+      "step": 12142
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.361536465092136e-05,
+      "loss": 0.5101,
+      "step": 12143
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3614787899760648e-05,
+      "loss": 0.4096,
+      "step": 12144
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3614211148599938e-05,
+      "loss": 0.552,
+      "step": 12145
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3613634397439225e-05,
+      "loss": 0.5811,
+      "step": 12146
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3613057646278513e-05,
+      "loss": 0.5681,
+      "step": 12147
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3612480895117804e-05,
+      "loss": 0.5136,
+      "step": 12148
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3611904143957092e-05,
+      "loss": 0.4696,
+      "step": 12149
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3611327392796378e-05,
+      "loss": 0.3708,
+      "step": 12150
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3610750641635669e-05,
+      "loss": 0.5793,
+      "step": 12151
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3610173890474956e-05,
+      "loss": 0.6525,
+      "step": 12152
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3609597139314244e-05,
+      "loss": 0.4948,
+      "step": 12153
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3609020388153534e-05,
+      "loss": 0.5129,
+      "step": 12154
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3608443636992821e-05,
+      "loss": 0.6028,
+      "step": 12155
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3607866885832109e-05,
+      "loss": 0.4806,
+      "step": 12156
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3607290134671398e-05,
+      "loss": 0.4704,
+      "step": 12157
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3606713383510686e-05,
+      "loss": 0.4596,
+      "step": 12158
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3606136632349974e-05,
+      "loss": 0.5381,
+      "step": 12159
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3605559881189263e-05,
+      "loss": 0.5771,
+      "step": 12160
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.360498313002855e-05,
+      "loss": 0.4298,
+      "step": 12161
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3604406378867838e-05,
+      "loss": 0.6079,
+      "step": 12162
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3603829627707126e-05,
+      "loss": 0.4157,
+      "step": 12163
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3603252876546415e-05,
+      "loss": 0.4803,
+      "step": 12164
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3602676125385703e-05,
+      "loss": 0.4645,
+      "step": 12165
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.360209937422499e-05,
+      "loss": 0.6001,
+      "step": 12166
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.360152262306428e-05,
+      "loss": 0.5141,
+      "step": 12167
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3600945871903568e-05,
+      "loss": 0.6335,
+      "step": 12168
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3600369120742856e-05,
+      "loss": 0.4522,
+      "step": 12169
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3599792369582145e-05,
+      "loss": 0.5693,
+      "step": 12170
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3599215618421433e-05,
+      "loss": 0.4186,
+      "step": 12171
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.359863886726072e-05,
+      "loss": 0.5337,
+      "step": 12172
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.359806211610001e-05,
+      "loss": 0.4112,
+      "step": 12173
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3597485364939297e-05,
+      "loss": 0.5543,
+      "step": 12174
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3596908613778585e-05,
+      "loss": 0.491,
+      "step": 12175
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3596331862617876e-05,
+      "loss": 0.5253,
+      "step": 12176
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3595755111457162e-05,
+      "loss": 0.5197,
+      "step": 12177
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.359517836029645e-05,
+      "loss": 0.3544,
+      "step": 12178
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3594601609135741e-05,
+      "loss": 0.4609,
+      "step": 12179
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3594024857975029e-05,
+      "loss": 0.4811,
+      "step": 12180
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3593448106814315e-05,
+      "loss": 0.5183,
+      "step": 12181
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3592871355653606e-05,
+      "loss": 0.4595,
+      "step": 12182
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3592294604492893e-05,
+      "loss": 0.5951,
+      "step": 12183
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3591717853332181e-05,
+      "loss": 0.6011,
+      "step": 12184
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.359114110217147e-05,
+      "loss": 0.5421,
+      "step": 12185
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3590564351010758e-05,
+      "loss": 0.4882,
+      "step": 12186
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3589987599850046e-05,
+      "loss": 0.5304,
+      "step": 12187
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3589410848689335e-05,
+      "loss": 0.541,
+      "step": 12188
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3588834097528623e-05,
+      "loss": 0.5429,
+      "step": 12189
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.358825734636791e-05,
+      "loss": 0.5866,
+      "step": 12190
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.35876805952072e-05,
+      "loss": 0.5475,
+      "step": 12191
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3587103844046488e-05,
+      "loss": 0.4099,
+      "step": 12192
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3586527092885775e-05,
+      "loss": 0.5212,
+      "step": 12193
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3585950341725065e-05,
+      "loss": 0.5211,
+      "step": 12194
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3585373590564352e-05,
+      "loss": 0.4407,
+      "step": 12195
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.358479683940364e-05,
+      "loss": 0.4457,
+      "step": 12196
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3584220088242928e-05,
+      "loss": 0.6665,
+      "step": 12197
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3583643337082217e-05,
+      "loss": 0.6055,
+      "step": 12198
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3583066585921505e-05,
+      "loss": 0.534,
+      "step": 12199
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3582489834760792e-05,
+      "loss": 0.6367,
+      "step": 12200
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3581913083600082e-05,
+      "loss": 0.4829,
+      "step": 12201
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.358133633243937e-05,
+      "loss": 0.6918,
+      "step": 12202
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3580759581278657e-05,
+      "loss": 0.5456,
+      "step": 12203
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3580182830117947e-05,
+      "loss": 0.5526,
+      "step": 12204
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3579606078957234e-05,
+      "loss": 0.652,
+      "step": 12205
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3579029327796522e-05,
+      "loss": 0.4974,
+      "step": 12206
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3578452576635813e-05,
+      "loss": 0.5615,
+      "step": 12207
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3577875825475099e-05,
+      "loss": 0.5604,
+      "step": 12208
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3577299074314387e-05,
+      "loss": 0.5668,
+      "step": 12209
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3576722323153678e-05,
+      "loss": 0.5231,
+      "step": 12210
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3576145571992965e-05,
+      "loss": 0.4319,
+      "step": 12211
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3575568820832251e-05,
+      "loss": 0.4604,
+      "step": 12212
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3574992069671543e-05,
+      "loss": 0.436,
+      "step": 12213
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.357441531851083e-05,
+      "loss": 0.5341,
+      "step": 12214
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3573838567350118e-05,
+      "loss": 0.578,
+      "step": 12215
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3573261816189407e-05,
+      "loss": 0.5523,
+      "step": 12216
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3572685065028695e-05,
+      "loss": 0.4811,
+      "step": 12217
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3572108313867983e-05,
+      "loss": 0.4437,
+      "step": 12218
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3571531562707272e-05,
+      "loss": 0.5751,
+      "step": 12219
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.357095481154656e-05,
+      "loss": 0.5752,
+      "step": 12220
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3570378060385847e-05,
+      "loss": 0.481,
+      "step": 12221
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3569801309225137e-05,
+      "loss": 0.5869,
+      "step": 12222
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3569224558064424e-05,
+      "loss": 0.4489,
+      "step": 12223
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3568647806903712e-05,
+      "loss": 0.3708,
+      "step": 12224
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3568071055743002e-05,
+      "loss": 0.5573,
+      "step": 12225
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.356749430458229e-05,
+      "loss": 0.5524,
+      "step": 12226
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3566917553421577e-05,
+      "loss": 0.543,
+      "step": 12227
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3566340802260866e-05,
+      "loss": 0.5724,
+      "step": 12228
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3565764051100154e-05,
+      "loss": 0.4785,
+      "step": 12229
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3565187299939442e-05,
+      "loss": 0.5638,
+      "step": 12230
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3564610548778731e-05,
+      "loss": 0.4948,
+      "step": 12231
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3564033797618019e-05,
+      "loss": 0.6639,
+      "step": 12232
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3563457046457306e-05,
+      "loss": 0.6294,
+      "step": 12233
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3562880295296594e-05,
+      "loss": 0.4649,
+      "step": 12234
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3562303544135883e-05,
+      "loss": 0.4207,
+      "step": 12235
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3561726792975171e-05,
+      "loss": 0.6003,
+      "step": 12236
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3561150041814459e-05,
+      "loss": 0.5043,
+      "step": 12237
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.356057329065375e-05,
+      "loss": 0.602,
+      "step": 12238
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3559996539493036e-05,
+      "loss": 0.5141,
+      "step": 12239
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3559419788332324e-05,
+      "loss": 0.5001,
+      "step": 12240
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3558843037171615e-05,
+      "loss": 0.4258,
+      "step": 12241
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3558266286010902e-05,
+      "loss": 0.4662,
+      "step": 12242
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.3557689534850188e-05,
+      "loss": 0.4258,
+      "step": 12243
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.355711278368948e-05,
+      "loss": 0.5226,
+      "step": 12244
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3556536032528767e-05,
+      "loss": 0.5183,
+      "step": 12245
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3555959281368055e-05,
+      "loss": 0.5001,
+      "step": 12246
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3555382530207344e-05,
+      "loss": 0.5533,
+      "step": 12247
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3554805779046632e-05,
+      "loss": 0.4985,
+      "step": 12248
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.355422902788592e-05,
+      "loss": 0.5362,
+      "step": 12249
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3553652276725209e-05,
+      "loss": 0.4209,
+      "step": 12250
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3553075525564497e-05,
+      "loss": 0.6236,
+      "step": 12251
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3552498774403784e-05,
+      "loss": 0.5042,
+      "step": 12252
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3551922023243074e-05,
+      "loss": 0.5371,
+      "step": 12253
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3551345272082361e-05,
+      "loss": 0.5088,
+      "step": 12254
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3550768520921649e-05,
+      "loss": 0.5642,
+      "step": 12255
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3550191769760938e-05,
+      "loss": 0.5864,
+      "step": 12256
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3549615018600226e-05,
+      "loss": 0.4856,
+      "step": 12257
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3549038267439514e-05,
+      "loss": 0.5095,
+      "step": 12258
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3548461516278803e-05,
+      "loss": 0.4711,
+      "step": 12259
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.354788476511809e-05,
+      "loss": 0.5329,
+      "step": 12260
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3547308013957379e-05,
+      "loss": 0.4179,
+      "step": 12261
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3546731262796668e-05,
+      "loss": 0.5909,
+      "step": 12262
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3546154511635956e-05,
+      "loss": 0.5564,
+      "step": 12263
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3545577760475243e-05,
+      "loss": 0.4501,
+      "step": 12264
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3545001009314533e-05,
+      "loss": 0.5406,
+      "step": 12265
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.354442425815382e-05,
+      "loss": 0.5393,
+      "step": 12266
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3543847506993108e-05,
+      "loss": 0.6255,
+      "step": 12267
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3543270755832396e-05,
+      "loss": 0.4294,
+      "step": 12268
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3542694004671687e-05,
+      "loss": 0.461,
+      "step": 12269
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3542117253510973e-05,
+      "loss": 0.5303,
+      "step": 12270
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.354154050235026e-05,
+      "loss": 0.4757,
+      "step": 12271
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3540963751189552e-05,
+      "loss": 0.4264,
+      "step": 12272
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.354038700002884e-05,
+      "loss": 0.5261,
+      "step": 12273
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3539810248868125e-05,
+      "loss": 0.4229,
+      "step": 12274
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3539233497707416e-05,
+      "loss": 0.5467,
+      "step": 12275
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3538656746546704e-05,
+      "loss": 0.4382,
+      "step": 12276
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3538079995385992e-05,
+      "loss": 0.4954,
+      "step": 12277
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3537503244225281e-05,
+      "loss": 0.4319,
+      "step": 12278
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3536926493064569e-05,
+      "loss": 0.5661,
+      "step": 12279
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3536349741903856e-05,
+      "loss": 0.5445,
+      "step": 12280
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3535772990743146e-05,
+      "loss": 0.7003,
+      "step": 12281
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3535196239582433e-05,
+      "loss": 0.574,
+      "step": 12282
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3534619488421721e-05,
+      "loss": 0.532,
+      "step": 12283
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.353404273726101e-05,
+      "loss": 0.5799,
+      "step": 12284
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3533465986100298e-05,
+      "loss": 0.4551,
+      "step": 12285
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3532889234939586e-05,
+      "loss": 0.4832,
+      "step": 12286
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3532312483778875e-05,
+      "loss": 0.5861,
+      "step": 12287
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3531735732618163e-05,
+      "loss": 0.5427,
+      "step": 12288
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.353115898145745e-05,
+      "loss": 0.6027,
+      "step": 12289
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.353058223029674e-05,
+      "loss": 0.5798,
+      "step": 12290
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3530005479136028e-05,
+      "loss": 0.3993,
+      "step": 12291
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3529428727975315e-05,
+      "loss": 0.4962,
+      "step": 12292
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3528851976814605e-05,
+      "loss": 0.6413,
+      "step": 12293
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3528275225653892e-05,
+      "loss": 0.5717,
+      "step": 12294
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.352769847449318e-05,
+      "loss": 0.6916,
+      "step": 12295
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.352712172333247e-05,
+      "loss": 0.4432,
+      "step": 12296
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3526544972171757e-05,
+      "loss": 0.5711,
+      "step": 12297
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3525968221011045e-05,
+      "loss": 0.5959,
+      "step": 12298
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3525391469850336e-05,
+      "loss": 0.5241,
+      "step": 12299
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3524814718689624e-05,
+      "loss": 0.6131,
+      "step": 12300
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.352423796752891e-05,
+      "loss": 0.4536,
+      "step": 12301
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3523661216368197e-05,
+      "loss": 0.5411,
+      "step": 12302
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3523084465207488e-05,
+      "loss": 0.4873,
+      "step": 12303
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3522507714046776e-05,
+      "loss": 0.4454,
+      "step": 12304
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3521930962886062e-05,
+      "loss": 0.6075,
+      "step": 12305
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3521354211725353e-05,
+      "loss": 0.5171,
+      "step": 12306
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.352077746056464e-05,
+      "loss": 0.4779,
+      "step": 12307
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3520200709403929e-05,
+      "loss": 0.4926,
+      "step": 12308
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3519623958243218e-05,
+      "loss": 0.3927,
+      "step": 12309
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3519047207082506e-05,
+      "loss": 0.6,
+      "step": 12310
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3518470455921793e-05,
+      "loss": 0.5749,
+      "step": 12311
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3517893704761083e-05,
+      "loss": 0.4378,
+      "step": 12312
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.351731695360037e-05,
+      "loss": 0.5213,
+      "step": 12313
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3516740202439658e-05,
+      "loss": 0.5188,
+      "step": 12314
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3516163451278947e-05,
+      "loss": 0.5896,
+      "step": 12315
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3515586700118235e-05,
+      "loss": 0.4682,
+      "step": 12316
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3515009948957523e-05,
+      "loss": 0.4368,
+      "step": 12317
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3514433197796812e-05,
+      "loss": 0.4083,
+      "step": 12318
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.35138564466361e-05,
+      "loss": 0.446,
+      "step": 12319
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3513279695475388e-05,
+      "loss": 0.5941,
+      "step": 12320
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3512702944314677e-05,
+      "loss": 0.4176,
+      "step": 12321
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3512126193153965e-05,
+      "loss": 0.4333,
+      "step": 12322
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3511549441993252e-05,
+      "loss": 0.5628,
+      "step": 12323
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3510972690832542e-05,
+      "loss": 0.6458,
+      "step": 12324
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.351039593967183e-05,
+      "loss": 0.3563,
+      "step": 12325
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3509819188511117e-05,
+      "loss": 0.5577,
+      "step": 12326
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3509242437350406e-05,
+      "loss": 0.4536,
+      "step": 12327
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3508665686189694e-05,
+      "loss": 0.4985,
+      "step": 12328
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3508088935028982e-05,
+      "loss": 0.5007,
+      "step": 12329
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3507512183868273e-05,
+      "loss": 0.3775,
+      "step": 12330
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.350693543270756e-05,
+      "loss": 0.5039,
+      "step": 12331
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3506358681546846e-05,
+      "loss": 0.5984,
+      "step": 12332
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3505781930386138e-05,
+      "loss": 0.3907,
+      "step": 12333
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3505205179225425e-05,
+      "loss": 0.5194,
+      "step": 12334
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3504628428064713e-05,
+      "loss": 0.496,
+      "step": 12335
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3504051676904002e-05,
+      "loss": 0.5344,
+      "step": 12336
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.350347492574329e-05,
+      "loss": 0.4977,
+      "step": 12337
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3502898174582578e-05,
+      "loss": 0.3971,
+      "step": 12338
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3502321423421865e-05,
+      "loss": 0.5325,
+      "step": 12339
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3501744672261155e-05,
+      "loss": 0.5737,
+      "step": 12340
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3501167921100442e-05,
+      "loss": 0.5386,
+      "step": 12341
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.350059116993973e-05,
+      "loss": 0.5521,
+      "step": 12342
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.350001441877902e-05,
+      "loss": 0.5579,
+      "step": 12343
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3499437667618307e-05,
+      "loss": 0.4998,
+      "step": 12344
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3498860916457595e-05,
+      "loss": 0.6539,
+      "step": 12345
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3498284165296884e-05,
+      "loss": 0.5812,
+      "step": 12346
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3497707414136172e-05,
+      "loss": 0.4245,
+      "step": 12347
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.349713066297546e-05,
+      "loss": 0.6286,
+      "step": 12348
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3496553911814749e-05,
+      "loss": 0.5939,
+      "step": 12349
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3495977160654037e-05,
+      "loss": 0.4617,
+      "step": 12350
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3495400409493324e-05,
+      "loss": 0.495,
+      "step": 12351
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3494823658332614e-05,
+      "loss": 0.6246,
+      "step": 12352
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3494246907171901e-05,
+      "loss": 0.5689,
+      "step": 12353
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3493670156011189e-05,
+      "loss": 0.4881,
+      "step": 12354
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3493093404850479e-05,
+      "loss": 0.642,
+      "step": 12355
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3492516653689766e-05,
+      "loss": 0.5438,
+      "step": 12356
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3491939902529054e-05,
+      "loss": 0.4553,
+      "step": 12357
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3491363151368343e-05,
+      "loss": 0.5176,
+      "step": 12358
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3490786400207631e-05,
+      "loss": 0.494,
+      "step": 12359
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3490209649046919e-05,
+      "loss": 0.5118,
+      "step": 12360
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.348963289788621e-05,
+      "loss": 0.4626,
+      "step": 12361
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3489056146725497e-05,
+      "loss": 0.5904,
+      "step": 12362
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3488479395564783e-05,
+      "loss": 0.4493,
+      "step": 12363
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3487902644404074e-05,
+      "loss": 0.4736,
+      "step": 12364
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3487325893243362e-05,
+      "loss": 0.539,
+      "step": 12365
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.348674914208265e-05,
+      "loss": 0.4451,
+      "step": 12366
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.348617239092194e-05,
+      "loss": 0.523,
+      "step": 12367
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3485595639761227e-05,
+      "loss": 0.4699,
+      "step": 12368
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3485018888600515e-05,
+      "loss": 0.6142,
+      "step": 12369
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3484442137439804e-05,
+      "loss": 0.417,
+      "step": 12370
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3483865386279092e-05,
+      "loss": 0.4344,
+      "step": 12371
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.348328863511838e-05,
+      "loss": 0.6048,
+      "step": 12372
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3482711883957667e-05,
+      "loss": 0.3649,
+      "step": 12373
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3482135132796956e-05,
+      "loss": 0.5471,
+      "step": 12374
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3481558381636244e-05,
+      "loss": 0.3996,
+      "step": 12375
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3480981630475532e-05,
+      "loss": 0.4598,
+      "step": 12376
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3480404879314821e-05,
+      "loss": 0.636,
+      "step": 12377
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3479828128154109e-05,
+      "loss": 0.4715,
+      "step": 12378
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3479251376993396e-05,
+      "loss": 0.4988,
+      "step": 12379
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3478674625832686e-05,
+      "loss": 0.5415,
+      "step": 12380
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3478097874671974e-05,
+      "loss": 0.4612,
+      "step": 12381
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3477521123511261e-05,
+      "loss": 0.5184,
+      "step": 12382
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.347694437235055e-05,
+      "loss": 0.5459,
+      "step": 12383
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3476367621189838e-05,
+      "loss": 0.6041,
+      "step": 12384
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3475790870029126e-05,
+      "loss": 0.4891,
+      "step": 12385
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3475214118868415e-05,
+      "loss": 0.4727,
+      "step": 12386
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3474637367707703e-05,
+      "loss": 0.3565,
+      "step": 12387
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.347406061654699e-05,
+      "loss": 0.4693,
+      "step": 12388
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.347348386538628e-05,
+      "loss": 0.5152,
+      "step": 12389
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3472907114225568e-05,
+      "loss": 0.4952,
+      "step": 12390
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3472330363064855e-05,
+      "loss": 0.4203,
+      "step": 12391
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3471753611904147e-05,
+      "loss": 0.5539,
+      "step": 12392
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3471176860743434e-05,
+      "loss": 0.4933,
+      "step": 12393
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.347060010958272e-05,
+      "loss": 0.4988,
+      "step": 12394
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3470023358422011e-05,
+      "loss": 0.4334,
+      "step": 12395
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3469446607261299e-05,
+      "loss": 0.5019,
+      "step": 12396
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3468869856100587e-05,
+      "loss": 0.3859,
+      "step": 12397
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3468293104939876e-05,
+      "loss": 0.5658,
+      "step": 12398
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3467716353779164e-05,
+      "loss": 0.5697,
+      "step": 12399
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3467139602618451e-05,
+      "loss": 0.4989,
+      "step": 12400
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.346656285145774e-05,
+      "loss": 0.5047,
+      "step": 12401
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3465986100297029e-05,
+      "loss": 0.4589,
+      "step": 12402
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3465409349136316e-05,
+      "loss": 0.4926,
+      "step": 12403
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3464832597975606e-05,
+      "loss": 0.522,
+      "step": 12404
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3464255846814893e-05,
+      "loss": 0.4769,
+      "step": 12405
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3463679095654181e-05,
+      "loss": 0.4611,
+      "step": 12406
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3463102344493469e-05,
+      "loss": 0.4781,
+      "step": 12407
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3462525593332758e-05,
+      "loss": 0.5802,
+      "step": 12408
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3461948842172046e-05,
+      "loss": 0.5285,
+      "step": 12409
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3461372091011333e-05,
+      "loss": 0.4673,
+      "step": 12410
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3460795339850623e-05,
+      "loss": 0.4578,
+      "step": 12411
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.346021858868991e-05,
+      "loss": 0.3592,
+      "step": 12412
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3459641837529198e-05,
+      "loss": 0.3692,
+      "step": 12413
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3459065086368487e-05,
+      "loss": 0.531,
+      "step": 12414
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3458488335207775e-05,
+      "loss": 0.412,
+      "step": 12415
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3457911584047063e-05,
+      "loss": 0.4388,
+      "step": 12416
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3457334832886352e-05,
+      "loss": 0.4779,
+      "step": 12417
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.345675808172564e-05,
+      "loss": 0.5636,
+      "step": 12418
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3456181330564928e-05,
+      "loss": 0.3721,
+      "step": 12419
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3455604579404217e-05,
+      "loss": 0.517,
+      "step": 12420
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3455027828243505e-05,
+      "loss": 0.4615,
+      "step": 12421
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3454451077082792e-05,
+      "loss": 0.4523,
+      "step": 12422
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 1.3453874325922083e-05,
+      "loss": 0.5014,
+      "step": 12423
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3453297574761371e-05,
+      "loss": 0.538,
+      "step": 12424
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3452720823600657e-05,
+      "loss": 0.5428,
+      "step": 12425
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3452144072439948e-05,
+      "loss": 0.5858,
+      "step": 12426
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3451567321279236e-05,
+      "loss": 0.5029,
+      "step": 12427
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3450990570118524e-05,
+      "loss": 0.5865,
+      "step": 12428
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3450413818957813e-05,
+      "loss": 0.6373,
+      "step": 12429
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.34498370677971e-05,
+      "loss": 0.6125,
+      "step": 12430
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3449260316636388e-05,
+      "loss": 0.5947,
+      "step": 12431
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3448683565475678e-05,
+      "loss": 0.4613,
+      "step": 12432
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3448106814314965e-05,
+      "loss": 0.6753,
+      "step": 12433
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3447530063154253e-05,
+      "loss": 0.3974,
+      "step": 12434
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3446953311993542e-05,
+      "loss": 0.4592,
+      "step": 12435
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.344637656083283e-05,
+      "loss": 0.42,
+      "step": 12436
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3445799809672118e-05,
+      "loss": 0.5191,
+      "step": 12437
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3445223058511407e-05,
+      "loss": 0.5784,
+      "step": 12438
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3444646307350695e-05,
+      "loss": 0.5313,
+      "step": 12439
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3444069556189983e-05,
+      "loss": 0.5889,
+      "step": 12440
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.344349280502927e-05,
+      "loss": 0.4988,
+      "step": 12441
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.344291605386856e-05,
+      "loss": 0.5687,
+      "step": 12442
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3442339302707847e-05,
+      "loss": 0.3697,
+      "step": 12443
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3441762551547135e-05,
+      "loss": 0.4905,
+      "step": 12444
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3441185800386424e-05,
+      "loss": 0.4696,
+      "step": 12445
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3440609049225712e-05,
+      "loss": 0.5022,
+      "step": 12446
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3440032298065e-05,
+      "loss": 0.4859,
+      "step": 12447
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3439455546904289e-05,
+      "loss": 0.3937,
+      "step": 12448
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3438878795743577e-05,
+      "loss": 0.4993,
+      "step": 12449
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3438302044582864e-05,
+      "loss": 0.6766,
+      "step": 12450
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3437725293422154e-05,
+      "loss": 0.4434,
+      "step": 12451
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3437148542261442e-05,
+      "loss": 0.4322,
+      "step": 12452
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.343657179110073e-05,
+      "loss": 0.5495,
+      "step": 12453
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.343599503994002e-05,
+      "loss": 0.6208,
+      "step": 12454
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3435418288779308e-05,
+      "loss": 0.5491,
+      "step": 12455
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3434841537618594e-05,
+      "loss": 0.5027,
+      "step": 12456
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3434264786457885e-05,
+      "loss": 0.5896,
+      "step": 12457
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3433688035297173e-05,
+      "loss": 0.478,
+      "step": 12458
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.343311128413646e-05,
+      "loss": 0.5548,
+      "step": 12459
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.343253453297575e-05,
+      "loss": 0.5074,
+      "step": 12460
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3431957781815037e-05,
+      "loss": 0.5502,
+      "step": 12461
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3431381030654325e-05,
+      "loss": 0.3843,
+      "step": 12462
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3430804279493615e-05,
+      "loss": 0.4605,
+      "step": 12463
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3430227528332902e-05,
+      "loss": 0.5195,
+      "step": 12464
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.342965077717219e-05,
+      "loss": 0.4664,
+      "step": 12465
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.342907402601148e-05,
+      "loss": 0.5298,
+      "step": 12466
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3428497274850767e-05,
+      "loss": 0.4745,
+      "step": 12467
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3427920523690055e-05,
+      "loss": 0.545,
+      "step": 12468
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3427343772529344e-05,
+      "loss": 0.6295,
+      "step": 12469
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3426767021368632e-05,
+      "loss": 0.5695,
+      "step": 12470
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.342619027020792e-05,
+      "loss": 0.6398,
+      "step": 12471
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3425613519047209e-05,
+      "loss": 0.5772,
+      "step": 12472
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3425036767886496e-05,
+      "loss": 0.3931,
+      "step": 12473
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3424460016725784e-05,
+      "loss": 0.4881,
+      "step": 12474
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3423883265565074e-05,
+      "loss": 0.4593,
+      "step": 12475
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3423306514404361e-05,
+      "loss": 0.5814,
+      "step": 12476
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3422729763243649e-05,
+      "loss": 0.465,
+      "step": 12477
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3422153012082937e-05,
+      "loss": 0.4837,
+      "step": 12478
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3421576260922226e-05,
+      "loss": 0.5584,
+      "step": 12479
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3420999509761514e-05,
+      "loss": 0.5477,
+      "step": 12480
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3420422758600801e-05,
+      "loss": 0.6034,
+      "step": 12481
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.341984600744009e-05,
+      "loss": 0.574,
+      "step": 12482
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3419269256279378e-05,
+      "loss": 0.5454,
+      "step": 12483
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3418692505118666e-05,
+      "loss": 0.5497,
+      "step": 12484
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3418115753957957e-05,
+      "loss": 0.5522,
+      "step": 12485
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3417539002797245e-05,
+      "loss": 0.6195,
+      "step": 12486
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3416962251636531e-05,
+      "loss": 0.4644,
+      "step": 12487
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3416385500475822e-05,
+      "loss": 0.5346,
+      "step": 12488
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.341580874931511e-05,
+      "loss": 0.5409,
+      "step": 12489
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3415231998154397e-05,
+      "loss": 0.5713,
+      "step": 12490
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3414655246993687e-05,
+      "loss": 0.4803,
+      "step": 12491
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3414078495832974e-05,
+      "loss": 0.4186,
+      "step": 12492
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3413501744672262e-05,
+      "loss": 0.4685,
+      "step": 12493
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3412924993511551e-05,
+      "loss": 0.4976,
+      "step": 12494
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3412348242350839e-05,
+      "loss": 0.5471,
+      "step": 12495
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3411771491190127e-05,
+      "loss": 0.6764,
+      "step": 12496
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3411194740029416e-05,
+      "loss": 0.5196,
+      "step": 12497
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3410617988868704e-05,
+      "loss": 0.4689,
+      "step": 12498
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3410041237707992e-05,
+      "loss": 0.5168,
+      "step": 12499
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3409464486547281e-05,
+      "loss": 0.3861,
+      "step": 12500
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3408887735386569e-05,
+      "loss": 0.4916,
+      "step": 12501
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3408310984225856e-05,
+      "loss": 0.6164,
+      "step": 12502
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3407734233065146e-05,
+      "loss": 0.4202,
+      "step": 12503
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3407157481904433e-05,
+      "loss": 0.5942,
+      "step": 12504
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3406580730743721e-05,
+      "loss": 0.5344,
+      "step": 12505
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.340600397958301e-05,
+      "loss": 0.4975,
+      "step": 12506
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3405427228422298e-05,
+      "loss": 0.5035,
+      "step": 12507
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3404850477261586e-05,
+      "loss": 0.6637,
+      "step": 12508
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3404273726100875e-05,
+      "loss": 0.4163,
+      "step": 12509
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3403696974940163e-05,
+      "loss": 0.4771,
+      "step": 12510
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.340312022377945e-05,
+      "loss": 0.5711,
+      "step": 12511
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3402543472618738e-05,
+      "loss": 0.6258,
+      "step": 12512
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3401966721458028e-05,
+      "loss": 0.617,
+      "step": 12513
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3401389970297315e-05,
+      "loss": 0.5806,
+      "step": 12514
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3400813219136603e-05,
+      "loss": 0.3934,
+      "step": 12515
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3400236467975894e-05,
+      "loss": 0.4712,
+      "step": 12516
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3399659716815182e-05,
+      "loss": 0.6019,
+      "step": 12517
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3399082965654468e-05,
+      "loss": 0.4772,
+      "step": 12518
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3398506214493759e-05,
+      "loss": 0.6092,
+      "step": 12519
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3397929463333046e-05,
+      "loss": 0.4384,
+      "step": 12520
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3397352712172334e-05,
+      "loss": 0.4811,
+      "step": 12521
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3396775961011624e-05,
+      "loss": 0.5358,
+      "step": 12522
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3396199209850911e-05,
+      "loss": 0.6307,
+      "step": 12523
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3395622458690199e-05,
+      "loss": 0.4203,
+      "step": 12524
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3395045707529488e-05,
+      "loss": 0.4807,
+      "step": 12525
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3394468956368776e-05,
+      "loss": 0.4485,
+      "step": 12526
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3393892205208064e-05,
+      "loss": 0.482,
+      "step": 12527
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3393315454047353e-05,
+      "loss": 0.6529,
+      "step": 12528
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.339273870288664e-05,
+      "loss": 0.5221,
+      "step": 12529
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3392161951725928e-05,
+      "loss": 0.4616,
+      "step": 12530
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3391585200565218e-05,
+      "loss": 0.5167,
+      "step": 12531
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3391008449404505e-05,
+      "loss": 0.519,
+      "step": 12532
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3390431698243793e-05,
+      "loss": 0.4403,
+      "step": 12533
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3389854947083083e-05,
+      "loss": 0.562,
+      "step": 12534
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.338927819592237e-05,
+      "loss": 0.5065,
+      "step": 12535
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3388701444761658e-05,
+      "loss": 0.4947,
+      "step": 12536
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3388124693600947e-05,
+      "loss": 0.5414,
+      "step": 12537
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3387547942440235e-05,
+      "loss": 0.5453,
+      "step": 12538
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3386971191279523e-05,
+      "loss": 0.4681,
+      "step": 12539
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3386394440118812e-05,
+      "loss": 0.5276,
+      "step": 12540
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.33858176889581e-05,
+      "loss": 0.3873,
+      "step": 12541
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3385240937797387e-05,
+      "loss": 0.508,
+      "step": 12542
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3384664186636678e-05,
+      "loss": 0.477,
+      "step": 12543
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3384087435475966e-05,
+      "loss": 0.685,
+      "step": 12544
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3383510684315252e-05,
+      "loss": 0.496,
+      "step": 12545
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.338293393315454e-05,
+      "loss": 0.5743,
+      "step": 12546
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3382357181993831e-05,
+      "loss": 0.4546,
+      "step": 12547
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3381780430833119e-05,
+      "loss": 0.5007,
+      "step": 12548
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3381203679672405e-05,
+      "loss": 0.4242,
+      "step": 12549
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3380626928511696e-05,
+      "loss": 0.5376,
+      "step": 12550
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3380050177350983e-05,
+      "loss": 0.4971,
+      "step": 12551
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3379473426190271e-05,
+      "loss": 0.5518,
+      "step": 12552
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.337889667502956e-05,
+      "loss": 0.4455,
+      "step": 12553
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3378319923868848e-05,
+      "loss": 0.3914,
+      "step": 12554
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3377743172708136e-05,
+      "loss": 0.5496,
+      "step": 12555
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3377166421547425e-05,
+      "loss": 0.5026,
+      "step": 12556
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3376589670386713e-05,
+      "loss": 0.5325,
+      "step": 12557
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3376012919226e-05,
+      "loss": 0.5275,
+      "step": 12558
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.337543616806529e-05,
+      "loss": 0.5648,
+      "step": 12559
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3374859416904578e-05,
+      "loss": 0.4846,
+      "step": 12560
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3374282665743865e-05,
+      "loss": 0.7404,
+      "step": 12561
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3373705914583155e-05,
+      "loss": 0.491,
+      "step": 12562
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3373129163422442e-05,
+      "loss": 0.5637,
+      "step": 12563
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.337255241226173e-05,
+      "loss": 0.5999,
+      "step": 12564
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.337197566110102e-05,
+      "loss": 0.4812,
+      "step": 12565
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3371398909940307e-05,
+      "loss": 0.3853,
+      "step": 12566
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3370822158779595e-05,
+      "loss": 0.5049,
+      "step": 12567
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3370245407618884e-05,
+      "loss": 0.4782,
+      "step": 12568
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3369668656458172e-05,
+      "loss": 0.5721,
+      "step": 12569
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.336909190529746e-05,
+      "loss": 0.5435,
+      "step": 12570
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3368515154136749e-05,
+      "loss": 0.4286,
+      "step": 12571
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3367938402976037e-05,
+      "loss": 0.5353,
+      "step": 12572
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3367361651815324e-05,
+      "loss": 0.4905,
+      "step": 12573
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3366784900654615e-05,
+      "loss": 0.4552,
+      "step": 12574
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3366208149493903e-05,
+      "loss": 0.4981,
+      "step": 12575
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3365631398333189e-05,
+      "loss": 0.775,
+      "step": 12576
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.336505464717248e-05,
+      "loss": 0.5736,
+      "step": 12577
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3364477896011768e-05,
+      "loss": 0.587,
+      "step": 12578
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3363901144851055e-05,
+      "loss": 0.6449,
+      "step": 12579
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3363324393690341e-05,
+      "loss": 0.5547,
+      "step": 12580
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3362747642529633e-05,
+      "loss": 0.4398,
+      "step": 12581
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.336217089136892e-05,
+      "loss": 0.5404,
+      "step": 12582
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3361594140208208e-05,
+      "loss": 0.5854,
+      "step": 12583
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3361017389047497e-05,
+      "loss": 0.3364,
+      "step": 12584
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3360440637886785e-05,
+      "loss": 0.5314,
+      "step": 12585
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3359863886726073e-05,
+      "loss": 0.4973,
+      "step": 12586
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3359287135565362e-05,
+      "loss": 0.4241,
+      "step": 12587
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.335871038440465e-05,
+      "loss": 0.4291,
+      "step": 12588
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3358133633243937e-05,
+      "loss": 0.6291,
+      "step": 12589
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3357556882083227e-05,
+      "loss": 0.5017,
+      "step": 12590
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3356980130922514e-05,
+      "loss": 0.4874,
+      "step": 12591
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3356403379761802e-05,
+      "loss": 0.5104,
+      "step": 12592
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3355826628601092e-05,
+      "loss": 0.5051,
+      "step": 12593
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.335524987744038e-05,
+      "loss": 0.4185,
+      "step": 12594
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3354673126279667e-05,
+      "loss": 0.4488,
+      "step": 12595
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3354096375118956e-05,
+      "loss": 0.2812,
+      "step": 12596
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3353519623958244e-05,
+      "loss": 0.5485,
+      "step": 12597
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3352942872797532e-05,
+      "loss": 0.5334,
+      "step": 12598
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3352366121636821e-05,
+      "loss": 0.396,
+      "step": 12599
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3351789370476109e-05,
+      "loss": 0.5412,
+      "step": 12600
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 1.3351212619315396e-05,
+      "loss": 0.4511,
+      "step": 12601
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3350635868154686e-05,
+      "loss": 0.4412,
+      "step": 12602
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3350059116993973e-05,
+      "loss": 0.5099,
+      "step": 12603
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3349482365833261e-05,
+      "loss": 0.3787,
+      "step": 12604
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3348905614672552e-05,
+      "loss": 0.4562,
+      "step": 12605
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.334832886351184e-05,
+      "loss": 0.5516,
+      "step": 12606
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3347752112351126e-05,
+      "loss": 0.4573,
+      "step": 12607
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3347175361190417e-05,
+      "loss": 0.4811,
+      "step": 12608
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3346598610029705e-05,
+      "loss": 0.5604,
+      "step": 12609
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3346021858868992e-05,
+      "loss": 0.5218,
+      "step": 12610
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3345445107708282e-05,
+      "loss": 0.5688,
+      "step": 12611
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.334486835654757e-05,
+      "loss": 0.5588,
+      "step": 12612
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3344291605386857e-05,
+      "loss": 0.5573,
+      "step": 12613
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3343714854226146e-05,
+      "loss": 0.5889,
+      "step": 12614
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3343138103065434e-05,
+      "loss": 0.5085,
+      "step": 12615
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3342561351904722e-05,
+      "loss": 0.4038,
+      "step": 12616
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.334198460074401e-05,
+      "loss": 0.5249,
+      "step": 12617
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3341407849583299e-05,
+      "loss": 0.4287,
+      "step": 12618
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3340831098422587e-05,
+      "loss": 0.5316,
+      "step": 12619
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3340254347261874e-05,
+      "loss": 0.5579,
+      "step": 12620
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3339677596101164e-05,
+      "loss": 0.4357,
+      "step": 12621
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3339100844940451e-05,
+      "loss": 0.4668,
+      "step": 12622
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3338524093779739e-05,
+      "loss": 0.4414,
+      "step": 12623
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3337947342619028e-05,
+      "loss": 0.5105,
+      "step": 12624
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3337370591458316e-05,
+      "loss": 0.4402,
+      "step": 12625
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3336793840297604e-05,
+      "loss": 0.5065,
+      "step": 12626
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3336217089136893e-05,
+      "loss": 0.6332,
+      "step": 12627
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3335640337976181e-05,
+      "loss": 0.4884,
+      "step": 12628
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3335063586815469e-05,
+      "loss": 0.5436,
+      "step": 12629
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3334486835654758e-05,
+      "loss": 0.5069,
+      "step": 12630
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3333910084494046e-05,
+      "loss": 0.5392,
+      "step": 12631
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.6261,
+      "step": 12632
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3332756582172623e-05,
+      "loss": 0.7151,
+      "step": 12633
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.333217983101191e-05,
+      "loss": 0.5067,
+      "step": 12634
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3331603079851198e-05,
+      "loss": 0.4186,
+      "step": 12635
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3331026328690489e-05,
+      "loss": 0.6478,
+      "step": 12636
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3330449577529777e-05,
+      "loss": 0.4736,
+      "step": 12637
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3329872826369063e-05,
+      "loss": 0.5017,
+      "step": 12638
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3329296075208354e-05,
+      "loss": 0.4834,
+      "step": 12639
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3328719324047642e-05,
+      "loss": 0.5864,
+      "step": 12640
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.332814257288693e-05,
+      "loss": 0.4985,
+      "step": 12641
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3327565821726219e-05,
+      "loss": 0.441,
+      "step": 12642
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3326989070565506e-05,
+      "loss": 0.5151,
+      "step": 12643
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3326412319404794e-05,
+      "loss": 0.4305,
+      "step": 12644
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3325835568244083e-05,
+      "loss": 0.5073,
+      "step": 12645
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3325258817083371e-05,
+      "loss": 0.4151,
+      "step": 12646
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3324682065922659e-05,
+      "loss": 0.5647,
+      "step": 12647
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3324105314761948e-05,
+      "loss": 0.4359,
+      "step": 12648
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3323528563601236e-05,
+      "loss": 0.4617,
+      "step": 12649
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3322951812440523e-05,
+      "loss": 0.4269,
+      "step": 12650
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3322375061279811e-05,
+      "loss": 0.5859,
+      "step": 12651
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.33217983101191e-05,
+      "loss": 0.4724,
+      "step": 12652
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3321221558958388e-05,
+      "loss": 0.4749,
+      "step": 12653
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3320644807797676e-05,
+      "loss": 0.487,
+      "step": 12654
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3320068056636965e-05,
+      "loss": 0.5531,
+      "step": 12655
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3319491305476253e-05,
+      "loss": 0.4394,
+      "step": 12656
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.331891455431554e-05,
+      "loss": 0.4551,
+      "step": 12657
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.331833780315483e-05,
+      "loss": 0.463,
+      "step": 12658
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3317761051994118e-05,
+      "loss": 0.5166,
+      "step": 12659
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3317184300833405e-05,
+      "loss": 0.5236,
+      "step": 12660
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3316607549672695e-05,
+      "loss": 0.6102,
+      "step": 12661
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3316030798511982e-05,
+      "loss": 0.485,
+      "step": 12662
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.331545404735127e-05,
+      "loss": 0.4352,
+      "step": 12663
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.331487729619056e-05,
+      "loss": 0.5927,
+      "step": 12664
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3314300545029847e-05,
+      "loss": 0.5918,
+      "step": 12665
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3313723793869135e-05,
+      "loss": 0.4258,
+      "step": 12666
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3313147042708426e-05,
+      "loss": 0.4491,
+      "step": 12667
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3312570291547714e-05,
+      "loss": 0.484,
+      "step": 12668
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3311993540387e-05,
+      "loss": 0.4867,
+      "step": 12669
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.331141678922629e-05,
+      "loss": 0.4192,
+      "step": 12670
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3310840038065578e-05,
+      "loss": 0.5932,
+      "step": 12671
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3310263286904866e-05,
+      "loss": 0.4677,
+      "step": 12672
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3309686535744155e-05,
+      "loss": 0.6073,
+      "step": 12673
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3309109784583443e-05,
+      "loss": 0.5355,
+      "step": 12674
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3308533033422731e-05,
+      "loss": 0.5996,
+      "step": 12675
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.330795628226202e-05,
+      "loss": 0.4048,
+      "step": 12676
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3307379531101308e-05,
+      "loss": 0.5353,
+      "step": 12677
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3306802779940596e-05,
+      "loss": 0.4763,
+      "step": 12678
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3306226028779885e-05,
+      "loss": 0.5068,
+      "step": 12679
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3305649277619173e-05,
+      "loss": 0.5582,
+      "step": 12680
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.330507252645846e-05,
+      "loss": 0.5812,
+      "step": 12681
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.330449577529775e-05,
+      "loss": 0.4552,
+      "step": 12682
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3303919024137037e-05,
+      "loss": 0.4612,
+      "step": 12683
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3303342272976325e-05,
+      "loss": 0.3546,
+      "step": 12684
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3302765521815613e-05,
+      "loss": 0.4843,
+      "step": 12685
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3302188770654902e-05,
+      "loss": 0.3686,
+      "step": 12686
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.330161201949419e-05,
+      "loss": 0.5687,
+      "step": 12687
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3301035268333478e-05,
+      "loss": 0.4294,
+      "step": 12688
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3300458517172767e-05,
+      "loss": 0.3925,
+      "step": 12689
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3299881766012055e-05,
+      "loss": 0.6068,
+      "step": 12690
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3299305014851342e-05,
+      "loss": 0.4937,
+      "step": 12691
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3298728263690632e-05,
+      "loss": 0.4726,
+      "step": 12692
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.329815151252992e-05,
+      "loss": 0.42,
+      "step": 12693
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3297574761369207e-05,
+      "loss": 0.5374,
+      "step": 12694
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3296998010208496e-05,
+      "loss": 0.5236,
+      "step": 12695
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3296421259047784e-05,
+      "loss": 0.5299,
+      "step": 12696
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3295844507887072e-05,
+      "loss": 0.4824,
+      "step": 12697
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3295267756726363e-05,
+      "loss": 0.547,
+      "step": 12698
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.329469100556565e-05,
+      "loss": 0.5898,
+      "step": 12699
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3294114254404937e-05,
+      "loss": 0.4943,
+      "step": 12700
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3293537503244228e-05,
+      "loss": 0.5137,
+      "step": 12701
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3292960752083515e-05,
+      "loss": 0.4959,
+      "step": 12702
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3292384000922803e-05,
+      "loss": 0.5222,
+      "step": 12703
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3291807249762092e-05,
+      "loss": 0.5347,
+      "step": 12704
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.329123049860138e-05,
+      "loss": 0.4879,
+      "step": 12705
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3290653747440668e-05,
+      "loss": 0.4736,
+      "step": 12706
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3290076996279957e-05,
+      "loss": 0.406,
+      "step": 12707
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3289500245119245e-05,
+      "loss": 0.4638,
+      "step": 12708
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3288923493958532e-05,
+      "loss": 0.4554,
+      "step": 12709
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3288346742797822e-05,
+      "loss": 0.4473,
+      "step": 12710
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.328776999163711e-05,
+      "loss": 0.5044,
+      "step": 12711
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3287193240476397e-05,
+      "loss": 0.5887,
+      "step": 12712
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3286616489315687e-05,
+      "loss": 0.612,
+      "step": 12713
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3286039738154974e-05,
+      "loss": 0.6674,
+      "step": 12714
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3285462986994262e-05,
+      "loss": 0.5459,
+      "step": 12715
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3284886235833551e-05,
+      "loss": 0.6248,
+      "step": 12716
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3284309484672839e-05,
+      "loss": 0.4663,
+      "step": 12717
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3283732733512127e-05,
+      "loss": 0.461,
+      "step": 12718
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3283155982351416e-05,
+      "loss": 0.4913,
+      "step": 12719
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3282579231190704e-05,
+      "loss": 0.6388,
+      "step": 12720
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3282002480029991e-05,
+      "loss": 0.512,
+      "step": 12721
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.328142572886928e-05,
+      "loss": 0.5699,
+      "step": 12722
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3280848977708569e-05,
+      "loss": 0.5331,
+      "step": 12723
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3280272226547856e-05,
+      "loss": 0.6234,
+      "step": 12724
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3279695475387144e-05,
+      "loss": 0.5686,
+      "step": 12725
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3279118724226433e-05,
+      "loss": 0.4932,
+      "step": 12726
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3278541973065721e-05,
+      "loss": 0.4479,
+      "step": 12727
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3277965221905009e-05,
+      "loss": 0.543,
+      "step": 12728
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.32773884707443e-05,
+      "loss": 0.4679,
+      "step": 12729
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3276811719583587e-05,
+      "loss": 0.5156,
+      "step": 12730
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3276234968422873e-05,
+      "loss": 0.4088,
+      "step": 12731
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3275658217262164e-05,
+      "loss": 0.5172,
+      "step": 12732
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3275081466101452e-05,
+      "loss": 0.5414,
+      "step": 12733
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.327450471494074e-05,
+      "loss": 0.4887,
+      "step": 12734
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.327392796378003e-05,
+      "loss": 0.5294,
+      "step": 12735
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3273351212619317e-05,
+      "loss": 0.5428,
+      "step": 12736
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3272774461458605e-05,
+      "loss": 0.4276,
+      "step": 12737
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3272197710297894e-05,
+      "loss": 0.4939,
+      "step": 12738
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3271620959137182e-05,
+      "loss": 0.5685,
+      "step": 12739
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.327104420797647e-05,
+      "loss": 0.4659,
+      "step": 12740
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3270467456815759e-05,
+      "loss": 0.5186,
+      "step": 12741
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3269890705655046e-05,
+      "loss": 0.5669,
+      "step": 12742
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3269313954494334e-05,
+      "loss": 0.5268,
+      "step": 12743
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3268737203333623e-05,
+      "loss": 0.4099,
+      "step": 12744
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3268160452172911e-05,
+      "loss": 0.4807,
+      "step": 12745
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3267583701012199e-05,
+      "loss": 0.4788,
+      "step": 12746
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3267006949851488e-05,
+      "loss": 0.6112,
+      "step": 12747
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3266430198690776e-05,
+      "loss": 0.4158,
+      "step": 12748
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3265853447530064e-05,
+      "loss": 0.6377,
+      "step": 12749
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3265276696369353e-05,
+      "loss": 0.3799,
+      "step": 12750
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.326469994520864e-05,
+      "loss": 0.545,
+      "step": 12751
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3264123194047928e-05,
+      "loss": 0.5901,
+      "step": 12752
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3263546442887218e-05,
+      "loss": 0.3951,
+      "step": 12753
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3262969691726505e-05,
+      "loss": 0.6014,
+      "step": 12754
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3262392940565793e-05,
+      "loss": 0.4905,
+      "step": 12755
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.326181618940508e-05,
+      "loss": 0.498,
+      "step": 12756
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.326123943824437e-05,
+      "loss": 0.5143,
+      "step": 12757
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3260662687083658e-05,
+      "loss": 0.4222,
+      "step": 12758
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3260085935922946e-05,
+      "loss": 0.4385,
+      "step": 12759
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3259509184762237e-05,
+      "loss": 0.4202,
+      "step": 12760
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3258932433601524e-05,
+      "loss": 0.4953,
+      "step": 12761
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.325835568244081e-05,
+      "loss": 0.4376,
+      "step": 12762
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3257778931280101e-05,
+      "loss": 0.4581,
+      "step": 12763
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3257202180119389e-05,
+      "loss": 0.4389,
+      "step": 12764
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3256625428958677e-05,
+      "loss": 0.4791,
+      "step": 12765
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3256048677797966e-05,
+      "loss": 0.5772,
+      "step": 12766
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3255471926637254e-05,
+      "loss": 0.4368,
+      "step": 12767
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3254895175476541e-05,
+      "loss": 0.5687,
+      "step": 12768
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3254318424315831e-05,
+      "loss": 0.6284,
+      "step": 12769
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3253741673155119e-05,
+      "loss": 0.3946,
+      "step": 12770
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3253164921994406e-05,
+      "loss": 0.4002,
+      "step": 12771
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3252588170833696e-05,
+      "loss": 0.5661,
+      "step": 12772
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3252011419672983e-05,
+      "loss": 0.5324,
+      "step": 12773
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3251434668512271e-05,
+      "loss": 0.4973,
+      "step": 12774
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.325085791735156e-05,
+      "loss": 0.4224,
+      "step": 12775
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3250281166190848e-05,
+      "loss": 0.502,
+      "step": 12776
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3249704415030136e-05,
+      "loss": 0.5814,
+      "step": 12777
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3249127663869425e-05,
+      "loss": 0.4648,
+      "step": 12778
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3248550912708713e-05,
+      "loss": 0.4622,
+      "step": 12779
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.3247974161548e-05,
+      "loss": 0.5741,
+      "step": 12780
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.324739741038729e-05,
+      "loss": 0.5612,
+      "step": 12781
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3246820659226578e-05,
+      "loss": 0.5344,
+      "step": 12782
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3246243908065865e-05,
+      "loss": 0.5039,
+      "step": 12783
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3245667156905155e-05,
+      "loss": 0.5351,
+      "step": 12784
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3245090405744442e-05,
+      "loss": 0.5193,
+      "step": 12785
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.324451365458373e-05,
+      "loss": 0.4142,
+      "step": 12786
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3243936903423021e-05,
+      "loss": 0.6636,
+      "step": 12787
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3243360152262307e-05,
+      "loss": 0.5174,
+      "step": 12788
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3242783401101595e-05,
+      "loss": 0.5088,
+      "step": 12789
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3242206649940882e-05,
+      "loss": 0.5022,
+      "step": 12790
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3241629898780173e-05,
+      "loss": 0.5516,
+      "step": 12791
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3241053147619461e-05,
+      "loss": 0.4222,
+      "step": 12792
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3240476396458747e-05,
+      "loss": 0.6274,
+      "step": 12793
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3239899645298038e-05,
+      "loss": 0.712,
+      "step": 12794
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3239322894137326e-05,
+      "loss": 0.5614,
+      "step": 12795
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3238746142976614e-05,
+      "loss": 0.5465,
+      "step": 12796
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3238169391815903e-05,
+      "loss": 0.4064,
+      "step": 12797
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.323759264065519e-05,
+      "loss": 0.4445,
+      "step": 12798
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3237015889494478e-05,
+      "loss": 0.4523,
+      "step": 12799
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3236439138333768e-05,
+      "loss": 0.498,
+      "step": 12800
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3235862387173055e-05,
+      "loss": 0.5083,
+      "step": 12801
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3235285636012343e-05,
+      "loss": 0.5219,
+      "step": 12802
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3234708884851632e-05,
+      "loss": 0.6472,
+      "step": 12803
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.323413213369092e-05,
+      "loss": 0.5032,
+      "step": 12804
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3233555382530208e-05,
+      "loss": 0.5441,
+      "step": 12805
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3232978631369497e-05,
+      "loss": 0.5266,
+      "step": 12806
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3232401880208785e-05,
+      "loss": 0.6011,
+      "step": 12807
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3231825129048073e-05,
+      "loss": 0.582,
+      "step": 12808
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3231248377887362e-05,
+      "loss": 0.5318,
+      "step": 12809
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.323067162672665e-05,
+      "loss": 0.6242,
+      "step": 12810
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3230094875565937e-05,
+      "loss": 0.5295,
+      "step": 12811
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3229518124405227e-05,
+      "loss": 0.5012,
+      "step": 12812
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3228941373244514e-05,
+      "loss": 0.4731,
+      "step": 12813
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3228364622083802e-05,
+      "loss": 0.575,
+      "step": 12814
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3227787870923091e-05,
+      "loss": 0.5499,
+      "step": 12815
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.322721111976238e-05,
+      "loss": 0.5266,
+      "step": 12816
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3226634368601667e-05,
+      "loss": 0.5499,
+      "step": 12817
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3226057617440958e-05,
+      "loss": 0.3732,
+      "step": 12818
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3225480866280244e-05,
+      "loss": 0.5166,
+      "step": 12819
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3224904115119532e-05,
+      "loss": 0.5292,
+      "step": 12820
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3224327363958823e-05,
+      "loss": 0.4305,
+      "step": 12821
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.322375061279811e-05,
+      "loss": 0.4888,
+      "step": 12822
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3223173861637398e-05,
+      "loss": 0.4135,
+      "step": 12823
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3222597110476684e-05,
+      "loss": 0.3983,
+      "step": 12824
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3222020359315975e-05,
+      "loss": 0.3597,
+      "step": 12825
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3221443608155263e-05,
+      "loss": 0.4434,
+      "step": 12826
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.322086685699455e-05,
+      "loss": 0.5556,
+      "step": 12827
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.322029010583384e-05,
+      "loss": 0.5248,
+      "step": 12828
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3219713354673128e-05,
+      "loss": 0.5459,
+      "step": 12829
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3219136603512415e-05,
+      "loss": 0.4612,
+      "step": 12830
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3218559852351705e-05,
+      "loss": 0.5907,
+      "step": 12831
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3217983101190992e-05,
+      "loss": 0.5961,
+      "step": 12832
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.321740635003028e-05,
+      "loss": 0.5953,
+      "step": 12833
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.321682959886957e-05,
+      "loss": 0.5314,
+      "step": 12834
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3216252847708857e-05,
+      "loss": 0.5241,
+      "step": 12835
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3215676096548145e-05,
+      "loss": 0.4047,
+      "step": 12836
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3215099345387434e-05,
+      "loss": 0.4464,
+      "step": 12837
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3214522594226722e-05,
+      "loss": 0.4713,
+      "step": 12838
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.321394584306601e-05,
+      "loss": 0.5013,
+      "step": 12839
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3213369091905299e-05,
+      "loss": 0.4646,
+      "step": 12840
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3212792340744587e-05,
+      "loss": 0.514,
+      "step": 12841
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3212215589583874e-05,
+      "loss": 0.6362,
+      "step": 12842
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3211638838423164e-05,
+      "loss": 0.6188,
+      "step": 12843
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3211062087262451e-05,
+      "loss": 0.5041,
+      "step": 12844
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3210485336101739e-05,
+      "loss": 0.5738,
+      "step": 12845
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3209908584941028e-05,
+      "loss": 0.5007,
+      "step": 12846
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3209331833780316e-05,
+      "loss": 0.518,
+      "step": 12847
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3208755082619604e-05,
+      "loss": 0.5016,
+      "step": 12848
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3208178331458895e-05,
+      "loss": 0.5997,
+      "step": 12849
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.320760158029818e-05,
+      "loss": 0.5465,
+      "step": 12850
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3207024829137468e-05,
+      "loss": 0.4989,
+      "step": 12851
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.320644807797676e-05,
+      "loss": 0.5993,
+      "step": 12852
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3205871326816047e-05,
+      "loss": 0.444,
+      "step": 12853
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3205294575655335e-05,
+      "loss": 0.5457,
+      "step": 12854
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3204717824494624e-05,
+      "loss": 0.4515,
+      "step": 12855
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3204141073333912e-05,
+      "loss": 0.5846,
+      "step": 12856
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.32035643221732e-05,
+      "loss": 0.6722,
+      "step": 12857
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3202987571012489e-05,
+      "loss": 0.5726,
+      "step": 12858
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3202410819851777e-05,
+      "loss": 0.5238,
+      "step": 12859
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3201834068691064e-05,
+      "loss": 0.5663,
+      "step": 12860
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3201257317530352e-05,
+      "loss": 0.5134,
+      "step": 12861
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3200680566369641e-05,
+      "loss": 0.4697,
+      "step": 12862
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.320010381520893e-05,
+      "loss": 0.5153,
+      "step": 12863
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3199527064048217e-05,
+      "loss": 0.6545,
+      "step": 12864
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3198950312887506e-05,
+      "loss": 0.4441,
+      "step": 12865
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3198373561726794e-05,
+      "loss": 0.4402,
+      "step": 12866
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3197796810566082e-05,
+      "loss": 0.4824,
+      "step": 12867
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3197220059405371e-05,
+      "loss": 0.4931,
+      "step": 12868
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3196643308244659e-05,
+      "loss": 0.5477,
+      "step": 12869
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3196066557083946e-05,
+      "loss": 0.373,
+      "step": 12870
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3195489805923236e-05,
+      "loss": 0.4508,
+      "step": 12871
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3194913054762523e-05,
+      "loss": 0.5206,
+      "step": 12872
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3194336303601811e-05,
+      "loss": 0.5082,
+      "step": 12873
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.31937595524411e-05,
+      "loss": 0.4584,
+      "step": 12874
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3193182801280388e-05,
+      "loss": 0.4422,
+      "step": 12875
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3192606050119676e-05,
+      "loss": 0.4248,
+      "step": 12876
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3192029298958965e-05,
+      "loss": 0.446,
+      "step": 12877
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3191452547798253e-05,
+      "loss": 0.4829,
+      "step": 12878
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.319087579663754e-05,
+      "loss": 0.5415,
+      "step": 12879
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3190299045476832e-05,
+      "loss": 0.583,
+      "step": 12880
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3189722294316118e-05,
+      "loss": 0.4524,
+      "step": 12881
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3189145543155405e-05,
+      "loss": 0.5432,
+      "step": 12882
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3188568791994696e-05,
+      "loss": 0.5029,
+      "step": 12883
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3187992040833984e-05,
+      "loss": 0.501,
+      "step": 12884
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3187415289673272e-05,
+      "loss": 0.5369,
+      "step": 12885
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3186838538512561e-05,
+      "loss": 0.4761,
+      "step": 12886
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3186261787351849e-05,
+      "loss": 0.3856,
+      "step": 12887
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3185685036191137e-05,
+      "loss": 0.616,
+      "step": 12888
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3185108285030426e-05,
+      "loss": 0.4824,
+      "step": 12889
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3184531533869714e-05,
+      "loss": 0.4618,
+      "step": 12890
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3183954782709001e-05,
+      "loss": 0.5389,
+      "step": 12891
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.318337803154829e-05,
+      "loss": 0.4692,
+      "step": 12892
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3182801280387578e-05,
+      "loss": 0.5052,
+      "step": 12893
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3182224529226866e-05,
+      "loss": 0.551,
+      "step": 12894
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3181647778066154e-05,
+      "loss": 0.4341,
+      "step": 12895
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3181071026905443e-05,
+      "loss": 0.3727,
+      "step": 12896
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.318049427574473e-05,
+      "loss": 0.4031,
+      "step": 12897
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3179917524584018e-05,
+      "loss": 0.5335,
+      "step": 12898
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3179340773423308e-05,
+      "loss": 0.5441,
+      "step": 12899
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3178764022262596e-05,
+      "loss": 0.4359,
+      "step": 12900
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3178187271101883e-05,
+      "loss": 0.5522,
+      "step": 12901
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3177610519941173e-05,
+      "loss": 0.5513,
+      "step": 12902
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.317703376878046e-05,
+      "loss": 0.5052,
+      "step": 12903
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3176457017619748e-05,
+      "loss": 0.4801,
+      "step": 12904
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3175880266459037e-05,
+      "loss": 0.4982,
+      "step": 12905
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3175303515298325e-05,
+      "loss": 0.5378,
+      "step": 12906
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3174726764137613e-05,
+      "loss": 0.4787,
+      "step": 12907
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3174150012976902e-05,
+      "loss": 0.5447,
+      "step": 12908
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.317357326181619e-05,
+      "loss": 0.5867,
+      "step": 12909
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3172996510655477e-05,
+      "loss": 0.4851,
+      "step": 12910
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3172419759494769e-05,
+      "loss": 0.4096,
+      "step": 12911
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3171843008334056e-05,
+      "loss": 0.4234,
+      "step": 12912
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3171266257173342e-05,
+      "loss": 0.5138,
+      "step": 12913
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3170689506012633e-05,
+      "loss": 0.5221,
+      "step": 12914
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3170112754851921e-05,
+      "loss": 0.58,
+      "step": 12915
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3169536003691209e-05,
+      "loss": 0.4586,
+      "step": 12916
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3168959252530498e-05,
+      "loss": 0.4613,
+      "step": 12917
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3168382501369786e-05,
+      "loss": 0.5968,
+      "step": 12918
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3167805750209073e-05,
+      "loss": 0.5725,
+      "step": 12919
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3167228999048363e-05,
+      "loss": 0.4529,
+      "step": 12920
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.316665224788765e-05,
+      "loss": 0.4574,
+      "step": 12921
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3166075496726938e-05,
+      "loss": 0.4664,
+      "step": 12922
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3165498745566228e-05,
+      "loss": 0.4862,
+      "step": 12923
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3164921994405515e-05,
+      "loss": 0.5033,
+      "step": 12924
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3164345243244803e-05,
+      "loss": 0.3802,
+      "step": 12925
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3163768492084092e-05,
+      "loss": 0.5903,
+      "step": 12926
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.316319174092338e-05,
+      "loss": 0.4087,
+      "step": 12927
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3162614989762668e-05,
+      "loss": 0.5454,
+      "step": 12928
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3162038238601955e-05,
+      "loss": 0.4844,
+      "step": 12929
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3161461487441245e-05,
+      "loss": 0.3535,
+      "step": 12930
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3160884736280532e-05,
+      "loss": 0.6681,
+      "step": 12931
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.316030798511982e-05,
+      "loss": 0.4831,
+      "step": 12932
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.315973123395911e-05,
+      "loss": 0.4601,
+      "step": 12933
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3159154482798397e-05,
+      "loss": 0.57,
+      "step": 12934
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3158577731637685e-05,
+      "loss": 0.5198,
+      "step": 12935
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3158000980476974e-05,
+      "loss": 0.4265,
+      "step": 12936
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3157424229316262e-05,
+      "loss": 0.5496,
+      "step": 12937
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.315684747815555e-05,
+      "loss": 0.5247,
+      "step": 12938
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3156270726994839e-05,
+      "loss": 0.4815,
+      "step": 12939
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3155693975834127e-05,
+      "loss": 0.5082,
+      "step": 12940
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3155117224673414e-05,
+      "loss": 0.4715,
+      "step": 12941
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3154540473512705e-05,
+      "loss": 0.4535,
+      "step": 12942
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3153963722351993e-05,
+      "loss": 0.5381,
+      "step": 12943
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3153386971191279e-05,
+      "loss": 0.5184,
+      "step": 12944
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.315281022003057e-05,
+      "loss": 0.5442,
+      "step": 12945
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3152233468869858e-05,
+      "loss": 0.5004,
+      "step": 12946
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3151656717709146e-05,
+      "loss": 0.4875,
+      "step": 12947
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3151079966548435e-05,
+      "loss": 0.4308,
+      "step": 12948
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3150503215387723e-05,
+      "loss": 0.4345,
+      "step": 12949
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.314992646422701e-05,
+      "loss": 0.5435,
+      "step": 12950
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.31493497130663e-05,
+      "loss": 0.4918,
+      "step": 12951
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3148772961905587e-05,
+      "loss": 0.5915,
+      "step": 12952
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3148196210744875e-05,
+      "loss": 0.4532,
+      "step": 12953
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3147619459584164e-05,
+      "loss": 0.5589,
+      "step": 12954
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3147042708423452e-05,
+      "loss": 0.5345,
+      "step": 12955
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.314646595726274e-05,
+      "loss": 0.4239,
+      "step": 12956
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.314588920610203e-05,
+      "loss": 0.5295,
+      "step": 12957
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3145312454941317e-05,
+      "loss": 0.4827,
+      "step": 12958
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 1.3144735703780605e-05,
+      "loss": 0.5982,
+      "step": 12959
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3144158952619894e-05,
+      "loss": 0.6506,
+      "step": 12960
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3143582201459182e-05,
+      "loss": 0.4596,
+      "step": 12961
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.314300545029847e-05,
+      "loss": 0.5172,
+      "step": 12962
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3142428699137757e-05,
+      "loss": 0.4923,
+      "step": 12963
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3141851947977046e-05,
+      "loss": 0.5518,
+      "step": 12964
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3141275196816334e-05,
+      "loss": 0.5483,
+      "step": 12965
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3140698445655622e-05,
+      "loss": 0.5367,
+      "step": 12966
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3140121694494911e-05,
+      "loss": 0.4407,
+      "step": 12967
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3139544943334199e-05,
+      "loss": 0.443,
+      "step": 12968
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3138968192173486e-05,
+      "loss": 0.606,
+      "step": 12969
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3138391441012776e-05,
+      "loss": 0.4171,
+      "step": 12970
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3137814689852064e-05,
+      "loss": 0.4638,
+      "step": 12971
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3137237938691351e-05,
+      "loss": 0.4656,
+      "step": 12972
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3136661187530642e-05,
+      "loss": 0.4733,
+      "step": 12973
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.313608443636993e-05,
+      "loss": 0.4077,
+      "step": 12974
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3135507685209216e-05,
+      "loss": 0.5563,
+      "step": 12975
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3134930934048507e-05,
+      "loss": 0.4809,
+      "step": 12976
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3134354182887795e-05,
+      "loss": 0.3863,
+      "step": 12977
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3133777431727082e-05,
+      "loss": 0.4219,
+      "step": 12978
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3133200680566372e-05,
+      "loss": 0.3898,
+      "step": 12979
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.313262392940566e-05,
+      "loss": 0.489,
+      "step": 12980
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3132047178244947e-05,
+      "loss": 0.5423,
+      "step": 12981
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3131470427084237e-05,
+      "loss": 0.5221,
+      "step": 12982
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3130893675923524e-05,
+      "loss": 0.5071,
+      "step": 12983
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3130316924762812e-05,
+      "loss": 0.4876,
+      "step": 12984
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3129740173602101e-05,
+      "loss": 0.5945,
+      "step": 12985
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3129163422441389e-05,
+      "loss": 0.4441,
+      "step": 12986
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3128586671280677e-05,
+      "loss": 0.4547,
+      "step": 12987
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3128009920119966e-05,
+      "loss": 0.4548,
+      "step": 12988
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3127433168959254e-05,
+      "loss": 0.461,
+      "step": 12989
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3126856417798541e-05,
+      "loss": 0.547,
+      "step": 12990
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.312627966663783e-05,
+      "loss": 0.5831,
+      "step": 12991
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3125702915477118e-05,
+      "loss": 0.4478,
+      "step": 12992
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3125126164316406e-05,
+      "loss": 0.4939,
+      "step": 12993
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3124549413155696e-05,
+      "loss": 0.5955,
+      "step": 12994
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3123972661994983e-05,
+      "loss": 0.4548,
+      "step": 12995
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3123395910834271e-05,
+      "loss": 0.5782,
+      "step": 12996
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.312281915967356e-05,
+      "loss": 0.4142,
+      "step": 12997
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3122242408512848e-05,
+      "loss": 0.3955,
+      "step": 12998
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3121665657352136e-05,
+      "loss": 0.6255,
+      "step": 12999
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3121088906191423e-05,
+      "loss": 0.5138,
+      "step": 13000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3120512155030713e-05,
+      "loss": 0.5588,
+      "step": 13001
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.311993540387e-05,
+      "loss": 0.467,
+      "step": 13002
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3119358652709288e-05,
+      "loss": 0.5168,
+      "step": 13003
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.311878190154858e-05,
+      "loss": 0.5196,
+      "step": 13004
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3118205150387867e-05,
+      "loss": 0.4751,
+      "step": 13005
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3117628399227153e-05,
+      "loss": 0.5037,
+      "step": 13006
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3117051648066444e-05,
+      "loss": 0.5441,
+      "step": 13007
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3116474896905732e-05,
+      "loss": 0.4204,
+      "step": 13008
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.311589814574502e-05,
+      "loss": 0.3948,
+      "step": 13009
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3115321394584309e-05,
+      "loss": 0.453,
+      "step": 13010
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3114744643423596e-05,
+      "loss": 0.4867,
+      "step": 13011
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3114167892262884e-05,
+      "loss": 0.4425,
+      "step": 13012
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3113591141102173e-05,
+      "loss": 0.4718,
+      "step": 13013
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3113014389941461e-05,
+      "loss": 0.6763,
+      "step": 13014
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3112437638780749e-05,
+      "loss": 0.4727,
+      "step": 13015
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3111860887620038e-05,
+      "loss": 0.4188,
+      "step": 13016
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3111284136459326e-05,
+      "loss": 0.5634,
+      "step": 13017
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3110707385298614e-05,
+      "loss": 0.4372,
+      "step": 13018
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3110130634137903e-05,
+      "loss": 0.564,
+      "step": 13019
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.310955388297719e-05,
+      "loss": 0.6162,
+      "step": 13020
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3108977131816478e-05,
+      "loss": 0.588,
+      "step": 13021
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3108400380655768e-05,
+      "loss": 0.3653,
+      "step": 13022
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3107823629495055e-05,
+      "loss": 0.6687,
+      "step": 13023
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3107246878334343e-05,
+      "loss": 0.5873,
+      "step": 13024
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3106670127173632e-05,
+      "loss": 0.4535,
+      "step": 13025
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.310609337601292e-05,
+      "loss": 0.438,
+      "step": 13026
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3105516624852208e-05,
+      "loss": 0.5644,
+      "step": 13027
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3104939873691497e-05,
+      "loss": 0.4958,
+      "step": 13028
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3104363122530785e-05,
+      "loss": 0.5221,
+      "step": 13029
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3103786371370073e-05,
+      "loss": 0.5094,
+      "step": 13030
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3103209620209364e-05,
+      "loss": 0.4627,
+      "step": 13031
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.310263286904865e-05,
+      "loss": 0.3875,
+      "step": 13032
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3102056117887937e-05,
+      "loss": 0.5375,
+      "step": 13033
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3101479366727225e-05,
+      "loss": 0.4089,
+      "step": 13034
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3100902615566516e-05,
+      "loss": 0.6801,
+      "step": 13035
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3100325864405804e-05,
+      "loss": 0.5307,
+      "step": 13036
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.309974911324509e-05,
+      "loss": 0.5721,
+      "step": 13037
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.309917236208438e-05,
+      "loss": 0.489,
+      "step": 13038
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3098595610923668e-05,
+      "loss": 0.5718,
+      "step": 13039
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3098018859762956e-05,
+      "loss": 0.4856,
+      "step": 13040
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3097442108602246e-05,
+      "loss": 0.4979,
+      "step": 13041
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3096865357441533e-05,
+      "loss": 0.4738,
+      "step": 13042
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3096288606280821e-05,
+      "loss": 0.5716,
+      "step": 13043
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.309571185512011e-05,
+      "loss": 0.5554,
+      "step": 13044
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3095135103959398e-05,
+      "loss": 0.4789,
+      "step": 13045
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3094558352798686e-05,
+      "loss": 0.4892,
+      "step": 13046
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3093981601637975e-05,
+      "loss": 0.6057,
+      "step": 13047
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3093404850477263e-05,
+      "loss": 0.3784,
+      "step": 13048
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.309282809931655e-05,
+      "loss": 0.5868,
+      "step": 13049
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.309225134815584e-05,
+      "loss": 0.5043,
+      "step": 13050
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3091674596995127e-05,
+      "loss": 0.5179,
+      "step": 13051
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3091097845834415e-05,
+      "loss": 0.5738,
+      "step": 13052
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3090521094673705e-05,
+      "loss": 0.4888,
+      "step": 13053
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3089944343512992e-05,
+      "loss": 0.4186,
+      "step": 13054
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.308936759235228e-05,
+      "loss": 0.5048,
+      "step": 13055
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.308879084119157e-05,
+      "loss": 0.5057,
+      "step": 13056
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3088214090030857e-05,
+      "loss": 0.3747,
+      "step": 13057
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3087637338870145e-05,
+      "loss": 0.5783,
+      "step": 13058
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3087060587709434e-05,
+      "loss": 0.477,
+      "step": 13059
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3086483836548722e-05,
+      "loss": 0.5372,
+      "step": 13060
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.308590708538801e-05,
+      "loss": 0.4778,
+      "step": 13061
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.30853303342273e-05,
+      "loss": 0.4322,
+      "step": 13062
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3084753583066586e-05,
+      "loss": 0.5445,
+      "step": 13063
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3084176831905874e-05,
+      "loss": 0.5687,
+      "step": 13064
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3083600080745165e-05,
+      "loss": 0.5035,
+      "step": 13065
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3083023329584453e-05,
+      "loss": 0.529,
+      "step": 13066
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.308244657842374e-05,
+      "loss": 0.5068,
+      "step": 13067
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3081869827263027e-05,
+      "loss": 0.3902,
+      "step": 13068
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3081293076102318e-05,
+      "loss": 0.4485,
+      "step": 13069
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3080716324941605e-05,
+      "loss": 0.4891,
+      "step": 13070
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3080139573780893e-05,
+      "loss": 0.5491,
+      "step": 13071
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3079562822620182e-05,
+      "loss": 0.6382,
+      "step": 13072
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.307898607145947e-05,
+      "loss": 0.5389,
+      "step": 13073
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3078409320298758e-05,
+      "loss": 0.5539,
+      "step": 13074
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3077832569138047e-05,
+      "loss": 0.4636,
+      "step": 13075
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3077255817977335e-05,
+      "loss": 0.526,
+      "step": 13076
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3076679066816623e-05,
+      "loss": 0.4527,
+      "step": 13077
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3076102315655912e-05,
+      "loss": 0.6208,
+      "step": 13078
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.30755255644952e-05,
+      "loss": 0.4808,
+      "step": 13079
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3074948813334487e-05,
+      "loss": 0.3879,
+      "step": 13080
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3074372062173777e-05,
+      "loss": 0.4786,
+      "step": 13081
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3073795311013064e-05,
+      "loss": 0.5326,
+      "step": 13082
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3073218559852352e-05,
+      "loss": 0.4349,
+      "step": 13083
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3072641808691641e-05,
+      "loss": 0.5196,
+      "step": 13084
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3072065057530929e-05,
+      "loss": 0.5156,
+      "step": 13085
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3071488306370217e-05,
+      "loss": 0.4685,
+      "step": 13086
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3070911555209506e-05,
+      "loss": 0.4133,
+      "step": 13087
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3070334804048794e-05,
+      "loss": 0.5559,
+      "step": 13088
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3069758052888082e-05,
+      "loss": 0.5745,
+      "step": 13089
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3069181301727371e-05,
+      "loss": 0.4389,
+      "step": 13090
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3068604550566659e-05,
+      "loss": 0.5575,
+      "step": 13091
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3068027799405946e-05,
+      "loss": 0.4734,
+      "step": 13092
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3067451048245237e-05,
+      "loss": 0.6819,
+      "step": 13093
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3066874297084523e-05,
+      "loss": 0.6467,
+      "step": 13094
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3066297545923811e-05,
+      "loss": 0.498,
+      "step": 13095
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3065720794763102e-05,
+      "loss": 0.5387,
+      "step": 13096
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.306514404360239e-05,
+      "loss": 0.3751,
+      "step": 13097
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3064567292441677e-05,
+      "loss": 0.5288,
+      "step": 13098
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3063990541280967e-05,
+      "loss": 0.5069,
+      "step": 13099
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3063413790120255e-05,
+      "loss": 0.5196,
+      "step": 13100
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3062837038959542e-05,
+      "loss": 0.4966,
+      "step": 13101
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3062260287798832e-05,
+      "loss": 0.5623,
+      "step": 13102
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.306168353663812e-05,
+      "loss": 0.4589,
+      "step": 13103
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3061106785477407e-05,
+      "loss": 0.6364,
+      "step": 13104
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3060530034316695e-05,
+      "loss": 0.6187,
+      "step": 13105
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3059953283155984e-05,
+      "loss": 0.4599,
+      "step": 13106
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3059376531995272e-05,
+      "loss": 0.5207,
+      "step": 13107
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.305879978083456e-05,
+      "loss": 0.5657,
+      "step": 13108
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3058223029673849e-05,
+      "loss": 0.4259,
+      "step": 13109
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3057646278513136e-05,
+      "loss": 0.4493,
+      "step": 13110
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3057069527352424e-05,
+      "loss": 0.4589,
+      "step": 13111
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3056492776191714e-05,
+      "loss": 0.4737,
+      "step": 13112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3055916025031001e-05,
+      "loss": 0.5784,
+      "step": 13113
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3055339273870289e-05,
+      "loss": 0.5179,
+      "step": 13114
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3054762522709578e-05,
+      "loss": 0.4805,
+      "step": 13115
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3054185771548866e-05,
+      "loss": 0.5092,
+      "step": 13116
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3053609020388154e-05,
+      "loss": 0.4351,
+      "step": 13117
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3053032269227443e-05,
+      "loss": 0.5646,
+      "step": 13118
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.305245551806673e-05,
+      "loss": 0.5098,
+      "step": 13119
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3051878766906018e-05,
+      "loss": 0.4677,
+      "step": 13120
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3051302015745308e-05,
+      "loss": 0.4919,
+      "step": 13121
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3050725264584595e-05,
+      "loss": 0.5523,
+      "step": 13122
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3050148513423883e-05,
+      "loss": 0.5597,
+      "step": 13123
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3049571762263174e-05,
+      "loss": 0.5403,
+      "step": 13124
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.304899501110246e-05,
+      "loss": 0.5637,
+      "step": 13125
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3048418259941748e-05,
+      "loss": 0.5485,
+      "step": 13126
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3047841508781039e-05,
+      "loss": 0.4491,
+      "step": 13127
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3047264757620327e-05,
+      "loss": 0.5352,
+      "step": 13128
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3046688006459614e-05,
+      "loss": 0.5262,
+      "step": 13129
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3046111255298904e-05,
+      "loss": 0.3219,
+      "step": 13130
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3045534504138191e-05,
+      "loss": 0.4423,
+      "step": 13131
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3044957752977479e-05,
+      "loss": 0.5646,
+      "step": 13132
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3044381001816768e-05,
+      "loss": 0.5887,
+      "step": 13133
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3043804250656056e-05,
+      "loss": 0.5485,
+      "step": 13134
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3043227499495344e-05,
+      "loss": 0.4376,
+      "step": 13135
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3042650748334633e-05,
+      "loss": 0.5055,
+      "step": 13136
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3042073997173921e-05,
+      "loss": 0.4473,
+      "step": 13137
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.3041497246013209e-05,
+      "loss": 0.5022,
+      "step": 13138
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3040920494852496e-05,
+      "loss": 0.5425,
+      "step": 13139
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3040343743691786e-05,
+      "loss": 0.5513,
+      "step": 13140
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3039766992531073e-05,
+      "loss": 0.414,
+      "step": 13141
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3039190241370361e-05,
+      "loss": 0.4487,
+      "step": 13142
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.303861349020965e-05,
+      "loss": 0.5531,
+      "step": 13143
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3038036739048938e-05,
+      "loss": 0.4775,
+      "step": 13144
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3037459987888226e-05,
+      "loss": 0.5126,
+      "step": 13145
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3036883236727515e-05,
+      "loss": 0.4563,
+      "step": 13146
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3036306485566803e-05,
+      "loss": 0.4795,
+      "step": 13147
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.303572973440609e-05,
+      "loss": 0.5902,
+      "step": 13148
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.303515298324538e-05,
+      "loss": 0.6081,
+      "step": 13149
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3034576232084668e-05,
+      "loss": 0.3945,
+      "step": 13150
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3033999480923955e-05,
+      "loss": 0.3498,
+      "step": 13151
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3033422729763245e-05,
+      "loss": 0.496,
+      "step": 13152
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3032845978602532e-05,
+      "loss": 0.6168,
+      "step": 13153
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.303226922744182e-05,
+      "loss": 0.4548,
+      "step": 13154
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3031692476281111e-05,
+      "loss": 0.5164,
+      "step": 13155
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3031115725120397e-05,
+      "loss": 0.4662,
+      "step": 13156
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3030538973959685e-05,
+      "loss": 0.4846,
+      "step": 13157
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3029962222798976e-05,
+      "loss": 0.5344,
+      "step": 13158
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3029385471638264e-05,
+      "loss": 0.5866,
+      "step": 13159
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3028808720477551e-05,
+      "loss": 0.4292,
+      "step": 13160
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.302823196931684e-05,
+      "loss": 0.4345,
+      "step": 13161
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3027655218156128e-05,
+      "loss": 0.6565,
+      "step": 13162
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3027078466995416e-05,
+      "loss": 0.5639,
+      "step": 13163
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3026501715834705e-05,
+      "loss": 0.3722,
+      "step": 13164
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3025924964673993e-05,
+      "loss": 0.4214,
+      "step": 13165
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.302534821351328e-05,
+      "loss": 0.5671,
+      "step": 13166
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.302477146235257e-05,
+      "loss": 0.5231,
+      "step": 13167
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3024194711191858e-05,
+      "loss": 0.4606,
+      "step": 13168
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3023617960031145e-05,
+      "loss": 0.5186,
+      "step": 13169
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3023041208870435e-05,
+      "loss": 0.5702,
+      "step": 13170
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3022464457709723e-05,
+      "loss": 0.5394,
+      "step": 13171
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.302188770654901e-05,
+      "loss": 0.4031,
+      "step": 13172
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3021310955388298e-05,
+      "loss": 0.5255,
+      "step": 13173
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3020734204227587e-05,
+      "loss": 0.5569,
+      "step": 13174
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3020157453066875e-05,
+      "loss": 0.5296,
+      "step": 13175
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3019580701906163e-05,
+      "loss": 0.5214,
+      "step": 13176
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3019003950745452e-05,
+      "loss": 0.5737,
+      "step": 13177
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.301842719958474e-05,
+      "loss": 0.5603,
+      "step": 13178
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3017850448424027e-05,
+      "loss": 0.5437,
+      "step": 13179
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3017273697263317e-05,
+      "loss": 0.5772,
+      "step": 13180
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3016696946102604e-05,
+      "loss": 0.5574,
+      "step": 13181
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3016120194941892e-05,
+      "loss": 0.378,
+      "step": 13182
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3015543443781182e-05,
+      "loss": 0.5542,
+      "step": 13183
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.301496669262047e-05,
+      "loss": 0.5718,
+      "step": 13184
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3014389941459757e-05,
+      "loss": 0.4722,
+      "step": 13185
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3013813190299048e-05,
+      "loss": 0.62,
+      "step": 13186
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3013236439138334e-05,
+      "loss": 0.4106,
+      "step": 13187
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3012659687977622e-05,
+      "loss": 0.352,
+      "step": 13188
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3012082936816913e-05,
+      "loss": 0.572,
+      "step": 13189
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.30115061856562e-05,
+      "loss": 0.3973,
+      "step": 13190
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3010929434495488e-05,
+      "loss": 0.4629,
+      "step": 13191
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3010352683334777e-05,
+      "loss": 0.5322,
+      "step": 13192
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3009775932174065e-05,
+      "loss": 0.5236,
+      "step": 13193
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3009199181013353e-05,
+      "loss": 0.4385,
+      "step": 13194
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3008622429852642e-05,
+      "loss": 0.5658,
+      "step": 13195
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.300804567869193e-05,
+      "loss": 0.5336,
+      "step": 13196
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3007468927531218e-05,
+      "loss": 0.4739,
+      "step": 13197
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3006892176370507e-05,
+      "loss": 0.5763,
+      "step": 13198
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3006315425209795e-05,
+      "loss": 0.5024,
+      "step": 13199
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3005738674049082e-05,
+      "loss": 0.5312,
+      "step": 13200
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3005161922888372e-05,
+      "loss": 0.4338,
+      "step": 13201
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.300458517172766e-05,
+      "loss": 0.4853,
+      "step": 13202
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3004008420566947e-05,
+      "loss": 0.5411,
+      "step": 13203
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3003431669406236e-05,
+      "loss": 0.3969,
+      "step": 13204
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3002854918245524e-05,
+      "loss": 0.5094,
+      "step": 13205
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3002278167084812e-05,
+      "loss": 0.4958,
+      "step": 13206
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.30017014159241e-05,
+      "loss": 0.4318,
+      "step": 13207
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3001124664763389e-05,
+      "loss": 0.6248,
+      "step": 13208
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.3000547913602677e-05,
+      "loss": 0.4558,
+      "step": 13209
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2999971162441964e-05,
+      "loss": 0.5905,
+      "step": 13210
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2999394411281254e-05,
+      "loss": 0.3683,
+      "step": 13211
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2998817660120541e-05,
+      "loss": 0.5186,
+      "step": 13212
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2998240908959829e-05,
+      "loss": 0.4554,
+      "step": 13213
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2997664157799118e-05,
+      "loss": 0.5515,
+      "step": 13214
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2997087406638406e-05,
+      "loss": 0.5754,
+      "step": 13215
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2996510655477694e-05,
+      "loss": 0.5901,
+      "step": 13216
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2995933904316985e-05,
+      "loss": 0.4192,
+      "step": 13217
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.299535715315627e-05,
+      "loss": 0.4609,
+      "step": 13218
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2994780401995559e-05,
+      "loss": 0.6293,
+      "step": 13219
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.299420365083485e-05,
+      "loss": 0.4582,
+      "step": 13220
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2993626899674137e-05,
+      "loss": 0.533,
+      "step": 13221
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2993050148513425e-05,
+      "loss": 0.4848,
+      "step": 13222
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2992473397352714e-05,
+      "loss": 0.5708,
+      "step": 13223
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2991896646192002e-05,
+      "loss": 0.4454,
+      "step": 13224
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.299131989503129e-05,
+      "loss": 0.4672,
+      "step": 13225
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2990743143870579e-05,
+      "loss": 0.4669,
+      "step": 13226
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2990166392709867e-05,
+      "loss": 0.5017,
+      "step": 13227
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2989589641549154e-05,
+      "loss": 0.575,
+      "step": 13228
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2989012890388444e-05,
+      "loss": 0.5242,
+      "step": 13229
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2988436139227732e-05,
+      "loss": 0.4664,
+      "step": 13230
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.298785938806702e-05,
+      "loss": 0.5617,
+      "step": 13231
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2987282636906309e-05,
+      "loss": 0.4187,
+      "step": 13232
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2986705885745596e-05,
+      "loss": 0.3772,
+      "step": 13233
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2986129134584884e-05,
+      "loss": 0.3829,
+      "step": 13234
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2985552383424173e-05,
+      "loss": 0.5432,
+      "step": 13235
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2984975632263461e-05,
+      "loss": 0.5537,
+      "step": 13236
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2984398881102749e-05,
+      "loss": 0.4873,
+      "step": 13237
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2983822129942038e-05,
+      "loss": 0.6129,
+      "step": 13238
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2983245378781326e-05,
+      "loss": 0.464,
+      "step": 13239
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2982668627620613e-05,
+      "loss": 0.5722,
+      "step": 13240
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2982091876459903e-05,
+      "loss": 0.453,
+      "step": 13241
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.298151512529919e-05,
+      "loss": 0.4944,
+      "step": 13242
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2980938374138478e-05,
+      "loss": 0.6109,
+      "step": 13243
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2980361622977766e-05,
+      "loss": 0.5393,
+      "step": 13244
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2979784871817055e-05,
+      "loss": 0.648,
+      "step": 13245
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2979208120656343e-05,
+      "loss": 0.545,
+      "step": 13246
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.297863136949563e-05,
+      "loss": 0.4353,
+      "step": 13247
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2978054618334922e-05,
+      "loss": 0.4746,
+      "step": 13248
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2977477867174208e-05,
+      "loss": 0.529,
+      "step": 13249
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2976901116013495e-05,
+      "loss": 0.4608,
+      "step": 13250
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2976324364852786e-05,
+      "loss": 0.4006,
+      "step": 13251
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2975747613692074e-05,
+      "loss": 0.4057,
+      "step": 13252
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2975170862531362e-05,
+      "loss": 0.4497,
+      "step": 13253
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2974594111370651e-05,
+      "loss": 0.5398,
+      "step": 13254
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2974017360209939e-05,
+      "loss": 0.4736,
+      "step": 13255
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2973440609049227e-05,
+      "loss": 0.5319,
+      "step": 13256
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2972863857888516e-05,
+      "loss": 0.4605,
+      "step": 13257
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2972287106727804e-05,
+      "loss": 0.4684,
+      "step": 13258
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2971710355567091e-05,
+      "loss": 0.5427,
+      "step": 13259
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.297113360440638e-05,
+      "loss": 0.4429,
+      "step": 13260
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2970556853245668e-05,
+      "loss": 0.4186,
+      "step": 13261
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2969980102084956e-05,
+      "loss": 0.5155,
+      "step": 13262
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2969403350924245e-05,
+      "loss": 0.5035,
+      "step": 13263
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2968826599763533e-05,
+      "loss": 0.4474,
+      "step": 13264
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296824984860282e-05,
+      "loss": 0.5261,
+      "step": 13265
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296767309744211e-05,
+      "loss": 0.4696,
+      "step": 13266
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2967096346281398e-05,
+      "loss": 0.5695,
+      "step": 13267
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2966519595120686e-05,
+      "loss": 0.4447,
+      "step": 13268
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2965942843959975e-05,
+      "loss": 0.3803,
+      "step": 13269
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2965366092799263e-05,
+      "loss": 0.4975,
+      "step": 13270
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296478934163855e-05,
+      "loss": 0.549,
+      "step": 13271
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296421259047784e-05,
+      "loss": 0.4196,
+      "step": 13272
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2963635839317127e-05,
+      "loss": 0.5063,
+      "step": 13273
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2963059088156415e-05,
+      "loss": 0.7123,
+      "step": 13274
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2962482336995706e-05,
+      "loss": 0.5645,
+      "step": 13275
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2961905585834992e-05,
+      "loss": 0.4363,
+      "step": 13276
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.296132883467428e-05,
+      "loss": 0.4446,
+      "step": 13277
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2960752083513568e-05,
+      "loss": 0.34,
+      "step": 13278
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2960175332352859e-05,
+      "loss": 0.4686,
+      "step": 13279
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2959598581192146e-05,
+      "loss": 0.4971,
+      "step": 13280
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2959021830031432e-05,
+      "loss": 0.495,
+      "step": 13281
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2958445078870723e-05,
+      "loss": 0.4044,
+      "step": 13282
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2957868327710011e-05,
+      "loss": 0.5596,
+      "step": 13283
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2957291576549299e-05,
+      "loss": 0.5683,
+      "step": 13284
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2956714825388588e-05,
+      "loss": 0.5346,
+      "step": 13285
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2956138074227876e-05,
+      "loss": 0.5951,
+      "step": 13286
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2955561323067163e-05,
+      "loss": 0.484,
+      "step": 13287
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2954984571906453e-05,
+      "loss": 0.3708,
+      "step": 13288
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.295440782074574e-05,
+      "loss": 0.4899,
+      "step": 13289
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2953831069585028e-05,
+      "loss": 0.4108,
+      "step": 13290
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2953254318424318e-05,
+      "loss": 0.4538,
+      "step": 13291
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2952677567263605e-05,
+      "loss": 0.5346,
+      "step": 13292
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2952100816102893e-05,
+      "loss": 0.5581,
+      "step": 13293
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2951524064942182e-05,
+      "loss": 0.4043,
+      "step": 13294
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.295094731378147e-05,
+      "loss": 0.6629,
+      "step": 13295
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2950370562620758e-05,
+      "loss": 0.4994,
+      "step": 13296
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2949793811460047e-05,
+      "loss": 0.5313,
+      "step": 13297
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2949217060299335e-05,
+      "loss": 0.5041,
+      "step": 13298
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2948640309138622e-05,
+      "loss": 0.3292,
+      "step": 13299
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2948063557977912e-05,
+      "loss": 0.7076,
+      "step": 13300
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.29474868068172e-05,
+      "loss": 0.5091,
+      "step": 13301
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2946910055656487e-05,
+      "loss": 0.4437,
+      "step": 13302
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2946333304495777e-05,
+      "loss": 0.5437,
+      "step": 13303
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2945756553335064e-05,
+      "loss": 0.4198,
+      "step": 13304
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2945179802174352e-05,
+      "loss": 0.6378,
+      "step": 13305
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2944603051013643e-05,
+      "loss": 0.4337,
+      "step": 13306
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2944026299852929e-05,
+      "loss": 0.4927,
+      "step": 13307
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2943449548692217e-05,
+      "loss": 0.5124,
+      "step": 13308
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2942872797531508e-05,
+      "loss": 0.5071,
+      "step": 13309
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2942296046370795e-05,
+      "loss": 0.4789,
+      "step": 13310
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2941719295210083e-05,
+      "loss": 0.4869,
+      "step": 13311
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2941142544049369e-05,
+      "loss": 0.6256,
+      "step": 13312
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.294056579288866e-05,
+      "loss": 0.5037,
+      "step": 13313
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2939989041727948e-05,
+      "loss": 0.5235,
+      "step": 13314
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2939412290567236e-05,
+      "loss": 0.5296,
+      "step": 13315
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 1.2938835539406525e-05,
+      "loss": 0.5327,
+      "step": 13316
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2938258788245813e-05,
+      "loss": 0.4948,
+      "step": 13317
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.29376820370851e-05,
+      "loss": 0.5169,
+      "step": 13318
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.293710528592439e-05,
+      "loss": 0.5492,
+      "step": 13319
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2936528534763677e-05,
+      "loss": 0.5001,
+      "step": 13320
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2935951783602965e-05,
+      "loss": 0.4285,
+      "step": 13321
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2935375032442254e-05,
+      "loss": 0.4048,
+      "step": 13322
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2934798281281542e-05,
+      "loss": 0.3702,
+      "step": 13323
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.293422153012083e-05,
+      "loss": 0.439,
+      "step": 13324
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.293364477896012e-05,
+      "loss": 0.5413,
+      "step": 13325
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2933068027799407e-05,
+      "loss": 0.5386,
+      "step": 13326
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2932491276638695e-05,
+      "loss": 0.5893,
+      "step": 13327
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2931914525477984e-05,
+      "loss": 0.612,
+      "step": 13328
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2931337774317272e-05,
+      "loss": 0.4767,
+      "step": 13329
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.293076102315656e-05,
+      "loss": 0.496,
+      "step": 13330
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2930184271995849e-05,
+      "loss": 0.5212,
+      "step": 13331
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2929607520835136e-05,
+      "loss": 0.5203,
+      "step": 13332
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2929030769674424e-05,
+      "loss": 0.4688,
+      "step": 13333
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2928454018513713e-05,
+      "loss": 0.5934,
+      "step": 13334
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2927877267353001e-05,
+      "loss": 0.6189,
+      "step": 13335
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2927300516192289e-05,
+      "loss": 0.5016,
+      "step": 13336
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.292672376503158e-05,
+      "loss": 0.5911,
+      "step": 13337
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2926147013870866e-05,
+      "loss": 0.5136,
+      "step": 13338
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2925570262710154e-05,
+      "loss": 0.6669,
+      "step": 13339
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2924993511549445e-05,
+      "loss": 0.4451,
+      "step": 13340
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2924416760388732e-05,
+      "loss": 0.5207,
+      "step": 13341
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.292384000922802e-05,
+      "loss": 0.5927,
+      "step": 13342
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.292326325806731e-05,
+      "loss": 0.5219,
+      "step": 13343
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2922686506906597e-05,
+      "loss": 0.5069,
+      "step": 13344
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2922109755745885e-05,
+      "loss": 0.4548,
+      "step": 13345
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2921533004585172e-05,
+      "loss": 0.5359,
+      "step": 13346
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2920956253424462e-05,
+      "loss": 0.389,
+      "step": 13347
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.292037950226375e-05,
+      "loss": 0.6056,
+      "step": 13348
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2919802751103037e-05,
+      "loss": 0.3723,
+      "step": 13349
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2919225999942327e-05,
+      "loss": 0.5395,
+      "step": 13350
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2918649248781614e-05,
+      "loss": 0.4988,
+      "step": 13351
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2918072497620902e-05,
+      "loss": 0.4526,
+      "step": 13352
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2917495746460191e-05,
+      "loss": 0.4429,
+      "step": 13353
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2916918995299479e-05,
+      "loss": 0.6264,
+      "step": 13354
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2916342244138767e-05,
+      "loss": 0.5798,
+      "step": 13355
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2915765492978056e-05,
+      "loss": 0.5298,
+      "step": 13356
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2915188741817344e-05,
+      "loss": 0.3583,
+      "step": 13357
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2914611990656631e-05,
+      "loss": 0.4778,
+      "step": 13358
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.291403523949592e-05,
+      "loss": 0.5843,
+      "step": 13359
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2913458488335209e-05,
+      "loss": 0.5106,
+      "step": 13360
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2912881737174496e-05,
+      "loss": 0.5551,
+      "step": 13361
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2912304986013786e-05,
+      "loss": 0.4765,
+      "step": 13362
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2911728234853073e-05,
+      "loss": 0.4751,
+      "step": 13363
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2911151483692361e-05,
+      "loss": 0.4693,
+      "step": 13364
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.291057473253165e-05,
+      "loss": 0.5485,
+      "step": 13365
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2909997981370938e-05,
+      "loss": 0.4282,
+      "step": 13366
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2909421230210226e-05,
+      "loss": 0.4945,
+      "step": 13367
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2908844479049517e-05,
+      "loss": 0.3944,
+      "step": 13368
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2908267727888803e-05,
+      "loss": 0.4148,
+      "step": 13369
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.290769097672809e-05,
+      "loss": 0.4507,
+      "step": 13370
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2907114225567382e-05,
+      "loss": 0.5461,
+      "step": 13371
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.290653747440667e-05,
+      "loss": 0.4962,
+      "step": 13372
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2905960723245957e-05,
+      "loss": 0.5125,
+      "step": 13373
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2905383972085246e-05,
+      "loss": 0.435,
+      "step": 13374
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2904807220924534e-05,
+      "loss": 0.49,
+      "step": 13375
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2904230469763822e-05,
+      "loss": 0.5358,
+      "step": 13376
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2903653718603111e-05,
+      "loss": 0.5736,
+      "step": 13377
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2903076967442399e-05,
+      "loss": 0.5453,
+      "step": 13378
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2902500216281686e-05,
+      "loss": 0.4033,
+      "step": 13379
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2901923465120976e-05,
+      "loss": 0.3851,
+      "step": 13380
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2901346713960263e-05,
+      "loss": 0.6121,
+      "step": 13381
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2900769962799551e-05,
+      "loss": 0.4327,
+      "step": 13382
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2900193211638839e-05,
+      "loss": 0.387,
+      "step": 13383
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2899616460478128e-05,
+      "loss": 0.4628,
+      "step": 13384
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2899039709317416e-05,
+      "loss": 0.5315,
+      "step": 13385
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2898462958156704e-05,
+      "loss": 0.5616,
+      "step": 13386
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2897886206995993e-05,
+      "loss": 0.5619,
+      "step": 13387
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.289730945583528e-05,
+      "loss": 0.5375,
+      "step": 13388
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2896732704674568e-05,
+      "loss": 0.5476,
+      "step": 13389
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2896155953513858e-05,
+      "loss": 0.4914,
+      "step": 13390
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2895579202353145e-05,
+      "loss": 0.5934,
+      "step": 13391
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2895002451192433e-05,
+      "loss": 0.5627,
+      "step": 13392
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2894425700031722e-05,
+      "loss": 0.535,
+      "step": 13393
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.289384894887101e-05,
+      "loss": 0.5192,
+      "step": 13394
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2893272197710298e-05,
+      "loss": 0.4444,
+      "step": 13395
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2892695446549587e-05,
+      "loss": 0.4297,
+      "step": 13396
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2892118695388875e-05,
+      "loss": 0.5486,
+      "step": 13397
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2891541944228163e-05,
+      "loss": 0.4812,
+      "step": 13398
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2890965193067454e-05,
+      "loss": 0.4009,
+      "step": 13399
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.289038844190674e-05,
+      "loss": 0.4601,
+      "step": 13400
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2889811690746027e-05,
+      "loss": 0.5245,
+      "step": 13401
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2889234939585318e-05,
+      "loss": 0.5711,
+      "step": 13402
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2888658188424606e-05,
+      "loss": 0.5324,
+      "step": 13403
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2888081437263894e-05,
+      "loss": 0.5,
+      "step": 13404
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2887504686103183e-05,
+      "loss": 0.6156,
+      "step": 13405
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.288692793494247e-05,
+      "loss": 0.589,
+      "step": 13406
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2886351183781759e-05,
+      "loss": 0.5014,
+      "step": 13407
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2885774432621048e-05,
+      "loss": 0.4825,
+      "step": 13408
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2885197681460336e-05,
+      "loss": 0.3501,
+      "step": 13409
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2884620930299623e-05,
+      "loss": 0.4432,
+      "step": 13410
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2884044179138913e-05,
+      "loss": 0.559,
+      "step": 13411
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.28834674279782e-05,
+      "loss": 0.4575,
+      "step": 13412
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2882890676817488e-05,
+      "loss": 0.5928,
+      "step": 13413
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2882313925656777e-05,
+      "loss": 0.3118,
+      "step": 13414
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2881737174496065e-05,
+      "loss": 0.5097,
+      "step": 13415
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2881160423335353e-05,
+      "loss": 0.4509,
+      "step": 13416
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.288058367217464e-05,
+      "loss": 0.5003,
+      "step": 13417
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.288000692101393e-05,
+      "loss": 0.5522,
+      "step": 13418
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2879430169853218e-05,
+      "loss": 0.5101,
+      "step": 13419
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2878853418692505e-05,
+      "loss": 0.4738,
+      "step": 13420
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2878276667531795e-05,
+      "loss": 0.529,
+      "step": 13421
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2877699916371082e-05,
+      "loss": 0.4683,
+      "step": 13422
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.287712316521037e-05,
+      "loss": 0.5358,
+      "step": 13423
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.287654641404966e-05,
+      "loss": 0.4753,
+      "step": 13424
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2875969662888947e-05,
+      "loss": 0.5482,
+      "step": 13425
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2875392911728235e-05,
+      "loss": 0.4854,
+      "step": 13426
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2874816160567524e-05,
+      "loss": 0.4933,
+      "step": 13427
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2874239409406812e-05,
+      "loss": 0.3988,
+      "step": 13428
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.28736626582461e-05,
+      "loss": 0.5833,
+      "step": 13429
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.287308590708539e-05,
+      "loss": 0.4844,
+      "step": 13430
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2872509155924677e-05,
+      "loss": 0.4573,
+      "step": 13431
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2871932404763964e-05,
+      "loss": 0.4828,
+      "step": 13432
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2871355653603255e-05,
+      "loss": 0.6257,
+      "step": 13433
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2870778902442543e-05,
+      "loss": 0.4303,
+      "step": 13434
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.287020215128183e-05,
+      "loss": 0.4996,
+      "step": 13435
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.286962540012112e-05,
+      "loss": 0.5786,
+      "step": 13436
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2869048648960408e-05,
+      "loss": 0.5534,
+      "step": 13437
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2868471897799695e-05,
+      "loss": 0.447,
+      "step": 13438
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2867895146638985e-05,
+      "loss": 0.3802,
+      "step": 13439
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2867318395478272e-05,
+      "loss": 0.6199,
+      "step": 13440
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.286674164431756e-05,
+      "loss": 0.5582,
+      "step": 13441
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.286616489315685e-05,
+      "loss": 0.5062,
+      "step": 13442
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2865588141996137e-05,
+      "loss": 0.5413,
+      "step": 13443
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2865011390835425e-05,
+      "loss": 0.3807,
+      "step": 13444
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2864434639674714e-05,
+      "loss": 0.5087,
+      "step": 13445
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2863857888514002e-05,
+      "loss": 0.453,
+      "step": 13446
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.286328113735329e-05,
+      "loss": 0.4568,
+      "step": 13447
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2862704386192579e-05,
+      "loss": 0.4316,
+      "step": 13448
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2862127635031867e-05,
+      "loss": 0.5713,
+      "step": 13449
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2861550883871154e-05,
+      "loss": 0.4259,
+      "step": 13450
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2860974132710442e-05,
+      "loss": 0.4172,
+      "step": 13451
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2860397381549731e-05,
+      "loss": 0.4712,
+      "step": 13452
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2859820630389019e-05,
+      "loss": 0.5738,
+      "step": 13453
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2859243879228307e-05,
+      "loss": 0.5472,
+      "step": 13454
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2858667128067596e-05,
+      "loss": 0.4772,
+      "step": 13455
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2858090376906884e-05,
+      "loss": 0.6093,
+      "step": 13456
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2857513625746172e-05,
+      "loss": 0.5115,
+      "step": 13457
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2856936874585461e-05,
+      "loss": 0.377,
+      "step": 13458
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2856360123424749e-05,
+      "loss": 0.6532,
+      "step": 13459
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2855783372264036e-05,
+      "loss": 0.5185,
+      "step": 13460
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2855206621103327e-05,
+      "loss": 0.4995,
+      "step": 13461
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2854629869942613e-05,
+      "loss": 0.4175,
+      "step": 13462
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2854053118781901e-05,
+      "loss": 0.65,
+      "step": 13463
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2853476367621192e-05,
+      "loss": 0.4902,
+      "step": 13464
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.285289961646048e-05,
+      "loss": 0.4868,
+      "step": 13465
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2852322865299768e-05,
+      "loss": 0.4798,
+      "step": 13466
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2851746114139057e-05,
+      "loss": 0.547,
+      "step": 13467
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2851169362978345e-05,
+      "loss": 0.4953,
+      "step": 13468
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2850592611817632e-05,
+      "loss": 0.517,
+      "step": 13469
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2850015860656922e-05,
+      "loss": 0.5095,
+      "step": 13470
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.284943910949621e-05,
+      "loss": 0.4833,
+      "step": 13471
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2848862358335497e-05,
+      "loss": 0.4799,
+      "step": 13472
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2848285607174786e-05,
+      "loss": 0.5621,
+      "step": 13473
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2847708856014074e-05,
+      "loss": 0.5077,
+      "step": 13474
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2847132104853362e-05,
+      "loss": 0.4651,
+      "step": 13475
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2846555353692651e-05,
+      "loss": 0.5494,
+      "step": 13476
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2845978602531939e-05,
+      "loss": 0.5842,
+      "step": 13477
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2845401851371226e-05,
+      "loss": 0.5404,
+      "step": 13478
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2844825100210516e-05,
+      "loss": 0.6186,
+      "step": 13479
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2844248349049804e-05,
+      "loss": 0.5365,
+      "step": 13480
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2843671597889091e-05,
+      "loss": 0.4894,
+      "step": 13481
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.284309484672838e-05,
+      "loss": 0.538,
+      "step": 13482
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2842518095567668e-05,
+      "loss": 0.6301,
+      "step": 13483
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2841941344406956e-05,
+      "loss": 0.5301,
+      "step": 13484
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2841364593246244e-05,
+      "loss": 0.5667,
+      "step": 13485
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2840787842085533e-05,
+      "loss": 0.5599,
+      "step": 13486
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.284021109092482e-05,
+      "loss": 0.4456,
+      "step": 13487
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2839634339764108e-05,
+      "loss": 0.4738,
+      "step": 13488
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2839057588603398e-05,
+      "loss": 0.3359,
+      "step": 13489
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2838480837442685e-05,
+      "loss": 0.6061,
+      "step": 13490
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2837904086281973e-05,
+      "loss": 0.4793,
+      "step": 13491
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2837327335121264e-05,
+      "loss": 0.4055,
+      "step": 13492
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.283675058396055e-05,
+      "loss": 0.4096,
+      "step": 13493
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2836173832799838e-05,
+      "loss": 0.6154,
+      "step": 13494
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 1.2835597081639129e-05,
+      "loss": 0.5608,
+      "step": 13495
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2835020330478417e-05,
+      "loss": 0.4347,
+      "step": 13496
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2834443579317704e-05,
+      "loss": 0.5759,
+      "step": 13497
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2833866828156994e-05,
+      "loss": 0.4578,
+      "step": 13498
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2833290076996281e-05,
+      "loss": 0.3538,
+      "step": 13499
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2832713325835569e-05,
+      "loss": 0.4491,
+      "step": 13500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2832136574674859e-05,
+      "loss": 0.4549,
+      "step": 13501
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2831559823514146e-05,
+      "loss": 0.4683,
+      "step": 13502
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2830983072353434e-05,
+      "loss": 0.5208,
+      "step": 13503
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2830406321192723e-05,
+      "loss": 0.5528,
+      "step": 13504
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2829829570032011e-05,
+      "loss": 0.5033,
+      "step": 13505
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2829252818871299e-05,
+      "loss": 0.3903,
+      "step": 13506
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2828676067710588e-05,
+      "loss": 0.6232,
+      "step": 13507
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2828099316549876e-05,
+      "loss": 0.5687,
+      "step": 13508
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2827522565389163e-05,
+      "loss": 0.5431,
+      "step": 13509
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2826945814228453e-05,
+      "loss": 0.4983,
+      "step": 13510
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.282636906306774e-05,
+      "loss": 0.4582,
+      "step": 13511
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2825792311907028e-05,
+      "loss": 0.4721,
+      "step": 13512
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2825215560746318e-05,
+      "loss": 0.4947,
+      "step": 13513
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2824638809585605e-05,
+      "loss": 0.5135,
+      "step": 13514
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2824062058424893e-05,
+      "loss": 0.6063,
+      "step": 13515
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2823485307264182e-05,
+      "loss": 0.5548,
+      "step": 13516
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.282290855610347e-05,
+      "loss": 0.4195,
+      "step": 13517
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2822331804942758e-05,
+      "loss": 0.5586,
+      "step": 13518
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2821755053782049e-05,
+      "loss": 0.4826,
+      "step": 13519
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2821178302621335e-05,
+      "loss": 0.4661,
+      "step": 13520
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2820601551460622e-05,
+      "loss": 0.5146,
+      "step": 13521
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.282002480029991e-05,
+      "loss": 0.4462,
+      "step": 13522
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2819448049139201e-05,
+      "loss": 0.569,
+      "step": 13523
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2818871297978487e-05,
+      "loss": 0.4871,
+      "step": 13524
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2818294546817775e-05,
+      "loss": 0.4131,
+      "step": 13525
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2817717795657066e-05,
+      "loss": 0.5788,
+      "step": 13526
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2817141044496354e-05,
+      "loss": 0.5144,
+      "step": 13527
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2816564293335641e-05,
+      "loss": 0.5404,
+      "step": 13528
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.281598754217493e-05,
+      "loss": 0.4332,
+      "step": 13529
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2815410791014218e-05,
+      "loss": 0.4999,
+      "step": 13530
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2814834039853506e-05,
+      "loss": 0.6103,
+      "step": 13531
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2814257288692795e-05,
+      "loss": 0.535,
+      "step": 13532
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2813680537532083e-05,
+      "loss": 0.526,
+      "step": 13533
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.281310378637137e-05,
+      "loss": 0.6189,
+      "step": 13534
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.281252703521066e-05,
+      "loss": 0.4983,
+      "step": 13535
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2811950284049948e-05,
+      "loss": 0.4704,
+      "step": 13536
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2811373532889235e-05,
+      "loss": 0.513,
+      "step": 13537
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2810796781728525e-05,
+      "loss": 0.4462,
+      "step": 13538
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2810220030567813e-05,
+      "loss": 0.4103,
+      "step": 13539
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.28096432794071e-05,
+      "loss": 0.6034,
+      "step": 13540
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.280906652824639e-05,
+      "loss": 0.4724,
+      "step": 13541
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2808489777085677e-05,
+      "loss": 0.4307,
+      "step": 13542
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2807913025924965e-05,
+      "loss": 0.4815,
+      "step": 13543
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2807336274764254e-05,
+      "loss": 0.6079,
+      "step": 13544
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2806759523603542e-05,
+      "loss": 0.4947,
+      "step": 13545
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.280618277244283e-05,
+      "loss": 0.5554,
+      "step": 13546
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2805606021282119e-05,
+      "loss": 0.5117,
+      "step": 13547
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2805029270121407e-05,
+      "loss": 0.576,
+      "step": 13548
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2804452518960694e-05,
+      "loss": 0.5239,
+      "step": 13549
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2803875767799986e-05,
+      "loss": 0.4811,
+      "step": 13550
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2803299016639272e-05,
+      "loss": 0.4866,
+      "step": 13551
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.280272226547856e-05,
+      "loss": 0.5765,
+      "step": 13552
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.280214551431785e-05,
+      "loss": 0.5257,
+      "step": 13553
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2801568763157138e-05,
+      "loss": 0.4403,
+      "step": 13554
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2800992011996424e-05,
+      "loss": 0.5091,
+      "step": 13555
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2800415260835712e-05,
+      "loss": 0.5377,
+      "step": 13556
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2799838509675003e-05,
+      "loss": 0.6154,
+      "step": 13557
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.279926175851429e-05,
+      "loss": 0.5212,
+      "step": 13558
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2798685007353578e-05,
+      "loss": 0.557,
+      "step": 13559
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2798108256192868e-05,
+      "loss": 0.5883,
+      "step": 13560
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2797531505032155e-05,
+      "loss": 0.5122,
+      "step": 13561
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2796954753871443e-05,
+      "loss": 0.3727,
+      "step": 13562
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2796378002710732e-05,
+      "loss": 0.5224,
+      "step": 13563
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.279580125155002e-05,
+      "loss": 0.5322,
+      "step": 13564
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2795224500389308e-05,
+      "loss": 0.5976,
+      "step": 13565
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2794647749228597e-05,
+      "loss": 0.4665,
+      "step": 13566
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2794070998067885e-05,
+      "loss": 0.4315,
+      "step": 13567
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2793494246907172e-05,
+      "loss": 0.5089,
+      "step": 13568
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2792917495746462e-05,
+      "loss": 0.5311,
+      "step": 13569
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.279234074458575e-05,
+      "loss": 0.4274,
+      "step": 13570
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2791763993425037e-05,
+      "loss": 0.5252,
+      "step": 13571
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2791187242264326e-05,
+      "loss": 0.5105,
+      "step": 13572
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2790610491103614e-05,
+      "loss": 0.5192,
+      "step": 13573
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2790033739942902e-05,
+      "loss": 0.5605,
+      "step": 13574
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2789456988782191e-05,
+      "loss": 0.5496,
+      "step": 13575
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2788880237621479e-05,
+      "loss": 0.4049,
+      "step": 13576
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2788303486460767e-05,
+      "loss": 0.4959,
+      "step": 13577
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2787726735300056e-05,
+      "loss": 0.4576,
+      "step": 13578
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2787149984139344e-05,
+      "loss": 0.4038,
+      "step": 13579
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2786573232978631e-05,
+      "loss": 0.4788,
+      "step": 13580
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2785996481817922e-05,
+      "loss": 0.3989,
+      "step": 13581
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2785419730657208e-05,
+      "loss": 0.5921,
+      "step": 13582
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2784842979496496e-05,
+      "loss": 0.5161,
+      "step": 13583
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2784266228335787e-05,
+      "loss": 0.4002,
+      "step": 13584
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2783689477175075e-05,
+      "loss": 0.5848,
+      "step": 13585
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2783112726014361e-05,
+      "loss": 0.4028,
+      "step": 13586
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2782535974853652e-05,
+      "loss": 0.5854,
+      "step": 13587
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.278195922369294e-05,
+      "loss": 0.4606,
+      "step": 13588
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2781382472532227e-05,
+      "loss": 0.5577,
+      "step": 13589
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2780805721371515e-05,
+      "loss": 0.3325,
+      "step": 13590
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2780228970210804e-05,
+      "loss": 0.4451,
+      "step": 13591
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2779652219050092e-05,
+      "loss": 0.4814,
+      "step": 13592
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.277907546788938e-05,
+      "loss": 0.4897,
+      "step": 13593
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2778498716728669e-05,
+      "loss": 0.3858,
+      "step": 13594
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2777921965567957e-05,
+      "loss": 0.5067,
+      "step": 13595
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2777345214407244e-05,
+      "loss": 0.5984,
+      "step": 13596
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2776768463246534e-05,
+      "loss": 0.5119,
+      "step": 13597
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2776191712085822e-05,
+      "loss": 0.5319,
+      "step": 13598
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.277561496092511e-05,
+      "loss": 0.4897,
+      "step": 13599
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2775038209764399e-05,
+      "loss": 0.5576,
+      "step": 13600
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2774461458603686e-05,
+      "loss": 0.5467,
+      "step": 13601
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2773884707442974e-05,
+      "loss": 0.5525,
+      "step": 13602
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2773307956282263e-05,
+      "loss": 0.6501,
+      "step": 13603
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2772731205121551e-05,
+      "loss": 0.4493,
+      "step": 13604
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2772154453960839e-05,
+      "loss": 0.46,
+      "step": 13605
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2771577702800128e-05,
+      "loss": 0.3377,
+      "step": 13606
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2771000951639416e-05,
+      "loss": 0.4851,
+      "step": 13607
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2770424200478703e-05,
+      "loss": 0.6836,
+      "step": 13608
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2769847449317993e-05,
+      "loss": 0.5525,
+      "step": 13609
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.276927069815728e-05,
+      "loss": 0.519,
+      "step": 13610
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2768693946996568e-05,
+      "loss": 0.5336,
+      "step": 13611
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.276811719583586e-05,
+      "loss": 0.4745,
+      "step": 13612
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2767540444675145e-05,
+      "loss": 0.5275,
+      "step": 13613
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2766963693514433e-05,
+      "loss": 0.6756,
+      "step": 13614
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2766386942353724e-05,
+      "loss": 0.4875,
+      "step": 13615
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2765810191193012e-05,
+      "loss": 0.6036,
+      "step": 13616
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2765233440032298e-05,
+      "loss": 0.6346,
+      "step": 13617
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2764656688871589e-05,
+      "loss": 0.5673,
+      "step": 13618
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2764079937710876e-05,
+      "loss": 0.4432,
+      "step": 13619
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2763503186550164e-05,
+      "loss": 0.4687,
+      "step": 13620
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2762926435389454e-05,
+      "loss": 0.592,
+      "step": 13621
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2762349684228741e-05,
+      "loss": 0.4875,
+      "step": 13622
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2761772933068029e-05,
+      "loss": 0.5809,
+      "step": 13623
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2761196181907318e-05,
+      "loss": 0.5758,
+      "step": 13624
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2760619430746606e-05,
+      "loss": 0.5063,
+      "step": 13625
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2760042679585894e-05,
+      "loss": 0.3948,
+      "step": 13626
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2759465928425181e-05,
+      "loss": 0.4368,
+      "step": 13627
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.275888917726447e-05,
+      "loss": 0.6759,
+      "step": 13628
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2758312426103758e-05,
+      "loss": 0.3731,
+      "step": 13629
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2757735674943046e-05,
+      "loss": 0.4691,
+      "step": 13630
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2757158923782335e-05,
+      "loss": 0.3978,
+      "step": 13631
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2756582172621623e-05,
+      "loss": 0.4771,
+      "step": 13632
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2756005421460911e-05,
+      "loss": 0.6968,
+      "step": 13633
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.27554286703002e-05,
+      "loss": 0.4025,
+      "step": 13634
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2754851919139488e-05,
+      "loss": 0.6045,
+      "step": 13635
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2754275167978776e-05,
+      "loss": 0.5266,
+      "step": 13636
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2753698416818065e-05,
+      "loss": 0.5945,
+      "step": 13637
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2753121665657353e-05,
+      "loss": 0.5612,
+      "step": 13638
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.275254491449664e-05,
+      "loss": 0.5438,
+      "step": 13639
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.275196816333593e-05,
+      "loss": 0.4757,
+      "step": 13640
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2751391412175217e-05,
+      "loss": 0.6165,
+      "step": 13641
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2750814661014505e-05,
+      "loss": 0.4127,
+      "step": 13642
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2750237909853796e-05,
+      "loss": 0.418,
+      "step": 13643
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2749661158693082e-05,
+      "loss": 0.4015,
+      "step": 13644
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.274908440753237e-05,
+      "loss": 0.5172,
+      "step": 13645
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2748507656371661e-05,
+      "loss": 0.4413,
+      "step": 13646
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2747930905210949e-05,
+      "loss": 0.3876,
+      "step": 13647
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2747354154050236e-05,
+      "loss": 0.4542,
+      "step": 13648
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2746777402889526e-05,
+      "loss": 0.4958,
+      "step": 13649
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2746200651728813e-05,
+      "loss": 0.491,
+      "step": 13650
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2745623900568101e-05,
+      "loss": 0.4785,
+      "step": 13651
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.274504714940739e-05,
+      "loss": 0.499,
+      "step": 13652
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2744470398246678e-05,
+      "loss": 0.4887,
+      "step": 13653
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2743893647085966e-05,
+      "loss": 0.4097,
+      "step": 13654
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2743316895925255e-05,
+      "loss": 0.3498,
+      "step": 13655
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2742740144764543e-05,
+      "loss": 0.5005,
+      "step": 13656
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.274216339360383e-05,
+      "loss": 0.3573,
+      "step": 13657
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.274158664244312e-05,
+      "loss": 0.4472,
+      "step": 13658
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2741009891282408e-05,
+      "loss": 0.6188,
+      "step": 13659
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2740433140121695e-05,
+      "loss": 0.466,
+      "step": 13660
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2739856388960983e-05,
+      "loss": 0.4489,
+      "step": 13661
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2739279637800272e-05,
+      "loss": 0.4524,
+      "step": 13662
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.273870288663956e-05,
+      "loss": 0.4171,
+      "step": 13663
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2738126135478848e-05,
+      "loss": 0.4568,
+      "step": 13664
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2737549384318137e-05,
+      "loss": 0.52,
+      "step": 13665
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2736972633157425e-05,
+      "loss": 0.5303,
+      "step": 13666
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2736395881996712e-05,
+      "loss": 0.4352,
+      "step": 13667
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2735819130836002e-05,
+      "loss": 0.5943,
+      "step": 13668
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.273524237967529e-05,
+      "loss": 0.4935,
+      "step": 13669
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2734665628514577e-05,
+      "loss": 0.4666,
+      "step": 13670
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2734088877353867e-05,
+      "loss": 0.5599,
+      "step": 13671
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2733512126193154e-05,
+      "loss": 0.654,
+      "step": 13672
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2732935375032442e-05,
+      "loss": 0.4925,
+      "step": 13673
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.2732358623871733e-05,
+      "loss": 0.4648,
+      "step": 13674
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2731781872711019e-05,
+      "loss": 0.5294,
+      "step": 13675
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2731205121550307e-05,
+      "loss": 0.4666,
+      "step": 13676
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2730628370389598e-05,
+      "loss": 0.4924,
+      "step": 13677
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2730051619228885e-05,
+      "loss": 0.5129,
+      "step": 13678
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2729474868068173e-05,
+      "loss": 0.4633,
+      "step": 13679
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2728898116907463e-05,
+      "loss": 0.4248,
+      "step": 13680
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272832136574675e-05,
+      "loss": 0.597,
+      "step": 13681
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2727744614586038e-05,
+      "loss": 0.4146,
+      "step": 13682
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2727167863425327e-05,
+      "loss": 0.5962,
+      "step": 13683
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2726591112264615e-05,
+      "loss": 0.4119,
+      "step": 13684
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2726014361103903e-05,
+      "loss": 0.5896,
+      "step": 13685
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2725437609943192e-05,
+      "loss": 0.5426,
+      "step": 13686
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272486085878248e-05,
+      "loss": 0.5243,
+      "step": 13687
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2724284107621767e-05,
+      "loss": 0.4184,
+      "step": 13688
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2723707356461057e-05,
+      "loss": 0.4821,
+      "step": 13689
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2723130605300344e-05,
+      "loss": 0.4268,
+      "step": 13690
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2722553854139632e-05,
+      "loss": 0.5278,
+      "step": 13691
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2721977102978922e-05,
+      "loss": 0.3874,
+      "step": 13692
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.272140035181821e-05,
+      "loss": 0.4309,
+      "step": 13693
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2720823600657497e-05,
+      "loss": 0.5862,
+      "step": 13694
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2720246849496785e-05,
+      "loss": 0.4989,
+      "step": 13695
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2719670098336074e-05,
+      "loss": 0.5787,
+      "step": 13696
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2719093347175362e-05,
+      "loss": 0.3679,
+      "step": 13697
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.271851659601465e-05,
+      "loss": 0.6619,
+      "step": 13698
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2717939844853939e-05,
+      "loss": 0.4558,
+      "step": 13699
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2717363093693226e-05,
+      "loss": 0.5733,
+      "step": 13700
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2716786342532514e-05,
+      "loss": 0.4456,
+      "step": 13701
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2716209591371803e-05,
+      "loss": 0.5116,
+      "step": 13702
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2715632840211091e-05,
+      "loss": 0.4739,
+      "step": 13703
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2715056089050379e-05,
+      "loss": 0.5456,
+      "step": 13704
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.271447933788967e-05,
+      "loss": 0.4263,
+      "step": 13705
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2713902586728956e-05,
+      "loss": 0.5189,
+      "step": 13706
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2713325835568244e-05,
+      "loss": 0.5746,
+      "step": 13707
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2712749084407535e-05,
+      "loss": 0.6325,
+      "step": 13708
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2712172333246822e-05,
+      "loss": 0.5389,
+      "step": 13709
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.271159558208611e-05,
+      "loss": 0.6457,
+      "step": 13710
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.27110188309254e-05,
+      "loss": 0.5099,
+      "step": 13711
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2710442079764687e-05,
+      "loss": 0.4716,
+      "step": 13712
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2709865328603975e-05,
+      "loss": 0.4915,
+      "step": 13713
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2709288577443264e-05,
+      "loss": 0.5654,
+      "step": 13714
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2708711826282552e-05,
+      "loss": 0.4679,
+      "step": 13715
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.270813507512184e-05,
+      "loss": 0.5326,
+      "step": 13716
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2707558323961129e-05,
+      "loss": 0.5424,
+      "step": 13717
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2706981572800417e-05,
+      "loss": 0.3777,
+      "step": 13718
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2706404821639704e-05,
+      "loss": 0.4564,
+      "step": 13719
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2705828070478994e-05,
+      "loss": 0.5282,
+      "step": 13720
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2705251319318281e-05,
+      "loss": 0.4636,
+      "step": 13721
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2704674568157569e-05,
+      "loss": 0.4232,
+      "step": 13722
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2704097816996858e-05,
+      "loss": 0.5293,
+      "step": 13723
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2703521065836146e-05,
+      "loss": 0.4872,
+      "step": 13724
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2702944314675434e-05,
+      "loss": 0.5447,
+      "step": 13725
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2702367563514723e-05,
+      "loss": 0.5656,
+      "step": 13726
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2701790812354011e-05,
+      "loss": 0.465,
+      "step": 13727
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2701214061193299e-05,
+      "loss": 0.5825,
+      "step": 13728
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2700637310032586e-05,
+      "loss": 0.5495,
+      "step": 13729
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2700060558871876e-05,
+      "loss": 0.5213,
+      "step": 13730
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2699483807711163e-05,
+      "loss": 0.56,
+      "step": 13731
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2698907056550451e-05,
+      "loss": 0.5181,
+      "step": 13732
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.269833030538974e-05,
+      "loss": 0.3748,
+      "step": 13733
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2697753554229028e-05,
+      "loss": 0.5054,
+      "step": 13734
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2697176803068316e-05,
+      "loss": 0.5861,
+      "step": 13735
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2696600051907607e-05,
+      "loss": 0.4084,
+      "step": 13736
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2696023300746893e-05,
+      "loss": 0.5038,
+      "step": 13737
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.269544654958618e-05,
+      "loss": 0.4796,
+      "step": 13738
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2694869798425472e-05,
+      "loss": 0.5308,
+      "step": 13739
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.269429304726476e-05,
+      "loss": 0.4146,
+      "step": 13740
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2693716296104047e-05,
+      "loss": 0.4489,
+      "step": 13741
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2693139544943336e-05,
+      "loss": 0.5367,
+      "step": 13742
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2692562793782624e-05,
+      "loss": 0.4217,
+      "step": 13743
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2691986042621912e-05,
+      "loss": 0.5047,
+      "step": 13744
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2691409291461201e-05,
+      "loss": 0.5127,
+      "step": 13745
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2690832540300489e-05,
+      "loss": 0.5255,
+      "step": 13746
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2690255789139776e-05,
+      "loss": 0.5281,
+      "step": 13747
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2689679037979066e-05,
+      "loss": 0.4301,
+      "step": 13748
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2689102286818353e-05,
+      "loss": 0.525,
+      "step": 13749
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2688525535657641e-05,
+      "loss": 0.4947,
+      "step": 13750
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.268794878449693e-05,
+      "loss": 0.5794,
+      "step": 13751
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2687372033336218e-05,
+      "loss": 0.4412,
+      "step": 13752
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2686795282175506e-05,
+      "loss": 0.468,
+      "step": 13753
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2686218531014795e-05,
+      "loss": 0.5197,
+      "step": 13754
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2685641779854083e-05,
+      "loss": 0.4515,
+      "step": 13755
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.268506502869337e-05,
+      "loss": 0.4354,
+      "step": 13756
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.268448827753266e-05,
+      "loss": 0.4957,
+      "step": 13757
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2683911526371948e-05,
+      "loss": 0.4301,
+      "step": 13758
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2683334775211235e-05,
+      "loss": 0.4972,
+      "step": 13759
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2682758024050525e-05,
+      "loss": 0.5456,
+      "step": 13760
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2682181272889812e-05,
+      "loss": 0.3785,
+      "step": 13761
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.26816045217291e-05,
+      "loss": 0.642,
+      "step": 13762
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2681027770568391e-05,
+      "loss": 0.4184,
+      "step": 13763
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2680451019407677e-05,
+      "loss": 0.5206,
+      "step": 13764
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2679874268246965e-05,
+      "loss": 0.5317,
+      "step": 13765
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2679297517086253e-05,
+      "loss": 0.6077,
+      "step": 13766
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2678720765925544e-05,
+      "loss": 0.5098,
+      "step": 13767
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.267814401476483e-05,
+      "loss": 0.6027,
+      "step": 13768
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2677567263604117e-05,
+      "loss": 0.4154,
+      "step": 13769
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2676990512443408e-05,
+      "loss": 0.4774,
+      "step": 13770
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2676413761282696e-05,
+      "loss": 0.578,
+      "step": 13771
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2675837010121984e-05,
+      "loss": 0.5568,
+      "step": 13772
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2675260258961273e-05,
+      "loss": 0.5018,
+      "step": 13773
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2674683507800561e-05,
+      "loss": 0.5368,
+      "step": 13774
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2674106756639849e-05,
+      "loss": 0.4682,
+      "step": 13775
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2673530005479138e-05,
+      "loss": 0.5814,
+      "step": 13776
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2672953254318426e-05,
+      "loss": 0.4515,
+      "step": 13777
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2672376503157713e-05,
+      "loss": 0.3573,
+      "step": 13778
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2671799751997003e-05,
+      "loss": 0.4826,
+      "step": 13779
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.267122300083629e-05,
+      "loss": 0.4295,
+      "step": 13780
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2670646249675578e-05,
+      "loss": 0.5686,
+      "step": 13781
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2670069498514867e-05,
+      "loss": 0.4252,
+      "step": 13782
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2669492747354155e-05,
+      "loss": 0.4766,
+      "step": 13783
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2668915996193443e-05,
+      "loss": 0.5315,
+      "step": 13784
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2668339245032732e-05,
+      "loss": 0.4215,
+      "step": 13785
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.266776249387202e-05,
+      "loss": 0.4358,
+      "step": 13786
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2667185742711308e-05,
+      "loss": 0.5056,
+      "step": 13787
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2666608991550597e-05,
+      "loss": 0.5596,
+      "step": 13788
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2666032240389885e-05,
+      "loss": 0.4961,
+      "step": 13789
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2665455489229172e-05,
+      "loss": 0.5026,
+      "step": 13790
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2664878738068462e-05,
+      "loss": 0.4503,
+      "step": 13791
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.266430198690775e-05,
+      "loss": 0.5024,
+      "step": 13792
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2663725235747037e-05,
+      "loss": 0.4106,
+      "step": 13793
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2663148484586328e-05,
+      "loss": 0.4621,
+      "step": 13794
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2662571733425614e-05,
+      "loss": 0.4419,
+      "step": 13795
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2661994982264902e-05,
+      "loss": 0.4881,
+      "step": 13796
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2661418231104193e-05,
+      "loss": 0.3306,
+      "step": 13797
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.266084147994348e-05,
+      "loss": 0.3796,
+      "step": 13798
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2660264728782767e-05,
+      "loss": 0.5026,
+      "step": 13799
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2659687977622054e-05,
+      "loss": 0.5583,
+      "step": 13800
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2659111226461345e-05,
+      "loss": 0.5104,
+      "step": 13801
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2658534475300633e-05,
+      "loss": 0.5757,
+      "step": 13802
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.265795772413992e-05,
+      "loss": 0.4658,
+      "step": 13803
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.265738097297921e-05,
+      "loss": 0.5017,
+      "step": 13804
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2656804221818498e-05,
+      "loss": 0.5998,
+      "step": 13805
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2656227470657785e-05,
+      "loss": 0.5025,
+      "step": 13806
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2655650719497075e-05,
+      "loss": 0.5803,
+      "step": 13807
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2655073968336362e-05,
+      "loss": 0.4688,
+      "step": 13808
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.265449721717565e-05,
+      "loss": 0.6295,
+      "step": 13809
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.265392046601494e-05,
+      "loss": 0.573,
+      "step": 13810
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2653343714854227e-05,
+      "loss": 0.4845,
+      "step": 13811
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2652766963693515e-05,
+      "loss": 0.5758,
+      "step": 13812
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2652190212532804e-05,
+      "loss": 0.5087,
+      "step": 13813
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2651613461372092e-05,
+      "loss": 0.4454,
+      "step": 13814
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.265103671021138e-05,
+      "loss": 0.5832,
+      "step": 13815
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2650459959050669e-05,
+      "loss": 0.5286,
+      "step": 13816
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2649883207889957e-05,
+      "loss": 0.4454,
+      "step": 13817
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2649306456729244e-05,
+      "loss": 0.5753,
+      "step": 13818
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2648729705568534e-05,
+      "loss": 0.5234,
+      "step": 13819
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2648152954407821e-05,
+      "loss": 0.5153,
+      "step": 13820
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.264757620324711e-05,
+      "loss": 0.4602,
+      "step": 13821
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2646999452086399e-05,
+      "loss": 0.5208,
+      "step": 13822
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2646422700925686e-05,
+      "loss": 0.546,
+      "step": 13823
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2645845949764974e-05,
+      "loss": 0.4503,
+      "step": 13824
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2645269198604265e-05,
+      "loss": 0.5499,
+      "step": 13825
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2644692447443551e-05,
+      "loss": 0.4379,
+      "step": 13826
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2644115696282839e-05,
+      "loss": 0.4468,
+      "step": 13827
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.264353894512213e-05,
+      "loss": 0.5863,
+      "step": 13828
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2642962193961417e-05,
+      "loss": 0.4786,
+      "step": 13829
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2642385442800703e-05,
+      "loss": 0.4113,
+      "step": 13830
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2641808691639994e-05,
+      "loss": 0.4065,
+      "step": 13831
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2641231940479282e-05,
+      "loss": 0.502,
+      "step": 13832
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.264065518931857e-05,
+      "loss": 0.4736,
+      "step": 13833
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2640078438157858e-05,
+      "loss": 0.4242,
+      "step": 13834
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2639501686997147e-05,
+      "loss": 0.4868,
+      "step": 13835
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2638924935836435e-05,
+      "loss": 0.5425,
+      "step": 13836
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2638348184675722e-05,
+      "loss": 0.4268,
+      "step": 13837
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2637771433515012e-05,
+      "loss": 0.5203,
+      "step": 13838
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.26371946823543e-05,
+      "loss": 0.5399,
+      "step": 13839
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2636617931193587e-05,
+      "loss": 0.5887,
+      "step": 13840
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2636041180032876e-05,
+      "loss": 0.5698,
+      "step": 13841
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2635464428872164e-05,
+      "loss": 0.4605,
+      "step": 13842
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2634887677711452e-05,
+      "loss": 0.4826,
+      "step": 13843
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2634310926550741e-05,
+      "loss": 0.5132,
+      "step": 13844
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2633734175390029e-05,
+      "loss": 0.4839,
+      "step": 13845
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2633157424229317e-05,
+      "loss": 0.4887,
+      "step": 13846
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2632580673068606e-05,
+      "loss": 0.5799,
+      "step": 13847
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2632003921907894e-05,
+      "loss": 0.4412,
+      "step": 13848
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2631427170747181e-05,
+      "loss": 0.4836,
+      "step": 13849
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.263085041958647e-05,
+      "loss": 0.5381,
+      "step": 13850
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2630273668425758e-05,
+      "loss": 0.4802,
+      "step": 13851
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2629696917265046e-05,
+      "loss": 0.5566,
+      "step": 13852
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 1.2629120166104335e-05,
+      "loss": 0.4796,
+      "step": 13853
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2628543414943623e-05,
+      "loss": 0.5192,
+      "step": 13854
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.262796666378291e-05,
+      "loss": 0.5222,
+      "step": 13855
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2627389912622202e-05,
+      "loss": 0.5092,
+      "step": 13856
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2626813161461488e-05,
+      "loss": 0.5835,
+      "step": 13857
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2626236410300776e-05,
+      "loss": 0.4603,
+      "step": 13858
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2625659659140067e-05,
+      "loss": 0.6397,
+      "step": 13859
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2625082907979354e-05,
+      "loss": 0.4489,
+      "step": 13860
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.262450615681864e-05,
+      "loss": 0.5327,
+      "step": 13861
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2623929405657931e-05,
+      "loss": 0.4733,
+      "step": 13862
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2623352654497219e-05,
+      "loss": 0.3976,
+      "step": 13863
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2622775903336507e-05,
+      "loss": 0.4698,
+      "step": 13864
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2622199152175796e-05,
+      "loss": 0.4419,
+      "step": 13865
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2621622401015084e-05,
+      "loss": 0.4769,
+      "step": 13866
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2621045649854371e-05,
+      "loss": 0.5934,
+      "step": 13867
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.262046889869366e-05,
+      "loss": 0.5677,
+      "step": 13868
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2619892147532949e-05,
+      "loss": 0.4496,
+      "step": 13869
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2619315396372236e-05,
+      "loss": 0.5113,
+      "step": 13870
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2618738645211524e-05,
+      "loss": 0.4924,
+      "step": 13871
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2618161894050813e-05,
+      "loss": 0.4769,
+      "step": 13872
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2617585142890101e-05,
+      "loss": 0.4338,
+      "step": 13873
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2617008391729389e-05,
+      "loss": 0.5924,
+      "step": 13874
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2616431640568678e-05,
+      "loss": 0.5093,
+      "step": 13875
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2615854889407966e-05,
+      "loss": 0.3906,
+      "step": 13876
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2615278138247253e-05,
+      "loss": 0.5736,
+      "step": 13877
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2614701387086543e-05,
+      "loss": 0.4916,
+      "step": 13878
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.261412463592583e-05,
+      "loss": 0.4431,
+      "step": 13879
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2613547884765118e-05,
+      "loss": 0.5669,
+      "step": 13880
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2612971133604408e-05,
+      "loss": 0.5319,
+      "step": 13881
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2612394382443695e-05,
+      "loss": 0.57,
+      "step": 13882
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2611817631282983e-05,
+      "loss": 0.5006,
+      "step": 13883
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2611240880122272e-05,
+      "loss": 0.4616,
+      "step": 13884
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.261066412896156e-05,
+      "loss": 0.4385,
+      "step": 13885
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2610087377800848e-05,
+      "loss": 0.4663,
+      "step": 13886
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2609510626640139e-05,
+      "loss": 0.5345,
+      "step": 13887
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2608933875479425e-05,
+      "loss": 0.541,
+      "step": 13888
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2608357124318712e-05,
+      "loss": 0.5093,
+      "step": 13889
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2607780373158003e-05,
+      "loss": 0.5239,
+      "step": 13890
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2607203621997291e-05,
+      "loss": 0.4921,
+      "step": 13891
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2606626870836577e-05,
+      "loss": 0.5561,
+      "step": 13892
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2606050119675868e-05,
+      "loss": 0.3871,
+      "step": 13893
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2605473368515156e-05,
+      "loss": 0.5011,
+      "step": 13894
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2604896617354444e-05,
+      "loss": 0.6265,
+      "step": 13895
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2604319866193733e-05,
+      "loss": 0.3976,
+      "step": 13896
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260374311503302e-05,
+      "loss": 0.5427,
+      "step": 13897
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2603166363872308e-05,
+      "loss": 0.4871,
+      "step": 13898
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2602589612711598e-05,
+      "loss": 0.5784,
+      "step": 13899
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2602012861550885e-05,
+      "loss": 0.3946,
+      "step": 13900
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2601436110390173e-05,
+      "loss": 0.5177,
+      "step": 13901
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2600859359229462e-05,
+      "loss": 0.3795,
+      "step": 13902
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.260028260806875e-05,
+      "loss": 0.5176,
+      "step": 13903
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2599705856908038e-05,
+      "loss": 0.5558,
+      "step": 13904
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2599129105747326e-05,
+      "loss": 0.4688,
+      "step": 13905
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2598552354586615e-05,
+      "loss": 0.5685,
+      "step": 13906
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2597975603425903e-05,
+      "loss": 0.5484,
+      "step": 13907
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259739885226519e-05,
+      "loss": 0.4408,
+      "step": 13908
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259682210110448e-05,
+      "loss": 0.5103,
+      "step": 13909
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2596245349943767e-05,
+      "loss": 0.3837,
+      "step": 13910
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2595668598783055e-05,
+      "loss": 0.5558,
+      "step": 13911
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2595091847622344e-05,
+      "loss": 0.5183,
+      "step": 13912
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2594515096461632e-05,
+      "loss": 0.6093,
+      "step": 13913
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259393834530092e-05,
+      "loss": 0.4877,
+      "step": 13914
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259336159414021e-05,
+      "loss": 0.4185,
+      "step": 13915
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2592784842979497e-05,
+      "loss": 0.4182,
+      "step": 13916
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2592208091818785e-05,
+      "loss": 0.5694,
+      "step": 13917
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2591631340658076e-05,
+      "loss": 0.4895,
+      "step": 13918
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2591054589497362e-05,
+      "loss": 0.4826,
+      "step": 13919
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.259047783833665e-05,
+      "loss": 0.5001,
+      "step": 13920
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258990108717594e-05,
+      "loss": 0.5697,
+      "step": 13921
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2589324336015228e-05,
+      "loss": 0.455,
+      "step": 13922
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2588747584854514e-05,
+      "loss": 0.451,
+      "step": 13923
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2588170833693805e-05,
+      "loss": 0.5024,
+      "step": 13924
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2587594082533093e-05,
+      "loss": 0.4511,
+      "step": 13925
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258701733137238e-05,
+      "loss": 0.528,
+      "step": 13926
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258644058021167e-05,
+      "loss": 0.4379,
+      "step": 13927
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2585863829050958e-05,
+      "loss": 0.5116,
+      "step": 13928
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2585287077890245e-05,
+      "loss": 0.4248,
+      "step": 13929
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2584710326729535e-05,
+      "loss": 0.5092,
+      "step": 13930
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2584133575568822e-05,
+      "loss": 0.4144,
+      "step": 13931
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258355682440811e-05,
+      "loss": 0.6786,
+      "step": 13932
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.25829800732474e-05,
+      "loss": 0.4646,
+      "step": 13933
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2582403322086687e-05,
+      "loss": 0.5236,
+      "step": 13934
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2581826570925975e-05,
+      "loss": 0.5506,
+      "step": 13935
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2581249819765264e-05,
+      "loss": 0.4118,
+      "step": 13936
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2580673068604552e-05,
+      "loss": 0.3567,
+      "step": 13937
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.258009631744384e-05,
+      "loss": 0.4876,
+      "step": 13938
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2579519566283127e-05,
+      "loss": 0.4646,
+      "step": 13939
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2578942815122417e-05,
+      "loss": 0.4436,
+      "step": 13940
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2578366063961704e-05,
+      "loss": 0.5433,
+      "step": 13941
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2577789312800992e-05,
+      "loss": 0.5035,
+      "step": 13942
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2577212561640281e-05,
+      "loss": 0.5901,
+      "step": 13943
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2576635810479569e-05,
+      "loss": 0.4782,
+      "step": 13944
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2576059059318857e-05,
+      "loss": 0.5714,
+      "step": 13945
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2575482308158146e-05,
+      "loss": 0.4766,
+      "step": 13946
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2574905556997434e-05,
+      "loss": 0.5201,
+      "step": 13947
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2574328805836721e-05,
+      "loss": 0.3865,
+      "step": 13948
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2573752054676012e-05,
+      "loss": 0.4576,
+      "step": 13949
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2573175303515298e-05,
+      "loss": 0.4841,
+      "step": 13950
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2572598552354586e-05,
+      "loss": 0.4387,
+      "step": 13951
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2572021801193877e-05,
+      "loss": 0.4762,
+      "step": 13952
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2571445050033165e-05,
+      "loss": 0.6427,
+      "step": 13953
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2570868298872451e-05,
+      "loss": 0.5248,
+      "step": 13954
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2570291547711742e-05,
+      "loss": 0.5075,
+      "step": 13955
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.256971479655103e-05,
+      "loss": 0.4145,
+      "step": 13956
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2569138045390317e-05,
+      "loss": 0.5154,
+      "step": 13957
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2568561294229607e-05,
+      "loss": 0.4165,
+      "step": 13958
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2567984543068894e-05,
+      "loss": 0.4693,
+      "step": 13959
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2567407791908182e-05,
+      "loss": 0.4519,
+      "step": 13960
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2566831040747471e-05,
+      "loss": 0.5052,
+      "step": 13961
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.256625428958676e-05,
+      "loss": 0.5455,
+      "step": 13962
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2565677538426047e-05,
+      "loss": 0.3636,
+      "step": 13963
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2565100787265336e-05,
+      "loss": 0.6291,
+      "step": 13964
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2564524036104624e-05,
+      "loss": 0.5093,
+      "step": 13965
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2563947284943912e-05,
+      "loss": 0.5188,
+      "step": 13966
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2563370533783201e-05,
+      "loss": 0.7024,
+      "step": 13967
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2562793782622489e-05,
+      "loss": 0.4887,
+      "step": 13968
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2562217031461776e-05,
+      "loss": 0.563,
+      "step": 13969
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2561640280301066e-05,
+      "loss": 0.5328,
+      "step": 13970
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2561063529140353e-05,
+      "loss": 0.391,
+      "step": 13971
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2560486777979641e-05,
+      "loss": 0.5233,
+      "step": 13972
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2559910026818929e-05,
+      "loss": 0.6288,
+      "step": 13973
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2559333275658218e-05,
+      "loss": 0.4693,
+      "step": 13974
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2558756524497506e-05,
+      "loss": 0.4328,
+      "step": 13975
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2558179773336794e-05,
+      "loss": 0.4685,
+      "step": 13976
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2557603022176083e-05,
+      "loss": 0.4818,
+      "step": 13977
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.255702627101537e-05,
+      "loss": 0.448,
+      "step": 13978
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2556449519854658e-05,
+      "loss": 0.5208,
+      "step": 13979
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.255587276869395e-05,
+      "loss": 0.4864,
+      "step": 13980
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2555296017533235e-05,
+      "loss": 0.4978,
+      "step": 13981
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2554719266372523e-05,
+      "loss": 0.5797,
+      "step": 13982
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2554142515211814e-05,
+      "loss": 0.4939,
+      "step": 13983
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2553565764051102e-05,
+      "loss": 0.5836,
+      "step": 13984
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2552989012890388e-05,
+      "loss": 0.4719,
+      "step": 13985
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2552412261729679e-05,
+      "loss": 0.5824,
+      "step": 13986
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2551835510568967e-05,
+      "loss": 0.5523,
+      "step": 13987
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2551258759408254e-05,
+      "loss": 0.5661,
+      "step": 13988
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2550682008247544e-05,
+      "loss": 0.4928,
+      "step": 13989
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2550105257086831e-05,
+      "loss": 0.4297,
+      "step": 13990
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2549528505926119e-05,
+      "loss": 0.4665,
+      "step": 13991
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2548951754765408e-05,
+      "loss": 0.5194,
+      "step": 13992
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2548375003604696e-05,
+      "loss": 0.4732,
+      "step": 13993
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2547798252443984e-05,
+      "loss": 0.517,
+      "step": 13994
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2547221501283273e-05,
+      "loss": 0.5649,
+      "step": 13995
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.254664475012256e-05,
+      "loss": 0.6194,
+      "step": 13996
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2546067998961848e-05,
+      "loss": 0.5491,
+      "step": 13997
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2545491247801138e-05,
+      "loss": 0.4882,
+      "step": 13998
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2544914496640426e-05,
+      "loss": 0.5202,
+      "step": 13999
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2544337745479713e-05,
+      "loss": 0.5109,
+      "step": 14000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2543760994319003e-05,
+      "loss": 0.5551,
+      "step": 14001
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.254318424315829e-05,
+      "loss": 0.4236,
+      "step": 14002
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2542607491997578e-05,
+      "loss": 0.4056,
+      "step": 14003
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2542030740836867e-05,
+      "loss": 0.6122,
+      "step": 14004
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2541453989676155e-05,
+      "loss": 0.2757,
+      "step": 14005
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2540877238515443e-05,
+      "loss": 0.5437,
+      "step": 14006
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2540300487354732e-05,
+      "loss": 0.5499,
+      "step": 14007
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.253972373619402e-05,
+      "loss": 0.5547,
+      "step": 14008
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2539146985033307e-05,
+      "loss": 0.5718,
+      "step": 14009
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2538570233872595e-05,
+      "loss": 0.4803,
+      "step": 14010
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2537993482711886e-05,
+      "loss": 0.6644,
+      "step": 14011
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2537416731551172e-05,
+      "loss": 0.5014,
+      "step": 14012
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.253683998039046e-05,
+      "loss": 0.4817,
+      "step": 14013
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2536263229229751e-05,
+      "loss": 0.4033,
+      "step": 14014
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2535686478069039e-05,
+      "loss": 0.5629,
+      "step": 14015
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2535109726908326e-05,
+      "loss": 0.4873,
+      "step": 14016
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2534532975747616e-05,
+      "loss": 0.4735,
+      "step": 14017
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2533956224586903e-05,
+      "loss": 0.5442,
+      "step": 14018
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2533379473426191e-05,
+      "loss": 0.414,
+      "step": 14019
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.253280272226548e-05,
+      "loss": 0.4308,
+      "step": 14020
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2532225971104768e-05,
+      "loss": 0.462,
+      "step": 14021
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2531649219944056e-05,
+      "loss": 0.5098,
+      "step": 14022
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2531072468783345e-05,
+      "loss": 0.4275,
+      "step": 14023
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2530495717622633e-05,
+      "loss": 0.4895,
+      "step": 14024
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.252991896646192e-05,
+      "loss": 0.5421,
+      "step": 14025
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.252934221530121e-05,
+      "loss": 0.4771,
+      "step": 14026
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2528765464140498e-05,
+      "loss": 0.627,
+      "step": 14027
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2528188712979785e-05,
+      "loss": 0.5182,
+      "step": 14028
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2527611961819075e-05,
+      "loss": 0.4758,
+      "step": 14029
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.2527035210658362e-05,
+      "loss": 0.5106,
+      "step": 14030
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 1.252645845949765e-05,
+      "loss": 0.4757,
+      "step": 14031
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.252588170833694e-05,
+      "loss": 0.516,
+      "step": 14032
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2525304957176227e-05,
+      "loss": 0.4133,
+      "step": 14033
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2524728206015515e-05,
+      "loss": 0.5633,
+      "step": 14034
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2524151454854804e-05,
+      "loss": 0.5104,
+      "step": 14035
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2523574703694092e-05,
+      "loss": 0.4365,
+      "step": 14036
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.252299795253338e-05,
+      "loss": 0.5019,
+      "step": 14037
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2522421201372669e-05,
+      "loss": 0.573,
+      "step": 14038
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2521844450211957e-05,
+      "loss": 0.3934,
+      "step": 14039
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2521267699051244e-05,
+      "loss": 0.3995,
+      "step": 14040
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2520690947890535e-05,
+      "loss": 0.4524,
+      "step": 14041
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2520114196729823e-05,
+      "loss": 0.5404,
+      "step": 14042
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2519537445569109e-05,
+      "loss": 0.5922,
+      "step": 14043
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2518960694408397e-05,
+      "loss": 0.5774,
+      "step": 14044
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2518383943247688e-05,
+      "loss": 0.5139,
+      "step": 14045
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2517807192086976e-05,
+      "loss": 0.3801,
+      "step": 14046
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2517230440926263e-05,
+      "loss": 0.4566,
+      "step": 14047
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2516653689765553e-05,
+      "loss": 0.5725,
+      "step": 14048
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.251607693860484e-05,
+      "loss": 0.4203,
+      "step": 14049
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2515500187444128e-05,
+      "loss": 0.5978,
+      "step": 14050
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2514923436283417e-05,
+      "loss": 0.6262,
+      "step": 14051
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2514346685122705e-05,
+      "loss": 0.5053,
+      "step": 14052
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2513769933961993e-05,
+      "loss": 0.5247,
+      "step": 14053
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2513193182801282e-05,
+      "loss": 0.5893,
+      "step": 14054
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.251261643164057e-05,
+      "loss": 0.4289,
+      "step": 14055
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2512039680479857e-05,
+      "loss": 0.6132,
+      "step": 14056
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2511462929319147e-05,
+      "loss": 0.4946,
+      "step": 14057
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2510886178158435e-05,
+      "loss": 0.503,
+      "step": 14058
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2510309426997722e-05,
+      "loss": 0.5614,
+      "step": 14059
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2509732675837012e-05,
+      "loss": 0.458,
+      "step": 14060
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.25091559246763e-05,
+      "loss": 0.5984,
+      "step": 14061
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2508579173515587e-05,
+      "loss": 0.5223,
+      "step": 14062
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2508002422354876e-05,
+      "loss": 0.4768,
+      "step": 14063
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2507425671194164e-05,
+      "loss": 0.4216,
+      "step": 14064
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2506848920033452e-05,
+      "loss": 0.4487,
+      "step": 14065
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2506272168872741e-05,
+      "loss": 0.4234,
+      "step": 14066
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2505695417712029e-05,
+      "loss": 0.564,
+      "step": 14067
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2505118666551316e-05,
+      "loss": 0.4537,
+      "step": 14068
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2504541915390606e-05,
+      "loss": 0.571,
+      "step": 14069
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2503965164229894e-05,
+      "loss": 0.3042,
+      "step": 14070
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2503388413069181e-05,
+      "loss": 0.5256,
+      "step": 14071
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2502811661908472e-05,
+      "loss": 0.4909,
+      "step": 14072
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.250223491074776e-05,
+      "loss": 0.5373,
+      "step": 14073
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2501658159587046e-05,
+      "loss": 0.5025,
+      "step": 14074
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2501081408426337e-05,
+      "loss": 0.4338,
+      "step": 14075
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2500504657265625e-05,
+      "loss": 0.4621,
+      "step": 14076
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2499927906104912e-05,
+      "loss": 0.6315,
+      "step": 14077
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.24993511549442e-05,
+      "loss": 0.607,
+      "step": 14078
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.249877440378349e-05,
+      "loss": 0.4846,
+      "step": 14079
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2498197652622777e-05,
+      "loss": 0.5401,
+      "step": 14080
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2497620901462065e-05,
+      "loss": 0.4983,
+      "step": 14081
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2497044150301354e-05,
+      "loss": 0.5161,
+      "step": 14082
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2496467399140642e-05,
+      "loss": 0.4131,
+      "step": 14083
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.249589064797993e-05,
+      "loss": 0.5463,
+      "step": 14084
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2495313896819219e-05,
+      "loss": 0.5294,
+      "step": 14085
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2494737145658507e-05,
+      "loss": 0.4613,
+      "step": 14086
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2494160394497794e-05,
+      "loss": 0.4774,
+      "step": 14087
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2493583643337084e-05,
+      "loss": 0.4463,
+      "step": 14088
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2493006892176371e-05,
+      "loss": 0.5525,
+      "step": 14089
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2492430141015659e-05,
+      "loss": 0.3961,
+      "step": 14090
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2491853389854948e-05,
+      "loss": 0.57,
+      "step": 14091
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2491276638694236e-05,
+      "loss": 0.596,
+      "step": 14092
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2490699887533524e-05,
+      "loss": 0.4245,
+      "step": 14093
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2490123136372813e-05,
+      "loss": 0.4657,
+      "step": 14094
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2489546385212101e-05,
+      "loss": 0.6132,
+      "step": 14095
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2488969634051389e-05,
+      "loss": 0.5251,
+      "step": 14096
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2488392882890678e-05,
+      "loss": 0.4496,
+      "step": 14097
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2487816131729966e-05,
+      "loss": 0.6029,
+      "step": 14098
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2487239380569253e-05,
+      "loss": 0.5034,
+      "step": 14099
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2486662629408544e-05,
+      "loss": 0.5025,
+      "step": 14100
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.248608587824783e-05,
+      "loss": 0.4912,
+      "step": 14101
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2485509127087118e-05,
+      "loss": 0.6389,
+      "step": 14102
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.248493237592641e-05,
+      "loss": 0.322,
+      "step": 14103
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2484355624765697e-05,
+      "loss": 0.4776,
+      "step": 14104
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2483778873604983e-05,
+      "loss": 0.4619,
+      "step": 14105
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2483202122444274e-05,
+      "loss": 0.5269,
+      "step": 14106
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2482625371283562e-05,
+      "loss": 0.456,
+      "step": 14107
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.248204862012285e-05,
+      "loss": 0.5037,
+      "step": 14108
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2481471868962139e-05,
+      "loss": 0.5084,
+      "step": 14109
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2480895117801426e-05,
+      "loss": 0.4139,
+      "step": 14110
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2480318366640714e-05,
+      "loss": 0.3739,
+      "step": 14111
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2479741615480002e-05,
+      "loss": 0.491,
+      "step": 14112
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2479164864319291e-05,
+      "loss": 0.4517,
+      "step": 14113
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2478588113158579e-05,
+      "loss": 0.5343,
+      "step": 14114
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2478011361997866e-05,
+      "loss": 0.495,
+      "step": 14115
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2477434610837156e-05,
+      "loss": 0.5278,
+      "step": 14116
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2476857859676444e-05,
+      "loss": 0.4424,
+      "step": 14117
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2476281108515731e-05,
+      "loss": 0.6505,
+      "step": 14118
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.247570435735502e-05,
+      "loss": 0.6539,
+      "step": 14119
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2475127606194308e-05,
+      "loss": 0.4065,
+      "step": 14120
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2474550855033596e-05,
+      "loss": 0.469,
+      "step": 14121
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2473974103872885e-05,
+      "loss": 0.4964,
+      "step": 14122
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2473397352712173e-05,
+      "loss": 0.4763,
+      "step": 14123
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.247282060155146e-05,
+      "loss": 0.4963,
+      "step": 14124
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.247224385039075e-05,
+      "loss": 0.5808,
+      "step": 14125
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2471667099230038e-05,
+      "loss": 0.5695,
+      "step": 14126
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2471090348069325e-05,
+      "loss": 0.6457,
+      "step": 14127
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2470513596908615e-05,
+      "loss": 0.4552,
+      "step": 14128
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2469936845747903e-05,
+      "loss": 0.4261,
+      "step": 14129
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.246936009458719e-05,
+      "loss": 0.5685,
+      "step": 14130
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2468783343426481e-05,
+      "loss": 0.4451,
+      "step": 14131
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2468206592265767e-05,
+      "loss": 0.5089,
+      "step": 14132
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2467629841105055e-05,
+      "loss": 0.4715,
+      "step": 14133
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2467053089944346e-05,
+      "loss": 0.5004,
+      "step": 14134
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2466476338783634e-05,
+      "loss": 0.486,
+      "step": 14135
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.246589958762292e-05,
+      "loss": 0.445,
+      "step": 14136
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.246532283646221e-05,
+      "loss": 0.4064,
+      "step": 14137
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2464746085301498e-05,
+      "loss": 0.5008,
+      "step": 14138
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2464169334140786e-05,
+      "loss": 0.3997,
+      "step": 14139
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2463592582980076e-05,
+      "loss": 0.5394,
+      "step": 14140
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2463015831819363e-05,
+      "loss": 0.4657,
+      "step": 14141
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2462439080658651e-05,
+      "loss": 0.5582,
+      "step": 14142
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.246186232949794e-05,
+      "loss": 0.5627,
+      "step": 14143
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2461285578337228e-05,
+      "loss": 0.5405,
+      "step": 14144
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2460708827176516e-05,
+      "loss": 0.4575,
+      "step": 14145
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2460132076015805e-05,
+      "loss": 0.4514,
+      "step": 14146
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2459555324855093e-05,
+      "loss": 0.5065,
+      "step": 14147
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.245897857369438e-05,
+      "loss": 0.4547,
+      "step": 14148
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2458401822533668e-05,
+      "loss": 0.6262,
+      "step": 14149
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2457825071372957e-05,
+      "loss": 0.5107,
+      "step": 14150
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2457248320212245e-05,
+      "loss": 0.6191,
+      "step": 14151
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2456671569051533e-05,
+      "loss": 0.577,
+      "step": 14152
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2456094817890822e-05,
+      "loss": 0.421,
+      "step": 14153
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.245551806673011e-05,
+      "loss": 0.4745,
+      "step": 14154
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2454941315569398e-05,
+      "loss": 0.5204,
+      "step": 14155
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2454364564408687e-05,
+      "loss": 0.4953,
+      "step": 14156
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2453787813247975e-05,
+      "loss": 0.5514,
+      "step": 14157
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2453211062087262e-05,
+      "loss": 0.4018,
+      "step": 14158
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2452634310926552e-05,
+      "loss": 0.4144,
+      "step": 14159
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.245205755976584e-05,
+      "loss": 0.436,
+      "step": 14160
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2451480808605127e-05,
+      "loss": 0.3836,
+      "step": 14161
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2450904057444418e-05,
+      "loss": 0.5548,
+      "step": 14162
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2450327306283704e-05,
+      "loss": 0.4523,
+      "step": 14163
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2449750555122992e-05,
+      "loss": 0.4555,
+      "step": 14164
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2449173803962283e-05,
+      "loss": 0.5132,
+      "step": 14165
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.244859705280157e-05,
+      "loss": 0.5418,
+      "step": 14166
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2448020301640857e-05,
+      "loss": 0.5351,
+      "step": 14167
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2447443550480148e-05,
+      "loss": 0.4752,
+      "step": 14168
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2446866799319435e-05,
+      "loss": 0.4764,
+      "step": 14169
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2446290048158723e-05,
+      "loss": 0.5096,
+      "step": 14170
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2445713296998012e-05,
+      "loss": 0.5317,
+      "step": 14171
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.24451365458373e-05,
+      "loss": 0.3721,
+      "step": 14172
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2444559794676588e-05,
+      "loss": 0.4958,
+      "step": 14173
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2443983043515877e-05,
+      "loss": 0.5195,
+      "step": 14174
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2443406292355165e-05,
+      "loss": 0.4582,
+      "step": 14175
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2442829541194453e-05,
+      "loss": 0.5528,
+      "step": 14176
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2442252790033742e-05,
+      "loss": 0.4983,
+      "step": 14177
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.244167603887303e-05,
+      "loss": 0.4826,
+      "step": 14178
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2441099287712317e-05,
+      "loss": 0.5616,
+      "step": 14179
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2440522536551607e-05,
+      "loss": 0.4316,
+      "step": 14180
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2439945785390894e-05,
+      "loss": 0.4129,
+      "step": 14181
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2439369034230182e-05,
+      "loss": 0.5481,
+      "step": 14182
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.243879228306947e-05,
+      "loss": 0.4322,
+      "step": 14183
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2438215531908759e-05,
+      "loss": 0.4444,
+      "step": 14184
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2437638780748047e-05,
+      "loss": 0.4655,
+      "step": 14185
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2437062029587334e-05,
+      "loss": 0.3662,
+      "step": 14186
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2436485278426624e-05,
+      "loss": 0.514,
+      "step": 14187
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2435908527265912e-05,
+      "loss": 0.4993,
+      "step": 14188
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.24353317761052e-05,
+      "loss": 0.4623,
+      "step": 14189
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2434755024944489e-05,
+      "loss": 0.5573,
+      "step": 14190
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2434178273783776e-05,
+      "loss": 0.6225,
+      "step": 14191
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2433601522623064e-05,
+      "loss": 0.5001,
+      "step": 14192
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2433024771462355e-05,
+      "loss": 0.5448,
+      "step": 14193
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2432448020301641e-05,
+      "loss": 0.6398,
+      "step": 14194
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2431871269140929e-05,
+      "loss": 0.2904,
+      "step": 14195
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.243129451798022e-05,
+      "loss": 0.524,
+      "step": 14196
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2430717766819507e-05,
+      "loss": 0.4849,
+      "step": 14197
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2430141015658793e-05,
+      "loss": 0.4866,
+      "step": 14198
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2429564264498085e-05,
+      "loss": 0.5641,
+      "step": 14199
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2428987513337372e-05,
+      "loss": 0.5129,
+      "step": 14200
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.242841076217666e-05,
+      "loss": 0.4039,
+      "step": 14201
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.242783401101595e-05,
+      "loss": 0.5135,
+      "step": 14202
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2427257259855237e-05,
+      "loss": 0.5504,
+      "step": 14203
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2426680508694525e-05,
+      "loss": 0.4747,
+      "step": 14204
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2426103757533814e-05,
+      "loss": 0.3245,
+      "step": 14205
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2425527006373102e-05,
+      "loss": 0.4593,
+      "step": 14206
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.242495025521239e-05,
+      "loss": 0.4909,
+      "step": 14207
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2424373504051679e-05,
+      "loss": 0.4482,
+      "step": 14208
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2423796752890966e-05,
+      "loss": 0.6483,
+      "step": 14209
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.2423220001730254e-05,
+      "loss": 0.3883,
+      "step": 14210
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2422643250569544e-05,
+      "loss": 0.5514,
+      "step": 14211
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2422066499408831e-05,
+      "loss": 0.4315,
+      "step": 14212
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2421489748248119e-05,
+      "loss": 0.5312,
+      "step": 14213
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2420912997087408e-05,
+      "loss": 0.624,
+      "step": 14214
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2420336245926696e-05,
+      "loss": 0.4233,
+      "step": 14215
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2419759494765984e-05,
+      "loss": 0.5365,
+      "step": 14216
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2419182743605271e-05,
+      "loss": 0.4476,
+      "step": 14217
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.241860599244456e-05,
+      "loss": 0.4839,
+      "step": 14218
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2418029241283848e-05,
+      "loss": 0.4461,
+      "step": 14219
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2417452490123136e-05,
+      "loss": 0.4891,
+      "step": 14220
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2416875738962425e-05,
+      "loss": 0.5117,
+      "step": 14221
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2416298987801713e-05,
+      "loss": 0.4946,
+      "step": 14222
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2415722236641e-05,
+      "loss": 0.5076,
+      "step": 14223
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2415145485480292e-05,
+      "loss": 0.5302,
+      "step": 14224
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2414568734319578e-05,
+      "loss": 0.4505,
+      "step": 14225
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2413991983158866e-05,
+      "loss": 0.509,
+      "step": 14226
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2413415231998157e-05,
+      "loss": 0.4542,
+      "step": 14227
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2412838480837444e-05,
+      "loss": 0.5315,
+      "step": 14228
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.241226172967673e-05,
+      "loss": 0.5088,
+      "step": 14229
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2411684978516021e-05,
+      "loss": 0.5304,
+      "step": 14230
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2411108227355309e-05,
+      "loss": 0.548,
+      "step": 14231
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2410531476194597e-05,
+      "loss": 0.4052,
+      "step": 14232
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2409954725033886e-05,
+      "loss": 0.5328,
+      "step": 14233
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2409377973873174e-05,
+      "loss": 0.4515,
+      "step": 14234
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2408801222712462e-05,
+      "loss": 0.5279,
+      "step": 14235
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2408224471551751e-05,
+      "loss": 0.543,
+      "step": 14236
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2407647720391039e-05,
+      "loss": 0.3751,
+      "step": 14237
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2407070969230326e-05,
+      "loss": 0.3956,
+      "step": 14238
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2406494218069616e-05,
+      "loss": 0.4882,
+      "step": 14239
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2405917466908903e-05,
+      "loss": 0.4378,
+      "step": 14240
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2405340715748191e-05,
+      "loss": 0.5271,
+      "step": 14241
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.240476396458748e-05,
+      "loss": 0.5446,
+      "step": 14242
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2404187213426768e-05,
+      "loss": 0.5115,
+      "step": 14243
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2403610462266056e-05,
+      "loss": 0.4858,
+      "step": 14244
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2403033711105345e-05,
+      "loss": 0.5989,
+      "step": 14245
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2402456959944633e-05,
+      "loss": 0.4892,
+      "step": 14246
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.240188020878392e-05,
+      "loss": 0.4843,
+      "step": 14247
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.240130345762321e-05,
+      "loss": 0.3822,
+      "step": 14248
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2400726706462498e-05,
+      "loss": 0.5177,
+      "step": 14249
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2400149955301785e-05,
+      "loss": 0.4984,
+      "step": 14250
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2399573204141073e-05,
+      "loss": 0.6982,
+      "step": 14251
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2398996452980362e-05,
+      "loss": 0.4914,
+      "step": 14252
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.239841970181965e-05,
+      "loss": 0.5044,
+      "step": 14253
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2397842950658938e-05,
+      "loss": 0.5006,
+      "step": 14254
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2397266199498229e-05,
+      "loss": 0.4235,
+      "step": 14255
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2396689448337515e-05,
+      "loss": 0.5912,
+      "step": 14256
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2396112697176802e-05,
+      "loss": 0.4441,
+      "step": 14257
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2395535946016094e-05,
+      "loss": 0.5872,
+      "step": 14258
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2394959194855381e-05,
+      "loss": 0.4704,
+      "step": 14259
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2394382443694667e-05,
+      "loss": 0.5633,
+      "step": 14260
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2393805692533958e-05,
+      "loss": 0.5524,
+      "step": 14261
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2393228941373246e-05,
+      "loss": 0.4973,
+      "step": 14262
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2392652190212534e-05,
+      "loss": 0.5453,
+      "step": 14263
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2392075439051823e-05,
+      "loss": 0.4693,
+      "step": 14264
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.239149868789111e-05,
+      "loss": 0.4564,
+      "step": 14265
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2390921936730398e-05,
+      "loss": 0.3967,
+      "step": 14266
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2390345185569688e-05,
+      "loss": 0.5121,
+      "step": 14267
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2389768434408975e-05,
+      "loss": 0.5105,
+      "step": 14268
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2389191683248263e-05,
+      "loss": 0.5553,
+      "step": 14269
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2388614932087553e-05,
+      "loss": 0.5929,
+      "step": 14270
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.238803818092684e-05,
+      "loss": 0.6274,
+      "step": 14271
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2387461429766128e-05,
+      "loss": 0.5235,
+      "step": 14272
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2386884678605417e-05,
+      "loss": 0.4688,
+      "step": 14273
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2386307927444705e-05,
+      "loss": 0.5121,
+      "step": 14274
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2385731176283993e-05,
+      "loss": 0.5138,
+      "step": 14275
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2385154425123282e-05,
+      "loss": 0.4577,
+      "step": 14276
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.238457767396257e-05,
+      "loss": 0.6713,
+      "step": 14277
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2384000922801857e-05,
+      "loss": 0.3386,
+      "step": 14278
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2383424171641147e-05,
+      "loss": 0.6035,
+      "step": 14279
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2382847420480434e-05,
+      "loss": 0.4573,
+      "step": 14280
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2382270669319722e-05,
+      "loss": 0.6259,
+      "step": 14281
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2381693918159012e-05,
+      "loss": 0.5905,
+      "step": 14282
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.23811171669983e-05,
+      "loss": 0.2898,
+      "step": 14283
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2380540415837587e-05,
+      "loss": 0.527,
+      "step": 14284
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2379963664676878e-05,
+      "loss": 0.5547,
+      "step": 14285
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2379386913516166e-05,
+      "loss": 0.4151,
+      "step": 14286
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2378810162355452e-05,
+      "loss": 0.42,
+      "step": 14287
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.237823341119474e-05,
+      "loss": 0.5394,
+      "step": 14288
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.237765666003403e-05,
+      "loss": 0.3921,
+      "step": 14289
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2377079908873318e-05,
+      "loss": 0.5178,
+      "step": 14290
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2376503157712604e-05,
+      "loss": 0.5177,
+      "step": 14291
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2375926406551895e-05,
+      "loss": 0.512,
+      "step": 14292
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2375349655391183e-05,
+      "loss": 0.57,
+      "step": 14293
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.237477290423047e-05,
+      "loss": 0.5139,
+      "step": 14294
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.237419615306976e-05,
+      "loss": 0.4714,
+      "step": 14295
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2373619401909048e-05,
+      "loss": 0.5529,
+      "step": 14296
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2373042650748335e-05,
+      "loss": 0.4724,
+      "step": 14297
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2372465899587625e-05,
+      "loss": 0.4753,
+      "step": 14298
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2371889148426912e-05,
+      "loss": 0.5236,
+      "step": 14299
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.23713123972662e-05,
+      "loss": 0.4112,
+      "step": 14300
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.237073564610549e-05,
+      "loss": 0.7735,
+      "step": 14301
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2370158894944777e-05,
+      "loss": 0.5377,
+      "step": 14302
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2369582143784065e-05,
+      "loss": 0.5893,
+      "step": 14303
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2369005392623354e-05,
+      "loss": 0.6044,
+      "step": 14304
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2368428641462642e-05,
+      "loss": 0.4746,
+      "step": 14305
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.236785189030193e-05,
+      "loss": 0.557,
+      "step": 14306
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2367275139141219e-05,
+      "loss": 0.4992,
+      "step": 14307
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2366698387980507e-05,
+      "loss": 0.4016,
+      "step": 14308
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2366121636819794e-05,
+      "loss": 0.4311,
+      "step": 14309
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2365544885659084e-05,
+      "loss": 0.5091,
+      "step": 14310
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2364968134498371e-05,
+      "loss": 0.4487,
+      "step": 14311
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2364391383337659e-05,
+      "loss": 0.3927,
+      "step": 14312
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2363814632176948e-05,
+      "loss": 0.5944,
+      "step": 14313
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2363237881016236e-05,
+      "loss": 0.5999,
+      "step": 14314
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2362661129855524e-05,
+      "loss": 0.5112,
+      "step": 14315
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2362084378694815e-05,
+      "loss": 0.458,
+      "step": 14316
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2361507627534103e-05,
+      "loss": 0.3926,
+      "step": 14317
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2360930876373389e-05,
+      "loss": 0.4532,
+      "step": 14318
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.236035412521268e-05,
+      "loss": 0.4655,
+      "step": 14319
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2359777374051967e-05,
+      "loss": 0.6132,
+      "step": 14320
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2359200622891255e-05,
+      "loss": 0.4234,
+      "step": 14321
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2358623871730541e-05,
+      "loss": 0.4504,
+      "step": 14322
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2358047120569832e-05,
+      "loss": 0.5317,
+      "step": 14323
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.235747036940912e-05,
+      "loss": 0.45,
+      "step": 14324
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2356893618248407e-05,
+      "loss": 0.4317,
+      "step": 14325
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2356316867087697e-05,
+      "loss": 0.4612,
+      "step": 14326
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2355740115926984e-05,
+      "loss": 0.5594,
+      "step": 14327
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2355163364766272e-05,
+      "loss": 0.3767,
+      "step": 14328
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2354586613605562e-05,
+      "loss": 0.5787,
+      "step": 14329
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.235400986244485e-05,
+      "loss": 0.4117,
+      "step": 14330
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2353433111284137e-05,
+      "loss": 0.5329,
+      "step": 14331
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2352856360123426e-05,
+      "loss": 0.423,
+      "step": 14332
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2352279608962714e-05,
+      "loss": 0.5126,
+      "step": 14333
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2351702857802002e-05,
+      "loss": 0.4244,
+      "step": 14334
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2351126106641291e-05,
+      "loss": 0.5242,
+      "step": 14335
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2350549355480579e-05,
+      "loss": 0.4724,
+      "step": 14336
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2349972604319866e-05,
+      "loss": 0.5574,
+      "step": 14337
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2349395853159156e-05,
+      "loss": 0.4623,
+      "step": 14338
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2348819101998443e-05,
+      "loss": 0.5347,
+      "step": 14339
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2348242350837731e-05,
+      "loss": 0.5379,
+      "step": 14340
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.234766559967702e-05,
+      "loss": 0.5822,
+      "step": 14341
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2347088848516308e-05,
+      "loss": 0.4161,
+      "step": 14342
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2346512097355596e-05,
+      "loss": 0.5127,
+      "step": 14343
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2345935346194885e-05,
+      "loss": 0.5096,
+      "step": 14344
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2345358595034173e-05,
+      "loss": 0.4938,
+      "step": 14345
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.234478184387346e-05,
+      "loss": 0.5622,
+      "step": 14346
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2344205092712752e-05,
+      "loss": 0.3845,
+      "step": 14347
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.234362834155204e-05,
+      "loss": 0.4218,
+      "step": 14348
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2343051590391325e-05,
+      "loss": 0.4866,
+      "step": 14349
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2342474839230616e-05,
+      "loss": 0.4701,
+      "step": 14350
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2341898088069904e-05,
+      "loss": 0.4038,
+      "step": 14351
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2341321336909192e-05,
+      "loss": 0.477,
+      "step": 14352
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2340744585748481e-05,
+      "loss": 0.4986,
+      "step": 14353
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2340167834587769e-05,
+      "loss": 0.4211,
+      "step": 14354
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2339591083427057e-05,
+      "loss": 0.5558,
+      "step": 14355
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2339014332266344e-05,
+      "loss": 0.4125,
+      "step": 14356
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2338437581105634e-05,
+      "loss": 0.4187,
+      "step": 14357
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2337860829944921e-05,
+      "loss": 0.4697,
+      "step": 14358
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2337284078784209e-05,
+      "loss": 0.5068,
+      "step": 14359
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2336707327623498e-05,
+      "loss": 0.4521,
+      "step": 14360
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2336130576462786e-05,
+      "loss": 0.5403,
+      "step": 14361
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2335553825302074e-05,
+      "loss": 0.3797,
+      "step": 14362
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2334977074141363e-05,
+      "loss": 0.5475,
+      "step": 14363
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.233440032298065e-05,
+      "loss": 0.5465,
+      "step": 14364
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2333823571819939e-05,
+      "loss": 0.492,
+      "step": 14365
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2333246820659228e-05,
+      "loss": 0.5616,
+      "step": 14366
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2332670069498516e-05,
+      "loss": 0.4974,
+      "step": 14367
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2332093318337803e-05,
+      "loss": 0.5576,
+      "step": 14368
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2331516567177093e-05,
+      "loss": 0.4433,
+      "step": 14369
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.233093981601638e-05,
+      "loss": 0.6027,
+      "step": 14370
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2330363064855668e-05,
+      "loss": 0.3965,
+      "step": 14371
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2329786313694957e-05,
+      "loss": 0.4535,
+      "step": 14372
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2329209562534245e-05,
+      "loss": 0.4219,
+      "step": 14373
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2328632811373533e-05,
+      "loss": 0.5274,
+      "step": 14374
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2328056060212822e-05,
+      "loss": 0.6338,
+      "step": 14375
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.232747930905211e-05,
+      "loss": 0.486,
+      "step": 14376
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2326902557891398e-05,
+      "loss": 0.4646,
+      "step": 14377
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2326325806730689e-05,
+      "loss": 0.5825,
+      "step": 14378
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2325749055569976e-05,
+      "loss": 0.4712,
+      "step": 14379
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2325172304409262e-05,
+      "loss": 0.5384,
+      "step": 14380
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2324595553248553e-05,
+      "loss": 0.4943,
+      "step": 14381
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2324018802087841e-05,
+      "loss": 0.4645,
+      "step": 14382
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2323442050927129e-05,
+      "loss": 0.6168,
+      "step": 14383
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2322865299766418e-05,
+      "loss": 0.6763,
+      "step": 14384
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2322288548605706e-05,
+      "loss": 0.5224,
+      "step": 14385
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2321711797444993e-05,
+      "loss": 0.4629,
+      "step": 14386
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2321135046284283e-05,
+      "loss": 0.5931,
+      "step": 14387
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.232055829512357e-05,
+      "loss": 0.5105,
+      "step": 14388
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 1.2319981543962858e-05,
+      "loss": 0.5004,
+      "step": 14389
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2319404792802148e-05,
+      "loss": 0.4372,
+      "step": 14390
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2318828041641435e-05,
+      "loss": 0.5138,
+      "step": 14391
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2318251290480723e-05,
+      "loss": 0.4868,
+      "step": 14392
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.231767453932001e-05,
+      "loss": 0.4178,
+      "step": 14393
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.23170977881593e-05,
+      "loss": 0.4372,
+      "step": 14394
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2316521036998588e-05,
+      "loss": 0.3242,
+      "step": 14395
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2315944285837875e-05,
+      "loss": 0.55,
+      "step": 14396
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2315367534677165e-05,
+      "loss": 0.5478,
+      "step": 14397
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2314790783516452e-05,
+      "loss": 0.4643,
+      "step": 14398
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.231421403235574e-05,
+      "loss": 0.5062,
+      "step": 14399
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.231363728119503e-05,
+      "loss": 0.5422,
+      "step": 14400
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2313060530034317e-05,
+      "loss": 0.4754,
+      "step": 14401
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2312483778873605e-05,
+      "loss": 0.5655,
+      "step": 14402
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2311907027712894e-05,
+      "loss": 0.5204,
+      "step": 14403
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2311330276552182e-05,
+      "loss": 0.5713,
+      "step": 14404
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.231075352539147e-05,
+      "loss": 0.5004,
+      "step": 14405
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2310176774230759e-05,
+      "loss": 0.4439,
+      "step": 14406
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2309600023070047e-05,
+      "loss": 0.4117,
+      "step": 14407
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2309023271909334e-05,
+      "loss": 0.5387,
+      "step": 14408
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2308446520748625e-05,
+      "loss": 0.4113,
+      "step": 14409
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2307869769587913e-05,
+      "loss": 0.4322,
+      "step": 14410
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2307293018427199e-05,
+      "loss": 0.412,
+      "step": 14411
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.230671626726649e-05,
+      "loss": 0.5338,
+      "step": 14412
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2306139516105778e-05,
+      "loss": 0.4659,
+      "step": 14413
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2305562764945066e-05,
+      "loss": 0.5023,
+      "step": 14414
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2304986013784355e-05,
+      "loss": 0.3854,
+      "step": 14415
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2304409262623643e-05,
+      "loss": 0.4749,
+      "step": 14416
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.230383251146293e-05,
+      "loss": 0.5616,
+      "step": 14417
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.230325576030222e-05,
+      "loss": 0.4886,
+      "step": 14418
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2302679009141507e-05,
+      "loss": 0.4527,
+      "step": 14419
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2302102257980795e-05,
+      "loss": 0.5254,
+      "step": 14420
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2301525506820084e-05,
+      "loss": 0.4148,
+      "step": 14421
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2300948755659372e-05,
+      "loss": 0.5189,
+      "step": 14422
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.230037200449866e-05,
+      "loss": 0.3743,
+      "step": 14423
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.229979525333795e-05,
+      "loss": 0.5403,
+      "step": 14424
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2299218502177237e-05,
+      "loss": 0.5233,
+      "step": 14425
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2298641751016525e-05,
+      "loss": 0.4267,
+      "step": 14426
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2298064999855812e-05,
+      "loss": 0.4674,
+      "step": 14427
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2297488248695102e-05,
+      "loss": 0.4912,
+      "step": 14428
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.229691149753439e-05,
+      "loss": 0.4733,
+      "step": 14429
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2296334746373677e-05,
+      "loss": 0.497,
+      "step": 14430
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2295757995212966e-05,
+      "loss": 0.5148,
+      "step": 14431
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2295181244052254e-05,
+      "loss": 0.4669,
+      "step": 14432
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2294604492891542e-05,
+      "loss": 0.4374,
+      "step": 14433
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2294027741730831e-05,
+      "loss": 0.5123,
+      "step": 14434
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2293450990570119e-05,
+      "loss": 0.5207,
+      "step": 14435
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2292874239409407e-05,
+      "loss": 0.4659,
+      "step": 14436
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2292297488248696e-05,
+      "loss": 0.4389,
+      "step": 14437
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2291720737087984e-05,
+      "loss": 0.4192,
+      "step": 14438
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2291143985927271e-05,
+      "loss": 0.4975,
+      "step": 14439
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2290567234766562e-05,
+      "loss": 0.5379,
+      "step": 14440
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.228999048360585e-05,
+      "loss": 0.5288,
+      "step": 14441
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2289413732445136e-05,
+      "loss": 0.5342,
+      "step": 14442
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2288836981284427e-05,
+      "loss": 0.5934,
+      "step": 14443
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2288260230123715e-05,
+      "loss": 0.54,
+      "step": 14444
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2287683478963002e-05,
+      "loss": 0.4146,
+      "step": 14445
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2287106727802292e-05,
+      "loss": 0.495,
+      "step": 14446
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.228652997664158e-05,
+      "loss": 0.3761,
+      "step": 14447
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2285953225480867e-05,
+      "loss": 0.445,
+      "step": 14448
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2285376474320157e-05,
+      "loss": 0.5405,
+      "step": 14449
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2284799723159444e-05,
+      "loss": 0.4395,
+      "step": 14450
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2284222971998732e-05,
+      "loss": 0.4808,
+      "step": 14451
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2283646220838021e-05,
+      "loss": 0.381,
+      "step": 14452
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2283069469677309e-05,
+      "loss": 0.5905,
+      "step": 14453
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2282492718516597e-05,
+      "loss": 0.4079,
+      "step": 14454
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2281915967355886e-05,
+      "loss": 0.5029,
+      "step": 14455
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2281339216195174e-05,
+      "loss": 0.5499,
+      "step": 14456
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2280762465034461e-05,
+      "loss": 0.546,
+      "step": 14457
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.228018571387375e-05,
+      "loss": 0.4743,
+      "step": 14458
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2279608962713039e-05,
+      "loss": 0.5607,
+      "step": 14459
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2279032211552326e-05,
+      "loss": 0.585,
+      "step": 14460
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2278455460391614e-05,
+      "loss": 0.4337,
+      "step": 14461
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2277878709230903e-05,
+      "loss": 0.3283,
+      "step": 14462
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2277301958070191e-05,
+      "loss": 0.3889,
+      "step": 14463
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2276725206909479e-05,
+      "loss": 0.4816,
+      "step": 14464
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2276148455748768e-05,
+      "loss": 0.6302,
+      "step": 14465
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2275571704588056e-05,
+      "loss": 0.6065,
+      "step": 14466
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2274994953427343e-05,
+      "loss": 0.4692,
+      "step": 14467
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2274418202266634e-05,
+      "loss": 0.4647,
+      "step": 14468
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.227384145110592e-05,
+      "loss": 0.4434,
+      "step": 14469
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2273264699945208e-05,
+      "loss": 0.4857,
+      "step": 14470
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.22726879487845e-05,
+      "loss": 0.4242,
+      "step": 14471
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2272111197623787e-05,
+      "loss": 0.4586,
+      "step": 14472
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2271534446463073e-05,
+      "loss": 0.5746,
+      "step": 14473
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2270957695302364e-05,
+      "loss": 0.4358,
+      "step": 14474
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2270380944141652e-05,
+      "loss": 0.4016,
+      "step": 14475
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.226980419298094e-05,
+      "loss": 0.4663,
+      "step": 14476
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2269227441820229e-05,
+      "loss": 0.5157,
+      "step": 14477
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2268650690659516e-05,
+      "loss": 0.4366,
+      "step": 14478
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2268073939498804e-05,
+      "loss": 0.3994,
+      "step": 14479
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2267497188338093e-05,
+      "loss": 0.4801,
+      "step": 14480
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2266920437177381e-05,
+      "loss": 0.4035,
+      "step": 14481
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2266343686016669e-05,
+      "loss": 0.573,
+      "step": 14482
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2265766934855958e-05,
+      "loss": 0.441,
+      "step": 14483
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2265190183695246e-05,
+      "loss": 0.5253,
+      "step": 14484
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2264613432534534e-05,
+      "loss": 0.5654,
+      "step": 14485
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2264036681373823e-05,
+      "loss": 0.5639,
+      "step": 14486
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.226345993021311e-05,
+      "loss": 0.4038,
+      "step": 14487
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2262883179052398e-05,
+      "loss": 0.5685,
+      "step": 14488
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2262306427891688e-05,
+      "loss": 0.5369,
+      "step": 14489
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2261729676730975e-05,
+      "loss": 0.5515,
+      "step": 14490
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2261152925570263e-05,
+      "loss": 0.5081,
+      "step": 14491
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2260576174409552e-05,
+      "loss": 0.5541,
+      "step": 14492
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.225999942324884e-05,
+      "loss": 0.5333,
+      "step": 14493
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2259422672088128e-05,
+      "loss": 0.5574,
+      "step": 14494
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2258845920927416e-05,
+      "loss": 0.5282,
+      "step": 14495
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2258269169766705e-05,
+      "loss": 0.4667,
+      "step": 14496
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2257692418605993e-05,
+      "loss": 0.4948,
+      "step": 14497
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.225711566744528e-05,
+      "loss": 0.4794,
+      "step": 14498
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2256538916284571e-05,
+      "loss": 0.5498,
+      "step": 14499
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2255962165123857e-05,
+      "loss": 0.55,
+      "step": 14500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2255385413963145e-05,
+      "loss": 0.5653,
+      "step": 14501
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2254808662802436e-05,
+      "loss": 0.583,
+      "step": 14502
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2254231911641724e-05,
+      "loss": 0.5011,
+      "step": 14503
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.225365516048101e-05,
+      "loss": 0.4301,
+      "step": 14504
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.22530784093203e-05,
+      "loss": 0.57,
+      "step": 14505
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2252501658159589e-05,
+      "loss": 0.3497,
+      "step": 14506
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2251924906998876e-05,
+      "loss": 0.4886,
+      "step": 14507
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2251348155838166e-05,
+      "loss": 0.4582,
+      "step": 14508
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2250771404677453e-05,
+      "loss": 0.4072,
+      "step": 14509
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2250194653516741e-05,
+      "loss": 0.5765,
+      "step": 14510
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.224961790235603e-05,
+      "loss": 0.4337,
+      "step": 14511
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2249041151195318e-05,
+      "loss": 0.5244,
+      "step": 14512
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2248464400034606e-05,
+      "loss": 0.5734,
+      "step": 14513
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2247887648873895e-05,
+      "loss": 0.4542,
+      "step": 14514
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2247310897713183e-05,
+      "loss": 0.5672,
+      "step": 14515
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.224673414655247e-05,
+      "loss": 0.5678,
+      "step": 14516
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.224615739539176e-05,
+      "loss": 0.5472,
+      "step": 14517
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2245580644231048e-05,
+      "loss": 0.3991,
+      "step": 14518
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2245003893070335e-05,
+      "loss": 0.5913,
+      "step": 14519
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2244427141909625e-05,
+      "loss": 0.464,
+      "step": 14520
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2243850390748912e-05,
+      "loss": 0.4936,
+      "step": 14521
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.22432736395882e-05,
+      "loss": 0.4596,
+      "step": 14522
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.224269688842749e-05,
+      "loss": 0.4405,
+      "step": 14523
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2242120137266777e-05,
+      "loss": 0.472,
+      "step": 14524
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2241543386106065e-05,
+      "loss": 0.4697,
+      "step": 14525
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2240966634945354e-05,
+      "loss": 0.3671,
+      "step": 14526
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2240389883784642e-05,
+      "loss": 0.4286,
+      "step": 14527
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.223981313262393e-05,
+      "loss": 0.5337,
+      "step": 14528
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.223923638146322e-05,
+      "loss": 0.4867,
+      "step": 14529
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2238659630302508e-05,
+      "loss": 0.481,
+      "step": 14530
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2238082879141794e-05,
+      "loss": 0.3907,
+      "step": 14531
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2237506127981082e-05,
+      "loss": 0.5927,
+      "step": 14532
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2236929376820373e-05,
+      "loss": 0.5449,
+      "step": 14533
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.223635262565966e-05,
+      "loss": 0.3807,
+      "step": 14534
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2235775874498947e-05,
+      "loss": 0.4227,
+      "step": 14535
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2235199123338238e-05,
+      "loss": 0.46,
+      "step": 14536
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2234622372177525e-05,
+      "loss": 0.4254,
+      "step": 14537
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2234045621016813e-05,
+      "loss": 0.6652,
+      "step": 14538
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2233468869856102e-05,
+      "loss": 0.3819,
+      "step": 14539
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.223289211869539e-05,
+      "loss": 0.4436,
+      "step": 14540
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2232315367534678e-05,
+      "loss": 0.581,
+      "step": 14541
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2231738616373967e-05,
+      "loss": 0.5497,
+      "step": 14542
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2231161865213255e-05,
+      "loss": 0.522,
+      "step": 14543
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2230585114052543e-05,
+      "loss": 0.487,
+      "step": 14544
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2230008362891832e-05,
+      "loss": 0.4845,
+      "step": 14545
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.222943161173112e-05,
+      "loss": 0.487,
+      "step": 14546
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2228854860570407e-05,
+      "loss": 0.5526,
+      "step": 14547
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2228278109409697e-05,
+      "loss": 0.4939,
+      "step": 14548
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2227701358248984e-05,
+      "loss": 0.5422,
+      "step": 14549
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2227124607088272e-05,
+      "loss": 0.3643,
+      "step": 14550
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2226547855927561e-05,
+      "loss": 0.4831,
+      "step": 14551
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2225971104766849e-05,
+      "loss": 0.5156,
+      "step": 14552
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2225394353606137e-05,
+      "loss": 0.4602,
+      "step": 14553
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2224817602445426e-05,
+      "loss": 0.5704,
+      "step": 14554
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2224240851284714e-05,
+      "loss": 0.6192,
+      "step": 14555
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2223664100124002e-05,
+      "loss": 0.4885,
+      "step": 14556
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2223087348963291e-05,
+      "loss": 0.5461,
+      "step": 14557
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2222510597802579e-05,
+      "loss": 0.4328,
+      "step": 14558
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2221933846641866e-05,
+      "loss": 0.5781,
+      "step": 14559
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2221357095481157e-05,
+      "loss": 0.3922,
+      "step": 14560
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2220780344320445e-05,
+      "loss": 0.5151,
+      "step": 14561
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2220203593159731e-05,
+      "loss": 0.5123,
+      "step": 14562
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2219626841999022e-05,
+      "loss": 0.5654,
+      "step": 14563
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.221905009083831e-05,
+      "loss": 0.3906,
+      "step": 14564
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2218473339677598e-05,
+      "loss": 0.4379,
+      "step": 14565
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2217896588516883e-05,
+      "loss": 0.4195,
+      "step": 14566
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2217319837356175e-05,
+      "loss": 0.5259,
+      "step": 14567
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 1.2216743086195462e-05,
+      "loss": 0.4958,
+      "step": 14568
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.221616633503475e-05,
+      "loss": 0.6069,
+      "step": 14569
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.221558958387404e-05,
+      "loss": 0.3702,
+      "step": 14570
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2215012832713327e-05,
+      "loss": 0.5411,
+      "step": 14571
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2214436081552615e-05,
+      "loss": 0.5226,
+      "step": 14572
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2213859330391904e-05,
+      "loss": 0.408,
+      "step": 14573
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2213282579231192e-05,
+      "loss": 0.3732,
+      "step": 14574
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.221270582807048e-05,
+      "loss": 0.4099,
+      "step": 14575
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2212129076909769e-05,
+      "loss": 0.5661,
+      "step": 14576
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2211552325749057e-05,
+      "loss": 0.4242,
+      "step": 14577
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2210975574588344e-05,
+      "loss": 0.6928,
+      "step": 14578
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2210398823427634e-05,
+      "loss": 0.6264,
+      "step": 14579
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2209822072266921e-05,
+      "loss": 0.5345,
+      "step": 14580
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2209245321106209e-05,
+      "loss": 0.5323,
+      "step": 14581
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2208668569945498e-05,
+      "loss": 0.4501,
+      "step": 14582
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2208091818784786e-05,
+      "loss": 0.5167,
+      "step": 14583
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2207515067624074e-05,
+      "loss": 0.7062,
+      "step": 14584
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2206938316463363e-05,
+      "loss": 0.4931,
+      "step": 14585
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.220636156530265e-05,
+      "loss": 0.553,
+      "step": 14586
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2205784814141938e-05,
+      "loss": 0.7561,
+      "step": 14587
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2205208062981228e-05,
+      "loss": 0.4216,
+      "step": 14588
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2204631311820515e-05,
+      "loss": 0.3891,
+      "step": 14589
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2204054560659803e-05,
+      "loss": 0.5052,
+      "step": 14590
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2203477809499094e-05,
+      "loss": 0.4784,
+      "step": 14591
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2202901058338382e-05,
+      "loss": 0.5037,
+      "step": 14592
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2202324307177668e-05,
+      "loss": 0.3712,
+      "step": 14593
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2201747556016959e-05,
+      "loss": 0.4842,
+      "step": 14594
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2201170804856247e-05,
+      "loss": 0.4456,
+      "step": 14595
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2200594053695534e-05,
+      "loss": 0.5235,
+      "step": 14596
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2200017302534824e-05,
+      "loss": 0.51,
+      "step": 14597
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2199440551374111e-05,
+      "loss": 0.5905,
+      "step": 14598
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2198863800213399e-05,
+      "loss": 0.5264,
+      "step": 14599
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2198287049052687e-05,
+      "loss": 0.5228,
+      "step": 14600
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2197710297891976e-05,
+      "loss": 0.4611,
+      "step": 14601
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2197133546731264e-05,
+      "loss": 0.4431,
+      "step": 14602
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2196556795570552e-05,
+      "loss": 0.5389,
+      "step": 14603
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2195980044409841e-05,
+      "loss": 0.4501,
+      "step": 14604
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2195403293249129e-05,
+      "loss": 0.4738,
+      "step": 14605
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2194826542088416e-05,
+      "loss": 0.5591,
+      "step": 14606
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2194249790927706e-05,
+      "loss": 0.4963,
+      "step": 14607
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2193673039766993e-05,
+      "loss": 0.5108,
+      "step": 14608
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2193096288606281e-05,
+      "loss": 0.523,
+      "step": 14609
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.219251953744557e-05,
+      "loss": 0.5522,
+      "step": 14610
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2191942786284858e-05,
+      "loss": 0.5876,
+      "step": 14611
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2191366035124146e-05,
+      "loss": 0.4973,
+      "step": 14612
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2190789283963435e-05,
+      "loss": 0.4132,
+      "step": 14613
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2190212532802723e-05,
+      "loss": 0.5287,
+      "step": 14614
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.218963578164201e-05,
+      "loss": 0.5106,
+      "step": 14615
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.21890590304813e-05,
+      "loss": 0.4127,
+      "step": 14616
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2188482279320588e-05,
+      "loss": 0.4935,
+      "step": 14617
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2187905528159875e-05,
+      "loss": 0.5336,
+      "step": 14618
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2187328776999165e-05,
+      "loss": 0.5762,
+      "step": 14619
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2186752025838452e-05,
+      "loss": 0.3097,
+      "step": 14620
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.218617527467774e-05,
+      "loss": 0.4833,
+      "step": 14621
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2185598523517031e-05,
+      "loss": 0.467,
+      "step": 14622
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2185021772356319e-05,
+      "loss": 0.4798,
+      "step": 14623
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2184445021195605e-05,
+      "loss": 0.5285,
+      "step": 14624
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2183868270034896e-05,
+      "loss": 0.4786,
+      "step": 14625
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2183291518874184e-05,
+      "loss": 0.4786,
+      "step": 14626
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2182714767713471e-05,
+      "loss": 0.5703,
+      "step": 14627
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.218213801655276e-05,
+      "loss": 0.4613,
+      "step": 14628
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2181561265392048e-05,
+      "loss": 0.411,
+      "step": 14629
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2180984514231336e-05,
+      "loss": 0.5749,
+      "step": 14630
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2180407763070625e-05,
+      "loss": 0.585,
+      "step": 14631
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2179831011909913e-05,
+      "loss": 0.4741,
+      "step": 14632
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.21792542607492e-05,
+      "loss": 0.4614,
+      "step": 14633
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2178677509588488e-05,
+      "loss": 0.4874,
+      "step": 14634
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2178100758427778e-05,
+      "loss": 0.5164,
+      "step": 14635
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2177524007267065e-05,
+      "loss": 0.5297,
+      "step": 14636
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2176947256106353e-05,
+      "loss": 0.5782,
+      "step": 14637
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2176370504945643e-05,
+      "loss": 0.5328,
+      "step": 14638
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.217579375378493e-05,
+      "loss": 0.4655,
+      "step": 14639
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2175217002624218e-05,
+      "loss": 0.4789,
+      "step": 14640
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2174640251463507e-05,
+      "loss": 0.4983,
+      "step": 14641
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2174063500302795e-05,
+      "loss": 0.3936,
+      "step": 14642
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2173486749142083e-05,
+      "loss": 0.5261,
+      "step": 14643
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2172909997981372e-05,
+      "loss": 0.5295,
+      "step": 14644
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.217233324682066e-05,
+      "loss": 0.5578,
+      "step": 14645
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2171756495659947e-05,
+      "loss": 0.4,
+      "step": 14646
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2171179744499237e-05,
+      "loss": 0.3997,
+      "step": 14647
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2170602993338524e-05,
+      "loss": 0.4828,
+      "step": 14648
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2170026242177812e-05,
+      "loss": 0.4303,
+      "step": 14649
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2169449491017102e-05,
+      "loss": 0.526,
+      "step": 14650
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.216887273985639e-05,
+      "loss": 0.3836,
+      "step": 14651
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2168295988695677e-05,
+      "loss": 0.4953,
+      "step": 14652
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2167719237534968e-05,
+      "loss": 0.3699,
+      "step": 14653
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2167142486374256e-05,
+      "loss": 0.4843,
+      "step": 14654
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2166565735213542e-05,
+      "loss": 0.4563,
+      "step": 14655
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2165988984052833e-05,
+      "loss": 0.52,
+      "step": 14656
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.216541223289212e-05,
+      "loss": 0.5492,
+      "step": 14657
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2164835481731408e-05,
+      "loss": 0.3886,
+      "step": 14658
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2164258730570698e-05,
+      "loss": 0.5213,
+      "step": 14659
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2163681979409985e-05,
+      "loss": 0.5151,
+      "step": 14660
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2163105228249273e-05,
+      "loss": 0.4267,
+      "step": 14661
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2162528477088562e-05,
+      "loss": 0.4146,
+      "step": 14662
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.216195172592785e-05,
+      "loss": 0.3991,
+      "step": 14663
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2161374974767138e-05,
+      "loss": 0.4492,
+      "step": 14664
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2160798223606427e-05,
+      "loss": 0.4968,
+      "step": 14665
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2160221472445715e-05,
+      "loss": 0.5521,
+      "step": 14666
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2159644721285002e-05,
+      "loss": 0.5056,
+      "step": 14667
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2159067970124292e-05,
+      "loss": 0.4665,
+      "step": 14668
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.215849121896358e-05,
+      "loss": 0.4659,
+      "step": 14669
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2157914467802867e-05,
+      "loss": 0.4639,
+      "step": 14670
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2157337716642155e-05,
+      "loss": 0.5116,
+      "step": 14671
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2156760965481444e-05,
+      "loss": 0.5137,
+      "step": 14672
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2156184214320732e-05,
+      "loss": 0.4928,
+      "step": 14673
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.215560746316002e-05,
+      "loss": 0.559,
+      "step": 14674
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2155030711999309e-05,
+      "loss": 0.5204,
+      "step": 14675
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2154453960838597e-05,
+      "loss": 0.4072,
+      "step": 14676
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2153877209677884e-05,
+      "loss": 0.4223,
+      "step": 14677
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2153300458517174e-05,
+      "loss": 0.5541,
+      "step": 14678
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2152723707356461e-05,
+      "loss": 0.4705,
+      "step": 14679
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2152146956195749e-05,
+      "loss": 0.553,
+      "step": 14680
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2151570205035038e-05,
+      "loss": 0.5871,
+      "step": 14681
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2150993453874326e-05,
+      "loss": 0.6024,
+      "step": 14682
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2150416702713614e-05,
+      "loss": 0.3797,
+      "step": 14683
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2149839951552905e-05,
+      "loss": 0.599,
+      "step": 14684
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2149263200392193e-05,
+      "loss": 0.5153,
+      "step": 14685
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2148686449231479e-05,
+      "loss": 0.4757,
+      "step": 14686
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.214810969807077e-05,
+      "loss": 0.5779,
+      "step": 14687
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2147532946910057e-05,
+      "loss": 0.3708,
+      "step": 14688
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2146956195749345e-05,
+      "loss": 0.4885,
+      "step": 14689
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2146379444588634e-05,
+      "loss": 0.6341,
+      "step": 14690
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2145802693427922e-05,
+      "loss": 0.4859,
+      "step": 14691
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.214522594226721e-05,
+      "loss": 0.4171,
+      "step": 14692
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2144649191106499e-05,
+      "loss": 0.3687,
+      "step": 14693
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2144072439945787e-05,
+      "loss": 0.5049,
+      "step": 14694
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2143495688785074e-05,
+      "loss": 0.5912,
+      "step": 14695
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2142918937624364e-05,
+      "loss": 0.4361,
+      "step": 14696
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2142342186463652e-05,
+      "loss": 0.4288,
+      "step": 14697
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.214176543530294e-05,
+      "loss": 0.5508,
+      "step": 14698
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2141188684142229e-05,
+      "loss": 0.4829,
+      "step": 14699
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2140611932981516e-05,
+      "loss": 0.5414,
+      "step": 14700
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2140035181820804e-05,
+      "loss": 0.5777,
+      "step": 14701
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2139458430660093e-05,
+      "loss": 0.5712,
+      "step": 14702
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2138881679499381e-05,
+      "loss": 0.5483,
+      "step": 14703
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2138304928338669e-05,
+      "loss": 0.4956,
+      "step": 14704
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2137728177177956e-05,
+      "loss": 0.4948,
+      "step": 14705
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2137151426017246e-05,
+      "loss": 0.5284,
+      "step": 14706
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2136574674856533e-05,
+      "loss": 0.438,
+      "step": 14707
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2135997923695821e-05,
+      "loss": 0.5706,
+      "step": 14708
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.213542117253511e-05,
+      "loss": 0.565,
+      "step": 14709
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2134844421374398e-05,
+      "loss": 0.3873,
+      "step": 14710
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2134267670213686e-05,
+      "loss": 0.525,
+      "step": 14711
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2133690919052975e-05,
+      "loss": 0.4835,
+      "step": 14712
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2133114167892263e-05,
+      "loss": 0.5335,
+      "step": 14713
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.213253741673155e-05,
+      "loss": 0.4447,
+      "step": 14714
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2131960665570842e-05,
+      "loss": 0.4745,
+      "step": 14715
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.213138391441013e-05,
+      "loss": 0.5014,
+      "step": 14716
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2130807163249415e-05,
+      "loss": 0.4684,
+      "step": 14717
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2130230412088706e-05,
+      "loss": 0.5122,
+      "step": 14718
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2129653660927994e-05,
+      "loss": 0.3467,
+      "step": 14719
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2129076909767282e-05,
+      "loss": 0.544,
+      "step": 14720
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2128500158606571e-05,
+      "loss": 0.5398,
+      "step": 14721
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2127923407445859e-05,
+      "loss": 0.5308,
+      "step": 14722
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2127346656285147e-05,
+      "loss": 0.4214,
+      "step": 14723
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2126769905124436e-05,
+      "loss": 0.4171,
+      "step": 14724
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2126193153963724e-05,
+      "loss": 0.4607,
+      "step": 14725
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2125616402803011e-05,
+      "loss": 0.4036,
+      "step": 14726
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.21250396516423e-05,
+      "loss": 0.4765,
+      "step": 14727
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2124462900481588e-05,
+      "loss": 0.4284,
+      "step": 14728
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2123886149320876e-05,
+      "loss": 0.4199,
+      "step": 14729
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2123309398160165e-05,
+      "loss": 0.4845,
+      "step": 14730
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2122732646999453e-05,
+      "loss": 0.5485,
+      "step": 14731
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2122155895838741e-05,
+      "loss": 0.6279,
+      "step": 14732
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.212157914467803e-05,
+      "loss": 0.4116,
+      "step": 14733
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2121002393517318e-05,
+      "loss": 0.4452,
+      "step": 14734
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2120425642356606e-05,
+      "loss": 0.5309,
+      "step": 14735
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2119848891195895e-05,
+      "loss": 0.5067,
+      "step": 14736
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2119272140035183e-05,
+      "loss": 0.4952,
+      "step": 14737
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.211869538887447e-05,
+      "loss": 0.5023,
+      "step": 14738
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2118118637713758e-05,
+      "loss": 0.475,
+      "step": 14739
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2117541886553047e-05,
+      "loss": 0.4257,
+      "step": 14740
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2116965135392335e-05,
+      "loss": 0.5393,
+      "step": 14741
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2116388384231623e-05,
+      "loss": 0.4896,
+      "step": 14742
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2115811633070912e-05,
+      "loss": 0.3461,
+      "step": 14743
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.21152348819102e-05,
+      "loss": 0.4284,
+      "step": 14744
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2114658130749488e-05,
+      "loss": 0.4692,
+      "step": 14745
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.2114081379588779e-05,
+      "loss": 0.4604,
+      "step": 14746
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2113504628428066e-05,
+      "loss": 0.5113,
+      "step": 14747
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2112927877267352e-05,
+      "loss": 0.566,
+      "step": 14748
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2112351126106643e-05,
+      "loss": 0.4093,
+      "step": 14749
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2111774374945931e-05,
+      "loss": 0.5532,
+      "step": 14750
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2111197623785219e-05,
+      "loss": 0.5629,
+      "step": 14751
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2110620872624508e-05,
+      "loss": 0.5044,
+      "step": 14752
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2110044121463796e-05,
+      "loss": 0.5315,
+      "step": 14753
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2109467370303083e-05,
+      "loss": 0.3193,
+      "step": 14754
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2108890619142373e-05,
+      "loss": 0.5266,
+      "step": 14755
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.210831386798166e-05,
+      "loss": 0.4869,
+      "step": 14756
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2107737116820948e-05,
+      "loss": 0.5126,
+      "step": 14757
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2107160365660238e-05,
+      "loss": 0.586,
+      "step": 14758
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2106583614499525e-05,
+      "loss": 0.4315,
+      "step": 14759
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2106006863338813e-05,
+      "loss": 0.4512,
+      "step": 14760
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2105430112178102e-05,
+      "loss": 0.5854,
+      "step": 14761
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.210485336101739e-05,
+      "loss": 0.5582,
+      "step": 14762
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2104276609856678e-05,
+      "loss": 0.5877,
+      "step": 14763
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2103699858695967e-05,
+      "loss": 0.5562,
+      "step": 14764
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2103123107535255e-05,
+      "loss": 0.3717,
+      "step": 14765
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2102546356374542e-05,
+      "loss": 0.433,
+      "step": 14766
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2101969605213832e-05,
+      "loss": 0.4413,
+      "step": 14767
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.210139285405312e-05,
+      "loss": 0.4704,
+      "step": 14768
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2100816102892407e-05,
+      "loss": 0.5007,
+      "step": 14769
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2100239351731697e-05,
+      "loss": 0.5304,
+      "step": 14770
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2099662600570984e-05,
+      "loss": 0.4797,
+      "step": 14771
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2099085849410272e-05,
+      "loss": 0.5061,
+      "step": 14772
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2098509098249563e-05,
+      "loss": 0.4201,
+      "step": 14773
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2097932347088849e-05,
+      "loss": 0.5393,
+      "step": 14774
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2097355595928137e-05,
+      "loss": 0.6139,
+      "step": 14775
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2096778844767424e-05,
+      "loss": 0.402,
+      "step": 14776
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2096202093606715e-05,
+      "loss": 0.3464,
+      "step": 14777
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2095625342446003e-05,
+      "loss": 0.3973,
+      "step": 14778
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.209504859128529e-05,
+      "loss": 0.4196,
+      "step": 14779
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.209447184012458e-05,
+      "loss": 0.4872,
+      "step": 14780
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2093895088963868e-05,
+      "loss": 0.6045,
+      "step": 14781
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2093318337803156e-05,
+      "loss": 0.4979,
+      "step": 14782
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2092741586642445e-05,
+      "loss": 0.3989,
+      "step": 14783
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2092164835481733e-05,
+      "loss": 0.6024,
+      "step": 14784
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.209158808432102e-05,
+      "loss": 0.6111,
+      "step": 14785
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.209101133316031e-05,
+      "loss": 0.5051,
+      "step": 14786
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2090434581999597e-05,
+      "loss": 0.5033,
+      "step": 14787
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2089857830838885e-05,
+      "loss": 0.4038,
+      "step": 14788
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2089281079678174e-05,
+      "loss": 0.5062,
+      "step": 14789
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2088704328517462e-05,
+      "loss": 0.4009,
+      "step": 14790
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.208812757735675e-05,
+      "loss": 0.5619,
+      "step": 14791
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.208755082619604e-05,
+      "loss": 0.4672,
+      "step": 14792
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2086974075035327e-05,
+      "loss": 0.4635,
+      "step": 14793
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2086397323874615e-05,
+      "loss": 0.6178,
+      "step": 14794
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2085820572713904e-05,
+      "loss": 0.5588,
+      "step": 14795
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2085243821553192e-05,
+      "loss": 0.4698,
+      "step": 14796
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.208466707039248e-05,
+      "loss": 0.5056,
+      "step": 14797
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2084090319231769e-05,
+      "loss": 0.4063,
+      "step": 14798
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2083513568071056e-05,
+      "loss": 0.5691,
+      "step": 14799
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2082936816910344e-05,
+      "loss": 0.4409,
+      "step": 14800
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2082360065749633e-05,
+      "loss": 0.4913,
+      "step": 14801
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2081783314588921e-05,
+      "loss": 0.4973,
+      "step": 14802
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2081206563428209e-05,
+      "loss": 0.5063,
+      "step": 14803
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.20806298122675e-05,
+      "loss": 0.4319,
+      "step": 14804
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2080053061106786e-05,
+      "loss": 0.4959,
+      "step": 14805
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2079476309946074e-05,
+      "loss": 0.4053,
+      "step": 14806
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2078899558785365e-05,
+      "loss": 0.4957,
+      "step": 14807
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2078322807624652e-05,
+      "loss": 0.4659,
+      "step": 14808
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.207774605646394e-05,
+      "loss": 0.4897,
+      "step": 14809
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2077169305303226e-05,
+      "loss": 0.471,
+      "step": 14810
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2076592554142517e-05,
+      "loss": 0.475,
+      "step": 14811
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2076015802981805e-05,
+      "loss": 0.5441,
+      "step": 14812
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2075439051821092e-05,
+      "loss": 0.4821,
+      "step": 14813
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2074862300660382e-05,
+      "loss": 0.5404,
+      "step": 14814
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.207428554949967e-05,
+      "loss": 0.554,
+      "step": 14815
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2073708798338957e-05,
+      "loss": 0.4842,
+      "step": 14816
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2073132047178247e-05,
+      "loss": 0.5047,
+      "step": 14817
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2072555296017534e-05,
+      "loss": 0.5134,
+      "step": 14818
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2071978544856822e-05,
+      "loss": 0.5282,
+      "step": 14819
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2071401793696111e-05,
+      "loss": 0.4529,
+      "step": 14820
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2070825042535399e-05,
+      "loss": 0.5044,
+      "step": 14821
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2070248291374687e-05,
+      "loss": 0.4358,
+      "step": 14822
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2069671540213976e-05,
+      "loss": 0.4198,
+      "step": 14823
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2069094789053264e-05,
+      "loss": 0.5548,
+      "step": 14824
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2068518037892551e-05,
+      "loss": 0.5204,
+      "step": 14825
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2067941286731841e-05,
+      "loss": 0.5254,
+      "step": 14826
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2067364535571129e-05,
+      "loss": 0.4311,
+      "step": 14827
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2066787784410416e-05,
+      "loss": 0.5631,
+      "step": 14828
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2066211033249706e-05,
+      "loss": 0.4851,
+      "step": 14829
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2065634282088993e-05,
+      "loss": 0.4414,
+      "step": 14830
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2065057530928281e-05,
+      "loss": 0.4877,
+      "step": 14831
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.206448077976757e-05,
+      "loss": 0.4981,
+      "step": 14832
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2063904028606858e-05,
+      "loss": 0.4868,
+      "step": 14833
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2063327277446146e-05,
+      "loss": 0.5246,
+      "step": 14834
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2062750526285437e-05,
+      "loss": 0.318,
+      "step": 14835
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2062173775124724e-05,
+      "loss": 0.3858,
+      "step": 14836
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.206159702396401e-05,
+      "loss": 0.5501,
+      "step": 14837
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2061020272803302e-05,
+      "loss": 0.4578,
+      "step": 14838
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.206044352164259e-05,
+      "loss": 0.3882,
+      "step": 14839
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2059866770481877e-05,
+      "loss": 0.5055,
+      "step": 14840
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2059290019321166e-05,
+      "loss": 0.4852,
+      "step": 14841
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2058713268160454e-05,
+      "loss": 0.5276,
+      "step": 14842
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2058136516999742e-05,
+      "loss": 0.5342,
+      "step": 14843
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.205755976583903e-05,
+      "loss": 0.3718,
+      "step": 14844
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2056983014678319e-05,
+      "loss": 0.5877,
+      "step": 14845
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2056406263517606e-05,
+      "loss": 0.4159,
+      "step": 14846
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2055829512356894e-05,
+      "loss": 0.3894,
+      "step": 14847
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2055252761196183e-05,
+      "loss": 0.4403,
+      "step": 14848
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2054676010035471e-05,
+      "loss": 0.4785,
+      "step": 14849
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2054099258874759e-05,
+      "loss": 0.5001,
+      "step": 14850
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2053522507714048e-05,
+      "loss": 0.644,
+      "step": 14851
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2052945756553336e-05,
+      "loss": 0.5793,
+      "step": 14852
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2052369005392624e-05,
+      "loss": 0.4955,
+      "step": 14853
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2051792254231913e-05,
+      "loss": 0.3932,
+      "step": 14854
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.20512155030712e-05,
+      "loss": 0.5079,
+      "step": 14855
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2050638751910488e-05,
+      "loss": 0.5096,
+      "step": 14856
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2050062000749778e-05,
+      "loss": 0.3323,
+      "step": 14857
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2049485249589065e-05,
+      "loss": 0.4154,
+      "step": 14858
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2048908498428353e-05,
+      "loss": 0.4275,
+      "step": 14859
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2048331747267642e-05,
+      "loss": 0.6595,
+      "step": 14860
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.204775499610693e-05,
+      "loss": 0.5728,
+      "step": 14861
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2047178244946218e-05,
+      "loss": 0.5177,
+      "step": 14862
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2046601493785507e-05,
+      "loss": 0.5289,
+      "step": 14863
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2046024742624795e-05,
+      "loss": 0.4537,
+      "step": 14864
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2045447991464083e-05,
+      "loss": 0.5337,
+      "step": 14865
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2044871240303374e-05,
+      "loss": 0.5509,
+      "step": 14866
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2044294489142661e-05,
+      "loss": 0.4111,
+      "step": 14867
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2043717737981947e-05,
+      "loss": 0.5373,
+      "step": 14868
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2043140986821238e-05,
+      "loss": 0.464,
+      "step": 14869
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2042564235660526e-05,
+      "loss": 0.517,
+      "step": 14870
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2041987484499814e-05,
+      "loss": 0.5862,
+      "step": 14871
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2041410733339103e-05,
+      "loss": 0.4605,
+      "step": 14872
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2040833982178391e-05,
+      "loss": 0.4209,
+      "step": 14873
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2040257231017679e-05,
+      "loss": 0.6097,
+      "step": 14874
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2039680479856968e-05,
+      "loss": 0.5085,
+      "step": 14875
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2039103728696256e-05,
+      "loss": 0.5233,
+      "step": 14876
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2038526977535543e-05,
+      "loss": 0.4745,
+      "step": 14877
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2037950226374831e-05,
+      "loss": 0.4731,
+      "step": 14878
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.203737347521412e-05,
+      "loss": 0.5132,
+      "step": 14879
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2036796724053408e-05,
+      "loss": 0.3517,
+      "step": 14880
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2036219972892696e-05,
+      "loss": 0.423,
+      "step": 14881
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2035643221731985e-05,
+      "loss": 0.4993,
+      "step": 14882
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2035066470571273e-05,
+      "loss": 0.5095,
+      "step": 14883
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.203448971941056e-05,
+      "loss": 0.4192,
+      "step": 14884
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.203391296824985e-05,
+      "loss": 0.5499,
+      "step": 14885
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2033336217089138e-05,
+      "loss": 0.4889,
+      "step": 14886
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2032759465928425e-05,
+      "loss": 0.4709,
+      "step": 14887
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2032182714767715e-05,
+      "loss": 0.559,
+      "step": 14888
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2031605963607002e-05,
+      "loss": 0.4834,
+      "step": 14889
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.203102921244629e-05,
+      "loss": 0.5633,
+      "step": 14890
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.203045246128558e-05,
+      "loss": 0.4864,
+      "step": 14891
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2029875710124867e-05,
+      "loss": 0.544,
+      "step": 14892
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2029298958964155e-05,
+      "loss": 0.4632,
+      "step": 14893
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2028722207803444e-05,
+      "loss": 0.4554,
+      "step": 14894
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2028145456642732e-05,
+      "loss": 0.5657,
+      "step": 14895
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202756870548202e-05,
+      "loss": 0.4722,
+      "step": 14896
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202699195432131e-05,
+      "loss": 0.5398,
+      "step": 14897
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2026415203160598e-05,
+      "loss": 0.4726,
+      "step": 14898
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2025838451999884e-05,
+      "loss": 0.463,
+      "step": 14899
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2025261700839175e-05,
+      "loss": 0.3779,
+      "step": 14900
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2024684949678463e-05,
+      "loss": 0.4224,
+      "step": 14901
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202410819851775e-05,
+      "loss": 0.4413,
+      "step": 14902
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202353144735704e-05,
+      "loss": 0.5354,
+      "step": 14903
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2022954696196328e-05,
+      "loss": 0.4544,
+      "step": 14904
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2022377945035615e-05,
+      "loss": 0.4076,
+      "step": 14905
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2021801193874905e-05,
+      "loss": 0.6081,
+      "step": 14906
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2021224442714192e-05,
+      "loss": 0.5115,
+      "step": 14907
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202064769155348e-05,
+      "loss": 0.5711,
+      "step": 14908
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.202007094039277e-05,
+      "loss": 0.4391,
+      "step": 14909
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2019494189232057e-05,
+      "loss": 0.421,
+      "step": 14910
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2018917438071345e-05,
+      "loss": 0.3926,
+      "step": 14911
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2018340686910634e-05,
+      "loss": 0.5305,
+      "step": 14912
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2017763935749922e-05,
+      "loss": 0.4582,
+      "step": 14913
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.201718718458921e-05,
+      "loss": 0.4492,
+      "step": 14914
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2016610433428497e-05,
+      "loss": 0.4661,
+      "step": 14915
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2016033682267787e-05,
+      "loss": 0.4774,
+      "step": 14916
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2015456931107074e-05,
+      "loss": 0.4855,
+      "step": 14917
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2014880179946362e-05,
+      "loss": 0.5218,
+      "step": 14918
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2014303428785651e-05,
+      "loss": 0.4409,
+      "step": 14919
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.201372667762494e-05,
+      "loss": 0.4122,
+      "step": 14920
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2013149926464227e-05,
+      "loss": 0.4023,
+      "step": 14921
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2012573175303516e-05,
+      "loss": 0.5331,
+      "step": 14922
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2011996424142804e-05,
+      "loss": 0.4511,
+      "step": 14923
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2011419672982092e-05,
+      "loss": 0.4937,
+      "step": 14924
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 1.2010842921821381e-05,
+      "loss": 0.4354,
+      "step": 14925
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2010266170660669e-05,
+      "loss": 0.4286,
+      "step": 14926
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2009689419499956e-05,
+      "loss": 0.5461,
+      "step": 14927
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2009112668339247e-05,
+      "loss": 0.5069,
+      "step": 14928
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2008535917178535e-05,
+      "loss": 0.497,
+      "step": 14929
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2007959166017821e-05,
+      "loss": 0.5179,
+      "step": 14930
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2007382414857112e-05,
+      "loss": 0.4523,
+      "step": 14931
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.20068056636964e-05,
+      "loss": 0.5755,
+      "step": 14932
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2006228912535688e-05,
+      "loss": 0.4335,
+      "step": 14933
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2005652161374977e-05,
+      "loss": 0.5257,
+      "step": 14934
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2005075410214265e-05,
+      "loss": 0.5249,
+      "step": 14935
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2004498659053552e-05,
+      "loss": 0.5285,
+      "step": 14936
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2003921907892842e-05,
+      "loss": 0.4859,
+      "step": 14937
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.200334515673213e-05,
+      "loss": 0.4771,
+      "step": 14938
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2002768405571417e-05,
+      "loss": 0.5711,
+      "step": 14939
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2002191654410706e-05,
+      "loss": 0.4874,
+      "step": 14940
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2001614903249994e-05,
+      "loss": 0.4514,
+      "step": 14941
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2001038152089282e-05,
+      "loss": 0.584,
+      "step": 14942
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.2000461400928571e-05,
+      "loss": 0.4964,
+      "step": 14943
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1999884649767859e-05,
+      "loss": 0.4888,
+      "step": 14944
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1999307898607147e-05,
+      "loss": 0.4763,
+      "step": 14945
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1998731147446436e-05,
+      "loss": 0.4075,
+      "step": 14946
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1998154396285724e-05,
+      "loss": 0.5795,
+      "step": 14947
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1997577645125011e-05,
+      "loss": 0.4999,
+      "step": 14948
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1997000893964299e-05,
+      "loss": 0.5224,
+      "step": 14949
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1996424142803588e-05,
+      "loss": 0.4502,
+      "step": 14950
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1995847391642876e-05,
+      "loss": 0.6177,
+      "step": 14951
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1995270640482164e-05,
+      "loss": 0.5617,
+      "step": 14952
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1994693889321453e-05,
+      "loss": 0.532,
+      "step": 14953
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.199411713816074e-05,
+      "loss": 0.4512,
+      "step": 14954
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1993540387000028e-05,
+      "loss": 0.4579,
+      "step": 14955
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1992963635839318e-05,
+      "loss": 0.51,
+      "step": 14956
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1992386884678606e-05,
+      "loss": 0.5049,
+      "step": 14957
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1991810133517893e-05,
+      "loss": 0.5873,
+      "step": 14958
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1991233382357184e-05,
+      "loss": 0.4052,
+      "step": 14959
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1990656631196472e-05,
+      "loss": 0.4751,
+      "step": 14960
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1990079880035758e-05,
+      "loss": 0.4687,
+      "step": 14961
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1989503128875049e-05,
+      "loss": 0.5325,
+      "step": 14962
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1988926377714337e-05,
+      "loss": 0.4369,
+      "step": 14963
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1988349626553624e-05,
+      "loss": 0.4615,
+      "step": 14964
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1987772875392914e-05,
+      "loss": 0.5522,
+      "step": 14965
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1987196124232201e-05,
+      "loss": 0.45,
+      "step": 14966
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.198661937307149e-05,
+      "loss": 0.3284,
+      "step": 14967
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1986042621910779e-05,
+      "loss": 0.4087,
+      "step": 14968
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1985465870750066e-05,
+      "loss": 0.5799,
+      "step": 14969
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1984889119589354e-05,
+      "loss": 0.4332,
+      "step": 14970
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1984312368428643e-05,
+      "loss": 0.5017,
+      "step": 14971
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1983735617267931e-05,
+      "loss": 0.4344,
+      "step": 14972
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1983158866107219e-05,
+      "loss": 0.5922,
+      "step": 14973
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1982582114946508e-05,
+      "loss": 0.4524,
+      "step": 14974
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1982005363785796e-05,
+      "loss": 0.4163,
+      "step": 14975
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1981428612625083e-05,
+      "loss": 0.5518,
+      "step": 14976
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1980851861464373e-05,
+      "loss": 0.5215,
+      "step": 14977
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.198027511030366e-05,
+      "loss": 0.417,
+      "step": 14978
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1979698359142948e-05,
+      "loss": 0.404,
+      "step": 14979
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1979121607982238e-05,
+      "loss": 0.6565,
+      "step": 14980
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1978544856821525e-05,
+      "loss": 0.5462,
+      "step": 14981
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1977968105660813e-05,
+      "loss": 0.4434,
+      "step": 14982
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.19773913545001e-05,
+      "loss": 0.4593,
+      "step": 14983
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.197681460333939e-05,
+      "loss": 0.5215,
+      "step": 14984
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1976237852178678e-05,
+      "loss": 0.5632,
+      "step": 14985
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1975661101017965e-05,
+      "loss": 0.4633,
+      "step": 14986
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1975084349857255e-05,
+      "loss": 0.5257,
+      "step": 14987
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1974507598696542e-05,
+      "loss": 0.4739,
+      "step": 14988
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.197393084753583e-05,
+      "loss": 0.6197,
+      "step": 14989
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1973354096375121e-05,
+      "loss": 0.4199,
+      "step": 14990
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1972777345214409e-05,
+      "loss": 0.3971,
+      "step": 14991
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1972200594053695e-05,
+      "loss": 0.3918,
+      "step": 14992
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1971623842892986e-05,
+      "loss": 0.5227,
+      "step": 14993
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1971047091732274e-05,
+      "loss": 0.5309,
+      "step": 14994
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1970470340571561e-05,
+      "loss": 0.5507,
+      "step": 14995
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196989358941085e-05,
+      "loss": 0.4738,
+      "step": 14996
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1969316838250138e-05,
+      "loss": 0.5456,
+      "step": 14997
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1968740087089426e-05,
+      "loss": 0.5402,
+      "step": 14998
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1968163335928715e-05,
+      "loss": 0.5273,
+      "step": 14999
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1967586584768003e-05,
+      "loss": 0.4925,
+      "step": 15000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196700983360729e-05,
+      "loss": 0.5573,
+      "step": 15001
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196643308244658e-05,
+      "loss": 0.4921,
+      "step": 15002
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1965856331285868e-05,
+      "loss": 0.4345,
+      "step": 15003
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1965279580125156e-05,
+      "loss": 0.4981,
+      "step": 15004
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1964702828964445e-05,
+      "loss": 0.5302,
+      "step": 15005
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1964126077803733e-05,
+      "loss": 0.461,
+      "step": 15006
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196354932664302e-05,
+      "loss": 0.4191,
+      "step": 15007
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196297257548231e-05,
+      "loss": 0.406,
+      "step": 15008
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1962395824321597e-05,
+      "loss": 0.3651,
+      "step": 15009
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1961819073160885e-05,
+      "loss": 0.4071,
+      "step": 15010
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1961242322000174e-05,
+      "loss": 0.4804,
+      "step": 15011
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1960665570839462e-05,
+      "loss": 0.3719,
+      "step": 15012
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.196008881967875e-05,
+      "loss": 0.4674,
+      "step": 15013
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.195951206851804e-05,
+      "loss": 0.4626,
+      "step": 15014
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1958935317357327e-05,
+      "loss": 0.5482,
+      "step": 15015
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1958358566196615e-05,
+      "loss": 0.5327,
+      "step": 15016
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1957781815035902e-05,
+      "loss": 0.5483,
+      "step": 15017
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1957205063875192e-05,
+      "loss": 0.4683,
+      "step": 15018
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.195662831271448e-05,
+      "loss": 0.4996,
+      "step": 15019
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1956051561553767e-05,
+      "loss": 0.463,
+      "step": 15020
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1955474810393058e-05,
+      "loss": 0.4615,
+      "step": 15021
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1954898059232346e-05,
+      "loss": 0.48,
+      "step": 15022
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1954321308071632e-05,
+      "loss": 0.3825,
+      "step": 15023
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1953744556910923e-05,
+      "loss": 0.5852,
+      "step": 15024
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.195316780575021e-05,
+      "loss": 0.5768,
+      "step": 15025
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1952591054589498e-05,
+      "loss": 0.4913,
+      "step": 15026
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1952014303428788e-05,
+      "loss": 0.3883,
+      "step": 15027
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1951437552268075e-05,
+      "loss": 0.4393,
+      "step": 15028
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1950860801107363e-05,
+      "loss": 0.5345,
+      "step": 15029
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1950284049946652e-05,
+      "loss": 0.5158,
+      "step": 15030
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.194970729878594e-05,
+      "loss": 0.4631,
+      "step": 15031
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1949130547625228e-05,
+      "loss": 0.6435,
+      "step": 15032
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1948553796464517e-05,
+      "loss": 0.479,
+      "step": 15033
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1947977045303805e-05,
+      "loss": 0.492,
+      "step": 15034
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1947400294143092e-05,
+      "loss": 0.4295,
+      "step": 15035
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1946823542982382e-05,
+      "loss": 0.4109,
+      "step": 15036
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.194624679182167e-05,
+      "loss": 0.5121,
+      "step": 15037
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1945670040660957e-05,
+      "loss": 0.5249,
+      "step": 15038
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1945093289500247e-05,
+      "loss": 0.4704,
+      "step": 15039
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1944516538339534e-05,
+      "loss": 0.5954,
+      "step": 15040
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1943939787178822e-05,
+      "loss": 0.4741,
+      "step": 15041
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1943363036018111e-05,
+      "loss": 0.4607,
+      "step": 15042
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1942786284857399e-05,
+      "loss": 0.4062,
+      "step": 15043
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1942209533696687e-05,
+      "loss": 0.4494,
+      "step": 15044
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1941632782535976e-05,
+      "loss": 0.52,
+      "step": 15045
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1941056031375264e-05,
+      "loss": 0.4872,
+      "step": 15046
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1940479280214551e-05,
+      "loss": 0.3648,
+      "step": 15047
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1939902529053842e-05,
+      "loss": 0.4833,
+      "step": 15048
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1939325777893128e-05,
+      "loss": 0.5028,
+      "step": 15049
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1938749026732416e-05,
+      "loss": 0.4329,
+      "step": 15050
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1938172275571707e-05,
+      "loss": 0.3794,
+      "step": 15051
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1937595524410995e-05,
+      "loss": 0.4172,
+      "step": 15052
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1937018773250283e-05,
+      "loss": 0.4189,
+      "step": 15053
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1936442022089569e-05,
+      "loss": 0.4618,
+      "step": 15054
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.193586527092886e-05,
+      "loss": 0.4603,
+      "step": 15055
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1935288519768147e-05,
+      "loss": 0.4173,
+      "step": 15056
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1934711768607435e-05,
+      "loss": 0.4653,
+      "step": 15057
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1934135017446724e-05,
+      "loss": 0.4075,
+      "step": 15058
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1933558266286012e-05,
+      "loss": 0.6412,
+      "step": 15059
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.19329815151253e-05,
+      "loss": 0.4884,
+      "step": 15060
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.193240476396459e-05,
+      "loss": 0.5389,
+      "step": 15061
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1931828012803877e-05,
+      "loss": 0.4749,
+      "step": 15062
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1931251261643165e-05,
+      "loss": 0.5444,
+      "step": 15063
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1930674510482454e-05,
+      "loss": 0.5342,
+      "step": 15064
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1930097759321742e-05,
+      "loss": 0.4887,
+      "step": 15065
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.192952100816103e-05,
+      "loss": 0.5392,
+      "step": 15066
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1928944257000319e-05,
+      "loss": 0.4594,
+      "step": 15067
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1928367505839606e-05,
+      "loss": 0.6165,
+      "step": 15068
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1927790754678894e-05,
+      "loss": 0.5053,
+      "step": 15069
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1927214003518183e-05,
+      "loss": 0.5246,
+      "step": 15070
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1926637252357471e-05,
+      "loss": 0.6013,
+      "step": 15071
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1926060501196759e-05,
+      "loss": 0.4443,
+      "step": 15072
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1925483750036048e-05,
+      "loss": 0.5062,
+      "step": 15073
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1924906998875336e-05,
+      "loss": 0.5457,
+      "step": 15074
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1924330247714624e-05,
+      "loss": 0.4371,
+      "step": 15075
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1923753496553913e-05,
+      "loss": 0.4132,
+      "step": 15076
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.19231767453932e-05,
+      "loss": 0.3899,
+      "step": 15077
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1922599994232488e-05,
+      "loss": 0.4148,
+      "step": 15078
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.192202324307178e-05,
+      "loss": 0.4436,
+      "step": 15079
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1921446491911065e-05,
+      "loss": 0.4206,
+      "step": 15080
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1920869740750353e-05,
+      "loss": 0.5231,
+      "step": 15081
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1920292989589644e-05,
+      "loss": 0.44,
+      "step": 15082
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1919716238428932e-05,
+      "loss": 0.4612,
+      "step": 15083
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.191913948726822e-05,
+      "loss": 0.3931,
+      "step": 15084
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1918562736107509e-05,
+      "loss": 0.5204,
+      "step": 15085
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1917985984946797e-05,
+      "loss": 0.5542,
+      "step": 15086
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1917409233786084e-05,
+      "loss": 0.441,
+      "step": 15087
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1916832482625372e-05,
+      "loss": 0.4132,
+      "step": 15088
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1916255731464661e-05,
+      "loss": 0.5384,
+      "step": 15089
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1915678980303949e-05,
+      "loss": 0.4482,
+      "step": 15090
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1915102229143237e-05,
+      "loss": 0.6154,
+      "step": 15091
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1914525477982526e-05,
+      "loss": 0.4958,
+      "step": 15092
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1913948726821814e-05,
+      "loss": 0.4567,
+      "step": 15093
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1913371975661101e-05,
+      "loss": 0.5414,
+      "step": 15094
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.191279522450039e-05,
+      "loss": 0.4345,
+      "step": 15095
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1912218473339678e-05,
+      "loss": 0.357,
+      "step": 15096
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1911641722178966e-05,
+      "loss": 0.4438,
+      "step": 15097
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1911064971018256e-05,
+      "loss": 0.6178,
+      "step": 15098
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1910488219857543e-05,
+      "loss": 0.4814,
+      "step": 15099
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1909911468696831e-05,
+      "loss": 0.5401,
+      "step": 15100
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.190933471753612e-05,
+      "loss": 0.4373,
+      "step": 15101
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1908757966375408e-05,
+      "loss": 0.4936,
+      "step": 15102
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1908181215214696e-05,
+      "loss": 0.4869,
+      "step": 15103
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.1907604464053985e-05,
+      "loss": 0.5189,
+      "step": 15104
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1907027712893273e-05,
+      "loss": 0.5331,
+      "step": 15105
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190645096173256e-05,
+      "loss": 0.3986,
+      "step": 15106
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190587421057185e-05,
+      "loss": 0.5796,
+      "step": 15107
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1905297459411137e-05,
+      "loss": 0.3819,
+      "step": 15108
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1904720708250425e-05,
+      "loss": 0.5068,
+      "step": 15109
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1904143957089716e-05,
+      "loss": 0.4952,
+      "step": 15110
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1903567205929002e-05,
+      "loss": 0.4322,
+      "step": 15111
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.190299045476829e-05,
+      "loss": 0.4932,
+      "step": 15112
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1902413703607581e-05,
+      "loss": 0.4369,
+      "step": 15113
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1901836952446869e-05,
+      "loss": 0.6096,
+      "step": 15114
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1901260201286156e-05,
+      "loss": 0.5457,
+      "step": 15115
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1900683450125446e-05,
+      "loss": 0.4453,
+      "step": 15116
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1900106698964733e-05,
+      "loss": 0.4594,
+      "step": 15117
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1899529947804021e-05,
+      "loss": 0.5805,
+      "step": 15118
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189895319664331e-05,
+      "loss": 0.4513,
+      "step": 15119
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1898376445482598e-05,
+      "loss": 0.4445,
+      "step": 15120
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1897799694321886e-05,
+      "loss": 0.5554,
+      "step": 15121
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1897222943161174e-05,
+      "loss": 0.4075,
+      "step": 15122
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1896646192000463e-05,
+      "loss": 0.4365,
+      "step": 15123
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189606944083975e-05,
+      "loss": 0.555,
+      "step": 15124
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1895492689679038e-05,
+      "loss": 0.6503,
+      "step": 15125
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1894915938518328e-05,
+      "loss": 0.589,
+      "step": 15126
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1894339187357615e-05,
+      "loss": 0.5754,
+      "step": 15127
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1893762436196903e-05,
+      "loss": 0.4173,
+      "step": 15128
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1893185685036192e-05,
+      "loss": 0.5245,
+      "step": 15129
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.189260893387548e-05,
+      "loss": 0.47,
+      "step": 15130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1892032182714768e-05,
+      "loss": 0.653,
+      "step": 15131
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1891455431554057e-05,
+      "loss": 0.4358,
+      "step": 15132
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1890878680393345e-05,
+      "loss": 0.4352,
+      "step": 15133
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1890301929232633e-05,
+      "loss": 0.4593,
+      "step": 15134
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1889725178071922e-05,
+      "loss": 0.4638,
+      "step": 15135
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.188914842691121e-05,
+      "loss": 0.5084,
+      "step": 15136
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1888571675750497e-05,
+      "loss": 0.4735,
+      "step": 15137
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1887994924589787e-05,
+      "loss": 0.4248,
+      "step": 15138
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1887418173429074e-05,
+      "loss": 0.4528,
+      "step": 15139
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1886841422268362e-05,
+      "loss": 0.5995,
+      "step": 15140
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1886264671107653e-05,
+      "loss": 0.4117,
+      "step": 15141
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1885687919946939e-05,
+      "loss": 0.5528,
+      "step": 15142
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1885111168786227e-05,
+      "loss": 0.5118,
+      "step": 15143
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1884534417625518e-05,
+      "loss": 0.6062,
+      "step": 15144
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1883957666464806e-05,
+      "loss": 0.3972,
+      "step": 15145
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1883380915304093e-05,
+      "loss": 0.4383,
+      "step": 15146
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1882804164143383e-05,
+      "loss": 0.5939,
+      "step": 15147
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.188222741298267e-05,
+      "loss": 0.5454,
+      "step": 15148
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1881650661821958e-05,
+      "loss": 0.4381,
+      "step": 15149
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1881073910661247e-05,
+      "loss": 0.4923,
+      "step": 15150
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1880497159500535e-05,
+      "loss": 0.4552,
+      "step": 15151
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1879920408339823e-05,
+      "loss": 0.4813,
+      "step": 15152
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1879343657179112e-05,
+      "loss": 0.6094,
+      "step": 15153
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18787669060184e-05,
+      "loss": 0.3961,
+      "step": 15154
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1878190154857687e-05,
+      "loss": 0.4345,
+      "step": 15155
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1877613403696977e-05,
+      "loss": 0.4321,
+      "step": 15156
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1877036652536265e-05,
+      "loss": 0.5321,
+      "step": 15157
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1876459901375552e-05,
+      "loss": 0.4671,
+      "step": 15158
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.187588315021484e-05,
+      "loss": 0.484,
+      "step": 15159
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.187530639905413e-05,
+      "loss": 0.5157,
+      "step": 15160
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1874729647893417e-05,
+      "loss": 0.5199,
+      "step": 15161
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1874152896732705e-05,
+      "loss": 0.488,
+      "step": 15162
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1873576145571994e-05,
+      "loss": 0.4093,
+      "step": 15163
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1872999394411282e-05,
+      "loss": 0.4946,
+      "step": 15164
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.187242264325057e-05,
+      "loss": 0.5366,
+      "step": 15165
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1871845892089859e-05,
+      "loss": 0.5901,
+      "step": 15166
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1871269140929146e-05,
+      "loss": 0.5003,
+      "step": 15167
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1870692389768434e-05,
+      "loss": 0.4858,
+      "step": 15168
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1870115638607724e-05,
+      "loss": 0.3614,
+      "step": 15169
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1869538887447011e-05,
+      "loss": 0.4878,
+      "step": 15170
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1868962136286299e-05,
+      "loss": 0.5589,
+      "step": 15171
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.186838538512559e-05,
+      "loss": 0.4632,
+      "step": 15172
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1867808633964878e-05,
+      "loss": 0.5983,
+      "step": 15173
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1867231882804164e-05,
+      "loss": 0.4745,
+      "step": 15174
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1866655131643455e-05,
+      "loss": 0.4834,
+      "step": 15175
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1866078380482742e-05,
+      "loss": 0.4439,
+      "step": 15176
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.186550162932203e-05,
+      "loss": 0.4306,
+      "step": 15177
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.186492487816132e-05,
+      "loss": 0.5392,
+      "step": 15178
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1864348127000607e-05,
+      "loss": 0.4571,
+      "step": 15179
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1863771375839895e-05,
+      "loss": 0.5293,
+      "step": 15180
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1863194624679184e-05,
+      "loss": 0.4629,
+      "step": 15181
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1862617873518472e-05,
+      "loss": 0.426,
+      "step": 15182
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.186204112235776e-05,
+      "loss": 0.5349,
+      "step": 15183
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1861464371197049e-05,
+      "loss": 0.5759,
+      "step": 15184
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1860887620036337e-05,
+      "loss": 0.363,
+      "step": 15185
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1860310868875624e-05,
+      "loss": 0.4939,
+      "step": 15186
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1859734117714914e-05,
+      "loss": 0.5264,
+      "step": 15187
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1859157366554201e-05,
+      "loss": 0.5007,
+      "step": 15188
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1858580615393489e-05,
+      "loss": 0.5396,
+      "step": 15189
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1858003864232778e-05,
+      "loss": 0.4515,
+      "step": 15190
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1857427113072066e-05,
+      "loss": 0.5155,
+      "step": 15191
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1856850361911354e-05,
+      "loss": 0.4298,
+      "step": 15192
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1856273610750642e-05,
+      "loss": 0.4366,
+      "step": 15193
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1855696859589931e-05,
+      "loss": 0.4854,
+      "step": 15194
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1855120108429219e-05,
+      "loss": 0.4356,
+      "step": 15195
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1854543357268506e-05,
+      "loss": 0.4823,
+      "step": 15196
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1853966606107796e-05,
+      "loss": 0.3491,
+      "step": 15197
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1853389854947083e-05,
+      "loss": 0.5801,
+      "step": 15198
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1852813103786371e-05,
+      "loss": 0.5579,
+      "step": 15199
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.185223635262566e-05,
+      "loss": 0.443,
+      "step": 15200
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1851659601464948e-05,
+      "loss": 0.4404,
+      "step": 15201
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1851082850304236e-05,
+      "loss": 0.4184,
+      "step": 15202
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1850506099143527e-05,
+      "loss": 0.5247,
+      "step": 15203
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1849929347982815e-05,
+      "loss": 0.3649,
+      "step": 15204
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18493525968221e-05,
+      "loss": 0.5233,
+      "step": 15205
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1848775845661392e-05,
+      "loss": 0.5034,
+      "step": 15206
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184819909450068e-05,
+      "loss": 0.5337,
+      "step": 15207
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1847622343339967e-05,
+      "loss": 0.4614,
+      "step": 15208
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1847045592179256e-05,
+      "loss": 0.5169,
+      "step": 15209
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1846468841018544e-05,
+      "loss": 0.501,
+      "step": 15210
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1845892089857832e-05,
+      "loss": 0.412,
+      "step": 15211
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1845315338697121e-05,
+      "loss": 0.4408,
+      "step": 15212
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1844738587536409e-05,
+      "loss": 0.5861,
+      "step": 15213
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1844161836375696e-05,
+      "loss": 0.3753,
+      "step": 15214
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1843585085214986e-05,
+      "loss": 0.4845,
+      "step": 15215
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1843008334054274e-05,
+      "loss": 0.3991,
+      "step": 15216
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1842431582893561e-05,
+      "loss": 0.5204,
+      "step": 15217
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.184185483173285e-05,
+      "loss": 0.4584,
+      "step": 15218
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1841278080572138e-05,
+      "loss": 0.5015,
+      "step": 15219
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1840701329411426e-05,
+      "loss": 0.5224,
+      "step": 15220
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1840124578250715e-05,
+      "loss": 0.4545,
+      "step": 15221
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1839547827090003e-05,
+      "loss": 0.3684,
+      "step": 15222
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183897107592929e-05,
+      "loss": 0.3004,
+      "step": 15223
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183839432476858e-05,
+      "loss": 0.5964,
+      "step": 15224
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1837817573607868e-05,
+      "loss": 0.4579,
+      "step": 15225
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1837240822447155e-05,
+      "loss": 0.5413,
+      "step": 15226
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1836664071286443e-05,
+      "loss": 0.4689,
+      "step": 15227
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1836087320125733e-05,
+      "loss": 0.5444,
+      "step": 15228
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.183551056896502e-05,
+      "loss": 0.5077,
+      "step": 15229
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1834933817804308e-05,
+      "loss": 0.3469,
+      "step": 15230
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1834357066643597e-05,
+      "loss": 0.4402,
+      "step": 15231
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1833780315482885e-05,
+      "loss": 0.3899,
+      "step": 15232
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1833203564322173e-05,
+      "loss": 0.4843,
+      "step": 15233
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1832626813161464e-05,
+      "loss": 0.6168,
+      "step": 15234
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1832050062000751e-05,
+      "loss": 0.4101,
+      "step": 15235
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1831473310840037e-05,
+      "loss": 0.3947,
+      "step": 15236
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1830896559679328e-05,
+      "loss": 0.483,
+      "step": 15237
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1830319808518616e-05,
+      "loss": 0.5686,
+      "step": 15238
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1829743057357904e-05,
+      "loss": 0.3949,
+      "step": 15239
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1829166306197193e-05,
+      "loss": 0.5732,
+      "step": 15240
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1828589555036481e-05,
+      "loss": 0.4358,
+      "step": 15241
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1828012803875769e-05,
+      "loss": 0.485,
+      "step": 15242
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1827436052715058e-05,
+      "loss": 0.5736,
+      "step": 15243
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1826859301554346e-05,
+      "loss": 0.3591,
+      "step": 15244
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1826282550393633e-05,
+      "loss": 0.6233,
+      "step": 15245
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1825705799232923e-05,
+      "loss": 0.4332,
+      "step": 15246
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.182512904807221e-05,
+      "loss": 0.6034,
+      "step": 15247
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1824552296911498e-05,
+      "loss": 0.5015,
+      "step": 15248
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1823975545750787e-05,
+      "loss": 0.4191,
+      "step": 15249
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1823398794590075e-05,
+      "loss": 0.4841,
+      "step": 15250
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1822822043429363e-05,
+      "loss": 0.4721,
+      "step": 15251
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1822245292268652e-05,
+      "loss": 0.5301,
+      "step": 15252
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.182166854110794e-05,
+      "loss": 0.5772,
+      "step": 15253
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1821091789947228e-05,
+      "loss": 0.6024,
+      "step": 15254
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1820515038786517e-05,
+      "loss": 0.5426,
+      "step": 15255
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1819938287625805e-05,
+      "loss": 0.4048,
+      "step": 15256
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1819361536465092e-05,
+      "loss": 0.5385,
+      "step": 15257
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1818784785304382e-05,
+      "loss": 0.364,
+      "step": 15258
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.181820803414367e-05,
+      "loss": 0.4929,
+      "step": 15259
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1817631282982957e-05,
+      "loss": 0.3577,
+      "step": 15260
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1817054531822245e-05,
+      "loss": 0.5851,
+      "step": 15261
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1816477780661534e-05,
+      "loss": 0.5423,
+      "step": 15262
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1815901029500822e-05,
+      "loss": 0.5363,
+      "step": 15263
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.181532427834011e-05,
+      "loss": 0.5648,
+      "step": 15264
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18147475271794e-05,
+      "loss": 0.4963,
+      "step": 15265
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1814170776018688e-05,
+      "loss": 0.4978,
+      "step": 15266
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1813594024857974e-05,
+      "loss": 0.545,
+      "step": 15267
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1813017273697265e-05,
+      "loss": 0.6029,
+      "step": 15268
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1812440522536553e-05,
+      "loss": 0.5632,
+      "step": 15269
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.181186377137584e-05,
+      "loss": 0.4462,
+      "step": 15270
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.181128702021513e-05,
+      "loss": 0.5241,
+      "step": 15271
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1810710269054418e-05,
+      "loss": 0.5359,
+      "step": 15272
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1810133517893705e-05,
+      "loss": 0.3902,
+      "step": 15273
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1809556766732995e-05,
+      "loss": 0.43,
+      "step": 15274
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1808980015572283e-05,
+      "loss": 0.448,
+      "step": 15275
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.180840326441157e-05,
+      "loss": 0.512,
+      "step": 15276
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.180782651325086e-05,
+      "loss": 0.4378,
+      "step": 15277
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1807249762090147e-05,
+      "loss": 0.3752,
+      "step": 15278
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1806673010929435e-05,
+      "loss": 0.5248,
+      "step": 15279
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1806096259768724e-05,
+      "loss": 0.3979,
+      "step": 15280
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1805519508608012e-05,
+      "loss": 0.509,
+      "step": 15281
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.18049427574473e-05,
+      "loss": 0.4267,
+      "step": 15282
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 1.1804366006286589e-05,
+      "loss": 0.4542,
+      "step": 15283
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1803789255125877e-05,
+      "loss": 0.4592,
+      "step": 15284
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1803212503965164e-05,
+      "loss": 0.3896,
+      "step": 15285
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1802635752804454e-05,
+      "loss": 0.503,
+      "step": 15286
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1802059001643742e-05,
+      "loss": 0.4798,
+      "step": 15287
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.180148225048303e-05,
+      "loss": 0.4231,
+      "step": 15288
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1800905499322319e-05,
+      "loss": 0.3932,
+      "step": 15289
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1800328748161606e-05,
+      "loss": 0.6057,
+      "step": 15290
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1799751997000894e-05,
+      "loss": 0.4146,
+      "step": 15291
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1799175245840185e-05,
+      "loss": 0.494,
+      "step": 15292
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1798598494679471e-05,
+      "loss": 0.6173,
+      "step": 15293
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1798021743518759e-05,
+      "loss": 0.4986,
+      "step": 15294
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.179744499235805e-05,
+      "loss": 0.5317,
+      "step": 15295
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1796868241197337e-05,
+      "loss": 0.5259,
+      "step": 15296
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1796291490036625e-05,
+      "loss": 0.4379,
+      "step": 15297
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1795714738875911e-05,
+      "loss": 0.4692,
+      "step": 15298
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1795137987715202e-05,
+      "loss": 0.4077,
+      "step": 15299
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.179456123655449e-05,
+      "loss": 0.4738,
+      "step": 15300
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1793984485393778e-05,
+      "loss": 0.5296,
+      "step": 15301
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1793407734233067e-05,
+      "loss": 0.4276,
+      "step": 15302
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1792830983072355e-05,
+      "loss": 0.5838,
+      "step": 15303
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1792254231911642e-05,
+      "loss": 0.397,
+      "step": 15304
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1791677480750932e-05,
+      "loss": 0.2972,
+      "step": 15305
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.179110072959022e-05,
+      "loss": 0.5045,
+      "step": 15306
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1790523978429507e-05,
+      "loss": 0.5884,
+      "step": 15307
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1789947227268796e-05,
+      "loss": 0.4333,
+      "step": 15308
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1789370476108084e-05,
+      "loss": 0.498,
+      "step": 15309
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1788793724947372e-05,
+      "loss": 0.4036,
+      "step": 15310
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1788216973786661e-05,
+      "loss": 0.5851,
+      "step": 15311
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1787640222625949e-05,
+      "loss": 0.6344,
+      "step": 15312
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1787063471465237e-05,
+      "loss": 0.3968,
+      "step": 15313
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1786486720304526e-05,
+      "loss": 0.5262,
+      "step": 15314
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1785909969143814e-05,
+      "loss": 0.4871,
+      "step": 15315
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1785333217983101e-05,
+      "loss": 0.6137,
+      "step": 15316
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.178475646682239e-05,
+      "loss": 0.4672,
+      "step": 15317
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1784179715661678e-05,
+      "loss": 0.4815,
+      "step": 15318
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1783602964500966e-05,
+      "loss": 0.4617,
+      "step": 15319
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1783026213340255e-05,
+      "loss": 0.5485,
+      "step": 15320
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1782449462179543e-05,
+      "loss": 0.4208,
+      "step": 15321
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.178187271101883e-05,
+      "loss": 0.5146,
+      "step": 15322
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1781295959858122e-05,
+      "loss": 0.4057,
+      "step": 15323
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1780719208697408e-05,
+      "loss": 0.5371,
+      "step": 15324
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1780142457536696e-05,
+      "loss": 0.5516,
+      "step": 15325
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1779565706375987e-05,
+      "loss": 0.4908,
+      "step": 15326
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1778988955215274e-05,
+      "loss": 0.5979,
+      "step": 15327
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1778412204054562e-05,
+      "loss": 0.4621,
+      "step": 15328
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1777835452893851e-05,
+      "loss": 0.4355,
+      "step": 15329
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1777258701733139e-05,
+      "loss": 0.6515,
+      "step": 15330
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1776681950572427e-05,
+      "loss": 0.4305,
+      "step": 15331
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1776105199411714e-05,
+      "loss": 0.4789,
+      "step": 15332
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1775528448251004e-05,
+      "loss": 0.396,
+      "step": 15333
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1774951697090292e-05,
+      "loss": 0.3756,
+      "step": 15334
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.177437494592958e-05,
+      "loss": 0.4392,
+      "step": 15335
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1773798194768869e-05,
+      "loss": 0.6156,
+      "step": 15336
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1773221443608156e-05,
+      "loss": 0.4819,
+      "step": 15337
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1772644692447444e-05,
+      "loss": 0.5488,
+      "step": 15338
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1772067941286733e-05,
+      "loss": 0.4726,
+      "step": 15339
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1771491190126021e-05,
+      "loss": 0.4789,
+      "step": 15340
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1770914438965309e-05,
+      "loss": 0.4339,
+      "step": 15341
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1770337687804598e-05,
+      "loss": 0.4137,
+      "step": 15342
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1769760936643886e-05,
+      "loss": 0.5027,
+      "step": 15343
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1769184185483173e-05,
+      "loss": 0.5856,
+      "step": 15344
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1768607434322463e-05,
+      "loss": 0.4706,
+      "step": 15345
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.176803068316175e-05,
+      "loss": 0.4638,
+      "step": 15346
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1767453932001038e-05,
+      "loss": 0.4473,
+      "step": 15347
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1766877180840328e-05,
+      "loss": 0.4823,
+      "step": 15348
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1766300429679615e-05,
+      "loss": 0.4369,
+      "step": 15349
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1765723678518903e-05,
+      "loss": 0.5038,
+      "step": 15350
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1765146927358192e-05,
+      "loss": 0.4327,
+      "step": 15351
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.176457017619748e-05,
+      "loss": 0.4502,
+      "step": 15352
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1763993425036768e-05,
+      "loss": 0.4681,
+      "step": 15353
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1763416673876059e-05,
+      "loss": 0.5232,
+      "step": 15354
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1762839922715345e-05,
+      "loss": 0.6554,
+      "step": 15355
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1762263171554632e-05,
+      "loss": 0.4176,
+      "step": 15356
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1761686420393924e-05,
+      "loss": 0.385,
+      "step": 15357
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1761109669233211e-05,
+      "loss": 0.4747,
+      "step": 15358
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1760532918072499e-05,
+      "loss": 0.4928,
+      "step": 15359
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1759956166911788e-05,
+      "loss": 0.493,
+      "step": 15360
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1759379415751076e-05,
+      "loss": 0.4709,
+      "step": 15361
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1758802664590364e-05,
+      "loss": 0.4452,
+      "step": 15362
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1758225913429653e-05,
+      "loss": 0.3773,
+      "step": 15363
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.175764916226894e-05,
+      "loss": 0.5242,
+      "step": 15364
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1757072411108228e-05,
+      "loss": 0.4275,
+      "step": 15365
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1756495659947516e-05,
+      "loss": 0.5957,
+      "step": 15366
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1755918908786805e-05,
+      "loss": 0.4583,
+      "step": 15367
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1755342157626093e-05,
+      "loss": 0.5599,
+      "step": 15368
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.175476540646538e-05,
+      "loss": 0.4842,
+      "step": 15369
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.175418865530467e-05,
+      "loss": 0.4446,
+      "step": 15370
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1753611904143958e-05,
+      "loss": 0.4724,
+      "step": 15371
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1753035152983246e-05,
+      "loss": 0.5069,
+      "step": 15372
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1752458401822535e-05,
+      "loss": 0.4372,
+      "step": 15373
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1751881650661823e-05,
+      "loss": 0.4931,
+      "step": 15374
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.175130489950111e-05,
+      "loss": 0.4197,
+      "step": 15375
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.17507281483404e-05,
+      "loss": 0.4484,
+      "step": 15376
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1750151397179687e-05,
+      "loss": 0.6208,
+      "step": 15377
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1749574646018975e-05,
+      "loss": 0.4874,
+      "step": 15378
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1748997894858264e-05,
+      "loss": 0.5067,
+      "step": 15379
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1748421143697552e-05,
+      "loss": 0.3477,
+      "step": 15380
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.174784439253684e-05,
+      "loss": 0.5171,
+      "step": 15381
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.174726764137613e-05,
+      "loss": 0.4649,
+      "step": 15382
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1746690890215417e-05,
+      "loss": 0.4535,
+      "step": 15383
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1746114139054705e-05,
+      "loss": 0.5162,
+      "step": 15384
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1745537387893996e-05,
+      "loss": 0.4497,
+      "step": 15385
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1744960636733282e-05,
+      "loss": 0.4449,
+      "step": 15386
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.174438388557257e-05,
+      "loss": 0.4397,
+      "step": 15387
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.174380713441186e-05,
+      "loss": 0.4546,
+      "step": 15388
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1743230383251148e-05,
+      "loss": 0.4346,
+      "step": 15389
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1742653632090436e-05,
+      "loss": 0.4958,
+      "step": 15390
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1742076880929725e-05,
+      "loss": 0.3857,
+      "step": 15391
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1741500129769013e-05,
+      "loss": 0.5449,
+      "step": 15392
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.17409233786083e-05,
+      "loss": 0.5678,
+      "step": 15393
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.174034662744759e-05,
+      "loss": 0.5191,
+      "step": 15394
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1739769876286878e-05,
+      "loss": 0.5628,
+      "step": 15395
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1739193125126165e-05,
+      "loss": 0.4263,
+      "step": 15396
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1738616373965455e-05,
+      "loss": 0.4437,
+      "step": 15397
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1738039622804742e-05,
+      "loss": 0.5545,
+      "step": 15398
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.173746287164403e-05,
+      "loss": 0.4292,
+      "step": 15399
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1736886120483318e-05,
+      "loss": 0.5596,
+      "step": 15400
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1736309369322607e-05,
+      "loss": 0.3975,
+      "step": 15401
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1735732618161895e-05,
+      "loss": 0.5641,
+      "step": 15402
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1735155867001182e-05,
+      "loss": 0.4992,
+      "step": 15403
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1734579115840472e-05,
+      "loss": 0.3632,
+      "step": 15404
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.173400236467976e-05,
+      "loss": 0.5036,
+      "step": 15405
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1733425613519047e-05,
+      "loss": 0.5273,
+      "step": 15406
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1732848862358337e-05,
+      "loss": 0.4945,
+      "step": 15407
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1732272111197624e-05,
+      "loss": 0.3641,
+      "step": 15408
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1731695360036912e-05,
+      "loss": 0.3927,
+      "step": 15409
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1731118608876201e-05,
+      "loss": 0.3828,
+      "step": 15410
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1730541857715489e-05,
+      "loss": 0.5296,
+      "step": 15411
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1729965106554777e-05,
+      "loss": 0.4399,
+      "step": 15412
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1729388355394066e-05,
+      "loss": 0.5386,
+      "step": 15413
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1728811604233354e-05,
+      "loss": 0.4588,
+      "step": 15414
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1728234853072641e-05,
+      "loss": 0.4948,
+      "step": 15415
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1727658101911933e-05,
+      "loss": 0.506,
+      "step": 15416
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1727081350751219e-05,
+      "loss": 0.4652,
+      "step": 15417
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1726504599590506e-05,
+      "loss": 0.2869,
+      "step": 15418
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1725927848429797e-05,
+      "loss": 0.4431,
+      "step": 15419
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1725351097269085e-05,
+      "loss": 0.4725,
+      "step": 15420
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1724774346108373e-05,
+      "loss": 0.5771,
+      "step": 15421
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1724197594947662e-05,
+      "loss": 0.5027,
+      "step": 15422
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.172362084378695e-05,
+      "loss": 0.6224,
+      "step": 15423
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1723044092626237e-05,
+      "loss": 0.5202,
+      "step": 15424
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1722467341465527e-05,
+      "loss": 0.4939,
+      "step": 15425
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1721890590304814e-05,
+      "loss": 0.4362,
+      "step": 15426
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1721313839144102e-05,
+      "loss": 0.4963,
+      "step": 15427
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1720737087983392e-05,
+      "loss": 0.5196,
+      "step": 15428
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.172016033682268e-05,
+      "loss": 0.4264,
+      "step": 15429
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1719583585661967e-05,
+      "loss": 0.3731,
+      "step": 15430
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1719006834501256e-05,
+      "loss": 0.5366,
+      "step": 15431
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1718430083340544e-05,
+      "loss": 0.5828,
+      "step": 15432
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1717853332179832e-05,
+      "loss": 0.4296,
+      "step": 15433
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1717276581019121e-05,
+      "loss": 0.4254,
+      "step": 15434
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1716699829858409e-05,
+      "loss": 0.5234,
+      "step": 15435
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1716123078697696e-05,
+      "loss": 0.6206,
+      "step": 15436
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1715546327536984e-05,
+      "loss": 0.4793,
+      "step": 15437
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1714969576376273e-05,
+      "loss": 0.5584,
+      "step": 15438
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1714392825215561e-05,
+      "loss": 0.3884,
+      "step": 15439
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1713816074054849e-05,
+      "loss": 0.5699,
+      "step": 15440
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1713239322894138e-05,
+      "loss": 0.4393,
+      "step": 15441
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1712662571733426e-05,
+      "loss": 0.5195,
+      "step": 15442
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1712085820572714e-05,
+      "loss": 0.7145,
+      "step": 15443
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1711509069412003e-05,
+      "loss": 0.3814,
+      "step": 15444
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.171093231825129e-05,
+      "loss": 0.5914,
+      "step": 15445
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1710355567090578e-05,
+      "loss": 0.5333,
+      "step": 15446
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.170977881592987e-05,
+      "loss": 0.4717,
+      "step": 15447
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1709202064769155e-05,
+      "loss": 0.5092,
+      "step": 15448
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1708625313608443e-05,
+      "loss": 0.4365,
+      "step": 15449
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1708048562447734e-05,
+      "loss": 0.3869,
+      "step": 15450
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1707471811287022e-05,
+      "loss": 0.462,
+      "step": 15451
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.170689506012631e-05,
+      "loss": 0.4424,
+      "step": 15452
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1706318308965599e-05,
+      "loss": 0.4659,
+      "step": 15453
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1705741557804887e-05,
+      "loss": 0.5666,
+      "step": 15454
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1705164806644174e-05,
+      "loss": 0.4169,
+      "step": 15455
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1704588055483464e-05,
+      "loss": 0.4678,
+      "step": 15456
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1704011304322751e-05,
+      "loss": 0.3667,
+      "step": 15457
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1703434553162039e-05,
+      "loss": 0.5223,
+      "step": 15458
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1702857802001328e-05,
+      "loss": 0.4856,
+      "step": 15459
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1702281050840616e-05,
+      "loss": 0.524,
+      "step": 15460
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 1.1701704299679904e-05,
+      "loss": 0.5858,
+      "step": 15461
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1701127548519193e-05,
+      "loss": 0.4348,
+      "step": 15462
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.170055079735848e-05,
+      "loss": 0.4419,
+      "step": 15463
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1699974046197769e-05,
+      "loss": 0.4839,
+      "step": 15464
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1699397295037058e-05,
+      "loss": 0.5257,
+      "step": 15465
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1698820543876346e-05,
+      "loss": 0.4147,
+      "step": 15466
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1698243792715633e-05,
+      "loss": 0.5741,
+      "step": 15467
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1697667041554923e-05,
+      "loss": 0.5318,
+      "step": 15468
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.169709029039421e-05,
+      "loss": 0.4724,
+      "step": 15469
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1696513539233498e-05,
+      "loss": 0.5482,
+      "step": 15470
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1695936788072786e-05,
+      "loss": 0.4216,
+      "step": 15471
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1695360036912075e-05,
+      "loss": 0.4922,
+      "step": 15472
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1694783285751363e-05,
+      "loss": 0.4476,
+      "step": 15473
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.169420653459065e-05,
+      "loss": 0.4489,
+      "step": 15474
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.169362978342994e-05,
+      "loss": 0.507,
+      "step": 15475
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1693053032269228e-05,
+      "loss": 0.4229,
+      "step": 15476
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1692476281108515e-05,
+      "loss": 0.5519,
+      "step": 15477
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1691899529947806e-05,
+      "loss": 0.5142,
+      "step": 15478
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1691322778787092e-05,
+      "loss": 0.5371,
+      "step": 15479
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.169074602762638e-05,
+      "loss": 0.4818,
+      "step": 15480
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1690169276465671e-05,
+      "loss": 0.5389,
+      "step": 15481
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1689592525304959e-05,
+      "loss": 0.5338,
+      "step": 15482
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1689015774144246e-05,
+      "loss": 0.5203,
+      "step": 15483
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1688439022983536e-05,
+      "loss": 0.4723,
+      "step": 15484
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1687862271822823e-05,
+      "loss": 0.4569,
+      "step": 15485
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1687285520662111e-05,
+      "loss": 0.5053,
+      "step": 15486
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.16867087695014e-05,
+      "loss": 0.3379,
+      "step": 15487
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1686132018340688e-05,
+      "loss": 0.5539,
+      "step": 15488
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1685555267179976e-05,
+      "loss": 0.4201,
+      "step": 15489
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1684978516019265e-05,
+      "loss": 0.5085,
+      "step": 15490
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1684401764858553e-05,
+      "loss": 0.4567,
+      "step": 15491
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.168382501369784e-05,
+      "loss": 0.5786,
+      "step": 15492
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.168324826253713e-05,
+      "loss": 0.443,
+      "step": 15493
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1682671511376418e-05,
+      "loss": 0.4654,
+      "step": 15494
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1682094760215705e-05,
+      "loss": 0.5429,
+      "step": 15495
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1681518009054995e-05,
+      "loss": 0.594,
+      "step": 15496
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1680941257894282e-05,
+      "loss": 0.4996,
+      "step": 15497
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.168036450673357e-05,
+      "loss": 0.4594,
+      "step": 15498
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.167978775557286e-05,
+      "loss": 0.4275,
+      "step": 15499
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1679211004412147e-05,
+      "loss": 0.5131,
+      "step": 15500
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1678634253251435e-05,
+      "loss": 0.4111,
+      "step": 15501
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1678057502090724e-05,
+      "loss": 0.505,
+      "step": 15502
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1677480750930012e-05,
+      "loss": 0.5387,
+      "step": 15503
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.16769039997693e-05,
+      "loss": 0.5247,
+      "step": 15504
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1676327248608587e-05,
+      "loss": 0.3977,
+      "step": 15505
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1675750497447877e-05,
+      "loss": 0.5046,
+      "step": 15506
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1675173746287164e-05,
+      "loss": 0.4378,
+      "step": 15507
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1674596995126452e-05,
+      "loss": 0.5776,
+      "step": 15508
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1674020243965743e-05,
+      "loss": 0.4847,
+      "step": 15509
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1673443492805029e-05,
+      "loss": 0.4379,
+      "step": 15510
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1672866741644317e-05,
+      "loss": 0.4262,
+      "step": 15511
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1672289990483608e-05,
+      "loss": 0.4328,
+      "step": 15512
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1671713239322896e-05,
+      "loss": 0.5771,
+      "step": 15513
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1671136488162183e-05,
+      "loss": 0.4897,
+      "step": 15514
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1670559737001473e-05,
+      "loss": 0.4456,
+      "step": 15515
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.166998298584076e-05,
+      "loss": 0.6377,
+      "step": 15516
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1669406234680048e-05,
+      "loss": 0.4108,
+      "step": 15517
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1668829483519337e-05,
+      "loss": 0.487,
+      "step": 15518
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1668252732358625e-05,
+      "loss": 0.509,
+      "step": 15519
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1667675981197913e-05,
+      "loss": 0.5395,
+      "step": 15520
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1667099230037202e-05,
+      "loss": 0.4803,
+      "step": 15521
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.166652247887649e-05,
+      "loss": 0.474,
+      "step": 15522
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1665945727715778e-05,
+      "loss": 0.4589,
+      "step": 15523
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1665368976555067e-05,
+      "loss": 0.5058,
+      "step": 15524
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1664792225394355e-05,
+      "loss": 0.4782,
+      "step": 15525
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1664215474233642e-05,
+      "loss": 0.4634,
+      "step": 15526
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1663638723072932e-05,
+      "loss": 0.4945,
+      "step": 15527
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.166306197191222e-05,
+      "loss": 0.4738,
+      "step": 15528
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1662485220751507e-05,
+      "loss": 0.4285,
+      "step": 15529
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1661908469590796e-05,
+      "loss": 0.4895,
+      "step": 15530
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1661331718430084e-05,
+      "loss": 0.3767,
+      "step": 15531
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1660754967269372e-05,
+      "loss": 0.424,
+      "step": 15532
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1660178216108661e-05,
+      "loss": 0.5677,
+      "step": 15533
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1659601464947949e-05,
+      "loss": 0.4667,
+      "step": 15534
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1659024713787237e-05,
+      "loss": 0.6205,
+      "step": 15535
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1658447962626528e-05,
+      "loss": 0.5272,
+      "step": 15536
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1657871211465814e-05,
+      "loss": 0.5085,
+      "step": 15537
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1657294460305101e-05,
+      "loss": 0.4952,
+      "step": 15538
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1656717709144392e-05,
+      "loss": 0.4601,
+      "step": 15539
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.165614095798368e-05,
+      "loss": 0.6026,
+      "step": 15540
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1655564206822968e-05,
+      "loss": 0.3432,
+      "step": 15541
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1654987455662254e-05,
+      "loss": 0.4897,
+      "step": 15542
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1654410704501545e-05,
+      "loss": 0.4145,
+      "step": 15543
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1653833953340832e-05,
+      "loss": 0.5589,
+      "step": 15544
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.165325720218012e-05,
+      "loss": 0.5845,
+      "step": 15545
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.165268045101941e-05,
+      "loss": 0.3427,
+      "step": 15546
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1652103699858697e-05,
+      "loss": 0.503,
+      "step": 15547
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1651526948697985e-05,
+      "loss": 0.5566,
+      "step": 15548
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1650950197537274e-05,
+      "loss": 0.5802,
+      "step": 15549
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1650373446376562e-05,
+      "loss": 0.4309,
+      "step": 15550
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.164979669521585e-05,
+      "loss": 0.5109,
+      "step": 15551
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1649219944055139e-05,
+      "loss": 0.4065,
+      "step": 15552
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1648643192894427e-05,
+      "loss": 0.5309,
+      "step": 15553
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1648066441733714e-05,
+      "loss": 0.476,
+      "step": 15554
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1647489690573004e-05,
+      "loss": 0.5364,
+      "step": 15555
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1646912939412291e-05,
+      "loss": 0.459,
+      "step": 15556
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1646336188251579e-05,
+      "loss": 0.4362,
+      "step": 15557
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1645759437090869e-05,
+      "loss": 0.5659,
+      "step": 15558
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1645182685930156e-05,
+      "loss": 0.4627,
+      "step": 15559
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1644605934769444e-05,
+      "loss": 0.5161,
+      "step": 15560
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1644029183608733e-05,
+      "loss": 0.4944,
+      "step": 15561
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1643452432448021e-05,
+      "loss": 0.4418,
+      "step": 15562
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1642875681287309e-05,
+      "loss": 0.4332,
+      "step": 15563
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1642298930126598e-05,
+      "loss": 0.5508,
+      "step": 15564
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1641722178965886e-05,
+      "loss": 0.4331,
+      "step": 15565
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1641145427805173e-05,
+      "loss": 0.4589,
+      "step": 15566
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1640568676644464e-05,
+      "loss": 0.3783,
+      "step": 15567
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.163999192548375e-05,
+      "loss": 0.4859,
+      "step": 15568
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1639415174323038e-05,
+      "loss": 0.5128,
+      "step": 15569
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.163883842316233e-05,
+      "loss": 0.5383,
+      "step": 15570
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1638261672001617e-05,
+      "loss": 0.4045,
+      "step": 15571
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1637684920840905e-05,
+      "loss": 0.5306,
+      "step": 15572
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1637108169680194e-05,
+      "loss": 0.4043,
+      "step": 15573
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1636531418519482e-05,
+      "loss": 0.5499,
+      "step": 15574
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.163595466735877e-05,
+      "loss": 0.5086,
+      "step": 15575
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1635377916198057e-05,
+      "loss": 0.4849,
+      "step": 15576
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1634801165037346e-05,
+      "loss": 0.4304,
+      "step": 15577
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1634224413876634e-05,
+      "loss": 0.4886,
+      "step": 15578
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1633647662715922e-05,
+      "loss": 0.4958,
+      "step": 15579
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1633070911555211e-05,
+      "loss": 0.518,
+      "step": 15580
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1632494160394499e-05,
+      "loss": 0.5039,
+      "step": 15581
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1631917409233787e-05,
+      "loss": 0.4504,
+      "step": 15582
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1631340658073076e-05,
+      "loss": 0.5796,
+      "step": 15583
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1630763906912364e-05,
+      "loss": 0.576,
+      "step": 15584
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1630187155751651e-05,
+      "loss": 0.4768,
+      "step": 15585
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.162961040459094e-05,
+      "loss": 0.4393,
+      "step": 15586
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1629033653430228e-05,
+      "loss": 0.445,
+      "step": 15587
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1628456902269516e-05,
+      "loss": 0.428,
+      "step": 15588
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1627880151108805e-05,
+      "loss": 0.5071,
+      "step": 15589
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1627303399948093e-05,
+      "loss": 0.4527,
+      "step": 15590
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.162672664878738e-05,
+      "loss": 0.4036,
+      "step": 15591
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.162614989762667e-05,
+      "loss": 0.4693,
+      "step": 15592
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1625573146465958e-05,
+      "loss": 0.4619,
+      "step": 15593
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1624996395305246e-05,
+      "loss": 0.4775,
+      "step": 15594
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1624419644144535e-05,
+      "loss": 0.4467,
+      "step": 15595
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1623842892983823e-05,
+      "loss": 0.4976,
+      "step": 15596
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.162326614182311e-05,
+      "loss": 0.5341,
+      "step": 15597
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1622689390662401e-05,
+      "loss": 0.424,
+      "step": 15598
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1622112639501687e-05,
+      "loss": 0.4053,
+      "step": 15599
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1621535888340975e-05,
+      "loss": 0.4147,
+      "step": 15600
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1620959137180266e-05,
+      "loss": 0.5816,
+      "step": 15601
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1620382386019554e-05,
+      "loss": 0.5583,
+      "step": 15602
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1619805634858841e-05,
+      "loss": 0.4374,
+      "step": 15603
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.161922888369813e-05,
+      "loss": 0.5012,
+      "step": 15604
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1618652132537419e-05,
+      "loss": 0.4872,
+      "step": 15605
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1618075381376706e-05,
+      "loss": 0.528,
+      "step": 15606
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1617498630215996e-05,
+      "loss": 0.364,
+      "step": 15607
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1616921879055283e-05,
+      "loss": 0.4697,
+      "step": 15608
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1616345127894571e-05,
+      "loss": 0.4896,
+      "step": 15609
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1615768376733859e-05,
+      "loss": 0.4322,
+      "step": 15610
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1615191625573148e-05,
+      "loss": 0.5508,
+      "step": 15611
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1614614874412436e-05,
+      "loss": 0.5779,
+      "step": 15612
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1614038123251723e-05,
+      "loss": 0.5971,
+      "step": 15613
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1613461372091013e-05,
+      "loss": 0.4125,
+      "step": 15614
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.16128846209303e-05,
+      "loss": 0.4369,
+      "step": 15615
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1612307869769588e-05,
+      "loss": 0.5596,
+      "step": 15616
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1611731118608878e-05,
+      "loss": 0.5141,
+      "step": 15617
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1611154367448165e-05,
+      "loss": 0.496,
+      "step": 15618
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1610577616287453e-05,
+      "loss": 0.4959,
+      "step": 15619
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1610000865126742e-05,
+      "loss": 0.4835,
+      "step": 15620
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.160942411396603e-05,
+      "loss": 0.3974,
+      "step": 15621
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1608847362805318e-05,
+      "loss": 0.5173,
+      "step": 15622
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1608270611644607e-05,
+      "loss": 0.4517,
+      "step": 15623
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1607693860483895e-05,
+      "loss": 0.4544,
+      "step": 15624
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1607117109323182e-05,
+      "loss": 0.5189,
+      "step": 15625
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1606540358162472e-05,
+      "loss": 0.4715,
+      "step": 15626
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.160596360700176e-05,
+      "loss": 0.4551,
+      "step": 15627
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1605386855841047e-05,
+      "loss": 0.3837,
+      "step": 15628
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1604810104680338e-05,
+      "loss": 0.5228,
+      "step": 15629
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1604233353519624e-05,
+      "loss": 0.4812,
+      "step": 15630
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1603656602358912e-05,
+      "loss": 0.5533,
+      "step": 15631
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1603079851198203e-05,
+      "loss": 0.3869,
+      "step": 15632
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.160250310003749e-05,
+      "loss": 0.4762,
+      "step": 15633
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1601926348876778e-05,
+      "loss": 0.4569,
+      "step": 15634
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1601349597716068e-05,
+      "loss": 0.5215,
+      "step": 15635
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1600772846555355e-05,
+      "loss": 0.4704,
+      "step": 15636
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1600196095394643e-05,
+      "loss": 0.4723,
+      "step": 15637
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1599619344233932e-05,
+      "loss": 0.3963,
+      "step": 15638
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.159904259307322e-05,
+      "loss": 0.5749,
+      "step": 15639
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.1598465841912508e-05,
+      "loss": 0.5571,
+      "step": 15640
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1597889090751797e-05,
+      "loss": 0.5896,
+      "step": 15641
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1597312339591085e-05,
+      "loss": 0.4232,
+      "step": 15642
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1596735588430373e-05,
+      "loss": 0.65,
+      "step": 15643
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.159615883726966e-05,
+      "loss": 0.4678,
+      "step": 15644
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.159558208610895e-05,
+      "loss": 0.5783,
+      "step": 15645
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1595005334948237e-05,
+      "loss": 0.5156,
+      "step": 15646
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1594428583787525e-05,
+      "loss": 0.4745,
+      "step": 15647
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1593851832626814e-05,
+      "loss": 0.4949,
+      "step": 15648
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1593275081466102e-05,
+      "loss": 0.5318,
+      "step": 15649
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.159269833030539e-05,
+      "loss": 0.5314,
+      "step": 15650
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1592121579144679e-05,
+      "loss": 0.5144,
+      "step": 15651
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1591544827983967e-05,
+      "loss": 0.5905,
+      "step": 15652
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1590968076823255e-05,
+      "loss": 0.3822,
+      "step": 15653
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1590391325662544e-05,
+      "loss": 0.5757,
+      "step": 15654
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1589814574501832e-05,
+      "loss": 0.5269,
+      "step": 15655
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.158923782334112e-05,
+      "loss": 0.5308,
+      "step": 15656
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1588661072180409e-05,
+      "loss": 0.4928,
+      "step": 15657
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1588084321019696e-05,
+      "loss": 0.4408,
+      "step": 15658
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1587507569858984e-05,
+      "loss": 0.5131,
+      "step": 15659
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1586930818698275e-05,
+      "loss": 0.4781,
+      "step": 15660
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1586354067537561e-05,
+      "loss": 0.4859,
+      "step": 15661
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1585777316376849e-05,
+      "loss": 0.4932,
+      "step": 15662
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.158520056521614e-05,
+      "loss": 0.655,
+      "step": 15663
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1584623814055428e-05,
+      "loss": 0.3976,
+      "step": 15664
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1584047062894715e-05,
+      "loss": 0.3908,
+      "step": 15665
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1583470311734005e-05,
+      "loss": 0.5192,
+      "step": 15666
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1582893560573292e-05,
+      "loss": 0.4158,
+      "step": 15667
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.158231680941258e-05,
+      "loss": 0.5137,
+      "step": 15668
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.158174005825187e-05,
+      "loss": 0.5193,
+      "step": 15669
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1581163307091157e-05,
+      "loss": 0.4561,
+      "step": 15670
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1580586555930445e-05,
+      "loss": 0.4993,
+      "step": 15671
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1580009804769734e-05,
+      "loss": 0.4192,
+      "step": 15672
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1579433053609022e-05,
+      "loss": 0.4447,
+      "step": 15673
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.157885630244831e-05,
+      "loss": 0.6202,
+      "step": 15674
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1578279551287599e-05,
+      "loss": 0.565,
+      "step": 15675
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1577702800126887e-05,
+      "loss": 0.4845,
+      "step": 15676
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1577126048966174e-05,
+      "loss": 0.4791,
+      "step": 15677
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1576549297805464e-05,
+      "loss": 0.4926,
+      "step": 15678
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1575972546644751e-05,
+      "loss": 0.4696,
+      "step": 15679
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1575395795484039e-05,
+      "loss": 0.529,
+      "step": 15680
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1574819044323327e-05,
+      "loss": 0.4752,
+      "step": 15681
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1574242293162616e-05,
+      "loss": 0.4089,
+      "step": 15682
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1573665542001904e-05,
+      "loss": 0.4207,
+      "step": 15683
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1573088790841191e-05,
+      "loss": 0.5552,
+      "step": 15684
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.157251203968048e-05,
+      "loss": 0.4943,
+      "step": 15685
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1571935288519768e-05,
+      "loss": 0.5326,
+      "step": 15686
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1571358537359056e-05,
+      "loss": 0.5462,
+      "step": 15687
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1570781786198346e-05,
+      "loss": 0.4857,
+      "step": 15688
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1570205035037633e-05,
+      "loss": 0.4348,
+      "step": 15689
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1569628283876921e-05,
+      "loss": 0.6679,
+      "step": 15690
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1569051532716212e-05,
+      "loss": 0.4821,
+      "step": 15691
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1568474781555498e-05,
+      "loss": 0.5126,
+      "step": 15692
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1567898030394786e-05,
+      "loss": 0.427,
+      "step": 15693
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1567321279234077e-05,
+      "loss": 0.4702,
+      "step": 15694
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1566744528073364e-05,
+      "loss": 0.5333,
+      "step": 15695
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1566167776912652e-05,
+      "loss": 0.6333,
+      "step": 15696
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1565591025751941e-05,
+      "loss": 0.561,
+      "step": 15697
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1565014274591229e-05,
+      "loss": 0.4462,
+      "step": 15698
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1564437523430517e-05,
+      "loss": 0.4282,
+      "step": 15699
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1563860772269806e-05,
+      "loss": 0.6043,
+      "step": 15700
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1563284021109094e-05,
+      "loss": 0.4901,
+      "step": 15701
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1562707269948382e-05,
+      "loss": 0.6293,
+      "step": 15702
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1562130518787671e-05,
+      "loss": 0.3786,
+      "step": 15703
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1561553767626959e-05,
+      "loss": 0.6244,
+      "step": 15704
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1560977016466246e-05,
+      "loss": 0.5831,
+      "step": 15705
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1560400265305536e-05,
+      "loss": 0.5773,
+      "step": 15706
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1559823514144823e-05,
+      "loss": 0.5218,
+      "step": 15707
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1559246762984111e-05,
+      "loss": 0.5779,
+      "step": 15708
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.15586700118234e-05,
+      "loss": 0.4972,
+      "step": 15709
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1558093260662688e-05,
+      "loss": 0.4326,
+      "step": 15710
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1557516509501976e-05,
+      "loss": 0.474,
+      "step": 15711
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1556939758341265e-05,
+      "loss": 0.5002,
+      "step": 15712
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1556363007180553e-05,
+      "loss": 0.5936,
+      "step": 15713
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.155578625601984e-05,
+      "loss": 0.4769,
+      "step": 15714
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1555209504859128e-05,
+      "loss": 0.5632,
+      "step": 15715
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1554632753698418e-05,
+      "loss": 0.5432,
+      "step": 15716
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1554056002537705e-05,
+      "loss": 0.4814,
+      "step": 15717
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1553479251376993e-05,
+      "loss": 0.5811,
+      "step": 15718
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1552902500216282e-05,
+      "loss": 0.4906,
+      "step": 15719
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.155232574905557e-05,
+      "loss": 0.5404,
+      "step": 15720
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1551748997894858e-05,
+      "loss": 0.5212,
+      "step": 15721
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1551172246734149e-05,
+      "loss": 0.5268,
+      "step": 15722
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1550595495573435e-05,
+      "loss": 0.4452,
+      "step": 15723
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1550018744412722e-05,
+      "loss": 0.4249,
+      "step": 15724
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1549441993252014e-05,
+      "loss": 0.5082,
+      "step": 15725
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1548865242091301e-05,
+      "loss": 0.4546,
+      "step": 15726
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1548288490930589e-05,
+      "loss": 0.3864,
+      "step": 15727
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1547711739769878e-05,
+      "loss": 0.4964,
+      "step": 15728
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1547134988609166e-05,
+      "loss": 0.4051,
+      "step": 15729
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1546558237448454e-05,
+      "loss": 0.4375,
+      "step": 15730
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1545981486287743e-05,
+      "loss": 0.5601,
+      "step": 15731
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.154540473512703e-05,
+      "loss": 0.3323,
+      "step": 15732
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1544827983966318e-05,
+      "loss": 0.461,
+      "step": 15733
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1544251232805608e-05,
+      "loss": 0.4757,
+      "step": 15734
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1543674481644896e-05,
+      "loss": 0.48,
+      "step": 15735
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1543097730484183e-05,
+      "loss": 0.5451,
+      "step": 15736
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1542520979323473e-05,
+      "loss": 0.431,
+      "step": 15737
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.154194422816276e-05,
+      "loss": 0.4242,
+      "step": 15738
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1541367477002048e-05,
+      "loss": 0.5396,
+      "step": 15739
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1540790725841337e-05,
+      "loss": 0.379,
+      "step": 15740
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1540213974680625e-05,
+      "loss": 0.4374,
+      "step": 15741
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1539637223519913e-05,
+      "loss": 0.4616,
+      "step": 15742
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1539060472359202e-05,
+      "loss": 0.5167,
+      "step": 15743
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.153848372119849e-05,
+      "loss": 0.5077,
+      "step": 15744
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1537906970037777e-05,
+      "loss": 0.4802,
+      "step": 15745
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1537330218877067e-05,
+      "loss": 0.4345,
+      "step": 15746
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1536753467716354e-05,
+      "loss": 0.4907,
+      "step": 15747
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1536176716555642e-05,
+      "loss": 0.576,
+      "step": 15748
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.153559996539493e-05,
+      "loss": 0.5525,
+      "step": 15749
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.153502321423422e-05,
+      "loss": 0.5203,
+      "step": 15750
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1534446463073507e-05,
+      "loss": 0.5028,
+      "step": 15751
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1533869711912795e-05,
+      "loss": 0.5716,
+      "step": 15752
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1533292960752086e-05,
+      "loss": 0.493,
+      "step": 15753
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1532716209591372e-05,
+      "loss": 0.4438,
+      "step": 15754
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.153213945843066e-05,
+      "loss": 0.5137,
+      "step": 15755
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.153156270726995e-05,
+      "loss": 0.4439,
+      "step": 15756
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1530985956109238e-05,
+      "loss": 0.4051,
+      "step": 15757
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1530409204948526e-05,
+      "loss": 0.3664,
+      "step": 15758
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1529832453787815e-05,
+      "loss": 0.4751,
+      "step": 15759
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1529255702627103e-05,
+      "loss": 0.4598,
+      "step": 15760
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.152867895146639e-05,
+      "loss": 0.471,
+      "step": 15761
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.152810220030568e-05,
+      "loss": 0.4299,
+      "step": 15762
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1527525449144968e-05,
+      "loss": 0.4269,
+      "step": 15763
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1526948697984255e-05,
+      "loss": 0.5954,
+      "step": 15764
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1526371946823545e-05,
+      "loss": 0.4144,
+      "step": 15765
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1525795195662832e-05,
+      "loss": 0.6315,
+      "step": 15766
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.152521844450212e-05,
+      "loss": 0.4709,
+      "step": 15767
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.152464169334141e-05,
+      "loss": 0.4494,
+      "step": 15768
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1524064942180697e-05,
+      "loss": 0.3681,
+      "step": 15769
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1523488191019985e-05,
+      "loss": 0.5519,
+      "step": 15770
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1522911439859274e-05,
+      "loss": 0.4858,
+      "step": 15771
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1522334688698562e-05,
+      "loss": 0.5434,
+      "step": 15772
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.152175793753785e-05,
+      "loss": 0.5807,
+      "step": 15773
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1521181186377139e-05,
+      "loss": 0.4702,
+      "step": 15774
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1520604435216427e-05,
+      "loss": 0.5349,
+      "step": 15775
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1520027684055714e-05,
+      "loss": 0.5393,
+      "step": 15776
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1519450932895004e-05,
+      "loss": 0.4003,
+      "step": 15777
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1518874181734291e-05,
+      "loss": 0.3826,
+      "step": 15778
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1518297430573579e-05,
+      "loss": 0.4298,
+      "step": 15779
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.151772067941287e-05,
+      "loss": 0.3673,
+      "step": 15780
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1517143928252156e-05,
+      "loss": 0.5916,
+      "step": 15781
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1516567177091444e-05,
+      "loss": 0.4881,
+      "step": 15782
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1515990425930731e-05,
+      "loss": 0.5682,
+      "step": 15783
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1515413674770023e-05,
+      "loss": 0.5708,
+      "step": 15784
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1514836923609309e-05,
+      "loss": 0.5505,
+      "step": 15785
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1514260172448596e-05,
+      "loss": 0.4841,
+      "step": 15786
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1513683421287887e-05,
+      "loss": 0.5753,
+      "step": 15787
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1513106670127175e-05,
+      "loss": 0.5226,
+      "step": 15788
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1512529918966463e-05,
+      "loss": 0.5206,
+      "step": 15789
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1511953167805752e-05,
+      "loss": 0.5689,
+      "step": 15790
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.151137641664504e-05,
+      "loss": 0.4092,
+      "step": 15791
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1510799665484327e-05,
+      "loss": 0.4368,
+      "step": 15792
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1510222914323617e-05,
+      "loss": 0.5148,
+      "step": 15793
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1509646163162904e-05,
+      "loss": 0.453,
+      "step": 15794
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1509069412002192e-05,
+      "loss": 0.4639,
+      "step": 15795
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1508492660841482e-05,
+      "loss": 0.4593,
+      "step": 15796
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.150791590968077e-05,
+      "loss": 0.501,
+      "step": 15797
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1507339158520057e-05,
+      "loss": 0.4007,
+      "step": 15798
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1506762407359346e-05,
+      "loss": 0.4867,
+      "step": 15799
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1506185656198634e-05,
+      "loss": 0.4068,
+      "step": 15800
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1505608905037922e-05,
+      "loss": 0.5698,
+      "step": 15801
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1505032153877211e-05,
+      "loss": 0.387,
+      "step": 15802
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1504455402716499e-05,
+      "loss": 0.4253,
+      "step": 15803
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1503878651555786e-05,
+      "loss": 0.4903,
+      "step": 15804
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1503301900395076e-05,
+      "loss": 0.4974,
+      "step": 15805
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1502725149234363e-05,
+      "loss": 0.3833,
+      "step": 15806
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1502148398073651e-05,
+      "loss": 0.3116,
+      "step": 15807
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.150157164691294e-05,
+      "loss": 0.6123,
+      "step": 15808
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1500994895752228e-05,
+      "loss": 0.4154,
+      "step": 15809
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1500418144591516e-05,
+      "loss": 0.5324,
+      "step": 15810
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1499841393430807e-05,
+      "loss": 0.5337,
+      "step": 15811
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1499264642270093e-05,
+      "loss": 0.5488,
+      "step": 15812
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.149868789110938e-05,
+      "loss": 0.4122,
+      "step": 15813
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1498111139948672e-05,
+      "loss": 0.464,
+      "step": 15814
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.149753438878796e-05,
+      "loss": 0.4828,
+      "step": 15815
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1496957637627245e-05,
+      "loss": 0.5618,
+      "step": 15816
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1496380886466537e-05,
+      "loss": 0.3921,
+      "step": 15817
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1495804135305824e-05,
+      "loss": 0.4981,
+      "step": 15818
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 1.1495227384145112e-05,
+      "loss": 0.4068,
+      "step": 15819
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.14946506329844e-05,
+      "loss": 0.6107,
+      "step": 15820
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1494073881823689e-05,
+      "loss": 0.4617,
+      "step": 15821
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1493497130662977e-05,
+      "loss": 0.4433,
+      "step": 15822
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1492920379502264e-05,
+      "loss": 0.5401,
+      "step": 15823
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1492343628341554e-05,
+      "loss": 0.5237,
+      "step": 15824
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1491766877180841e-05,
+      "loss": 0.3816,
+      "step": 15825
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1491190126020129e-05,
+      "loss": 0.4313,
+      "step": 15826
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1490613374859418e-05,
+      "loss": 0.3321,
+      "step": 15827
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1490036623698706e-05,
+      "loss": 0.4442,
+      "step": 15828
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1489459872537994e-05,
+      "loss": 0.4491,
+      "step": 15829
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1488883121377283e-05,
+      "loss": 0.4933,
+      "step": 15830
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1488306370216571e-05,
+      "loss": 0.5995,
+      "step": 15831
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1487729619055859e-05,
+      "loss": 0.4234,
+      "step": 15832
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1487152867895148e-05,
+      "loss": 0.4709,
+      "step": 15833
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1486576116734436e-05,
+      "loss": 0.4297,
+      "step": 15834
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1485999365573723e-05,
+      "loss": 0.5325,
+      "step": 15835
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1485422614413013e-05,
+      "loss": 0.4717,
+      "step": 15836
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.14848458632523e-05,
+      "loss": 0.5348,
+      "step": 15837
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1484269112091588e-05,
+      "loss": 0.5725,
+      "step": 15838
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1483692360930877e-05,
+      "loss": 0.5103,
+      "step": 15839
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1483115609770165e-05,
+      "loss": 0.341,
+      "step": 15840
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1482538858609453e-05,
+      "loss": 0.4292,
+      "step": 15841
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1481962107448744e-05,
+      "loss": 0.4418,
+      "step": 15842
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.148138535628803e-05,
+      "loss": 0.4577,
+      "step": 15843
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1480808605127318e-05,
+      "loss": 0.4776,
+      "step": 15844
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1480231853966609e-05,
+      "loss": 0.4686,
+      "step": 15845
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1479655102805896e-05,
+      "loss": 0.5812,
+      "step": 15846
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1479078351645182e-05,
+      "loss": 0.4131,
+      "step": 15847
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1478501600484473e-05,
+      "loss": 0.3856,
+      "step": 15848
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1477924849323761e-05,
+      "loss": 0.5421,
+      "step": 15849
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1477348098163049e-05,
+      "loss": 0.4249,
+      "step": 15850
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1476771347002338e-05,
+      "loss": 0.4018,
+      "step": 15851
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1476194595841626e-05,
+      "loss": 0.5201,
+      "step": 15852
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1475617844680913e-05,
+      "loss": 0.4437,
+      "step": 15853
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1475041093520201e-05,
+      "loss": 0.413,
+      "step": 15854
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.147446434235949e-05,
+      "loss": 0.4785,
+      "step": 15855
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1473887591198778e-05,
+      "loss": 0.5078,
+      "step": 15856
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1473310840038066e-05,
+      "loss": 0.4271,
+      "step": 15857
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1472734088877355e-05,
+      "loss": 0.5106,
+      "step": 15858
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1472157337716643e-05,
+      "loss": 0.5005,
+      "step": 15859
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.147158058655593e-05,
+      "loss": 0.4507,
+      "step": 15860
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.147100383539522e-05,
+      "loss": 0.5244,
+      "step": 15861
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1470427084234508e-05,
+      "loss": 0.5396,
+      "step": 15862
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1469850333073795e-05,
+      "loss": 0.5043,
+      "step": 15863
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1469273581913085e-05,
+      "loss": 0.5067,
+      "step": 15864
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1468696830752372e-05,
+      "loss": 0.5843,
+      "step": 15865
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146812007959166e-05,
+      "loss": 0.5738,
+      "step": 15866
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146754332843095e-05,
+      "loss": 0.4845,
+      "step": 15867
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1466966577270237e-05,
+      "loss": 0.5162,
+      "step": 15868
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1466389826109525e-05,
+      "loss": 0.4525,
+      "step": 15869
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1465813074948814e-05,
+      "loss": 0.5848,
+      "step": 15870
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1465236323788102e-05,
+      "loss": 0.4058,
+      "step": 15871
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146465957262739e-05,
+      "loss": 0.4077,
+      "step": 15872
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146408282146668e-05,
+      "loss": 0.4723,
+      "step": 15873
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1463506070305967e-05,
+      "loss": 0.4472,
+      "step": 15874
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1462929319145254e-05,
+      "loss": 0.5335,
+      "step": 15875
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1462352567984545e-05,
+      "loss": 0.5201,
+      "step": 15876
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1461775816823833e-05,
+      "loss": 0.5325,
+      "step": 15877
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146119906566312e-05,
+      "loss": 0.4758,
+      "step": 15878
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.146062231450241e-05,
+      "loss": 0.5353,
+      "step": 15879
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1460045563341698e-05,
+      "loss": 0.4943,
+      "step": 15880
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1459468812180986e-05,
+      "loss": 0.4973,
+      "step": 15881
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1458892061020275e-05,
+      "loss": 0.5272,
+      "step": 15882
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1458315309859563e-05,
+      "loss": 0.4385,
+      "step": 15883
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.145773855869885e-05,
+      "loss": 0.4795,
+      "step": 15884
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.145716180753814e-05,
+      "loss": 0.6029,
+      "step": 15885
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1456585056377427e-05,
+      "loss": 0.4447,
+      "step": 15886
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1456008305216715e-05,
+      "loss": 0.4752,
+      "step": 15887
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1455431554056003e-05,
+      "loss": 0.5237,
+      "step": 15888
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1454854802895292e-05,
+      "loss": 0.409,
+      "step": 15889
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.145427805173458e-05,
+      "loss": 0.5068,
+      "step": 15890
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1453701300573868e-05,
+      "loss": 0.5861,
+      "step": 15891
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1453124549413157e-05,
+      "loss": 0.4743,
+      "step": 15892
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1452547798252445e-05,
+      "loss": 0.4664,
+      "step": 15893
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1451971047091732e-05,
+      "loss": 0.3883,
+      "step": 15894
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1451394295931022e-05,
+      "loss": 0.3928,
+      "step": 15895
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.145081754477031e-05,
+      "loss": 0.4174,
+      "step": 15896
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1450240793609597e-05,
+      "loss": 0.4743,
+      "step": 15897
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1449664042448886e-05,
+      "loss": 0.4617,
+      "step": 15898
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1449087291288174e-05,
+      "loss": 0.5059,
+      "step": 15899
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1448510540127462e-05,
+      "loss": 0.6064,
+      "step": 15900
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1447933788966751e-05,
+      "loss": 0.5551,
+      "step": 15901
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1447357037806039e-05,
+      "loss": 0.4156,
+      "step": 15902
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1446780286645327e-05,
+      "loss": 0.5309,
+      "step": 15903
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1446203535484618e-05,
+      "loss": 0.5059,
+      "step": 15904
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1445626784323904e-05,
+      "loss": 0.4664,
+      "step": 15905
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1445050033163191e-05,
+      "loss": 0.4409,
+      "step": 15906
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1444473282002482e-05,
+      "loss": 0.5047,
+      "step": 15907
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.144389653084177e-05,
+      "loss": 0.495,
+      "step": 15908
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1443319779681058e-05,
+      "loss": 0.4574,
+      "step": 15909
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1442743028520347e-05,
+      "loss": 0.6223,
+      "step": 15910
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1442166277359635e-05,
+      "loss": 0.4905,
+      "step": 15911
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1441589526198922e-05,
+      "loss": 0.5722,
+      "step": 15912
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1441012775038212e-05,
+      "loss": 0.5046,
+      "step": 15913
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.14404360238775e-05,
+      "loss": 0.4328,
+      "step": 15914
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1439859272716787e-05,
+      "loss": 0.4971,
+      "step": 15915
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1439282521556077e-05,
+      "loss": 0.3763,
+      "step": 15916
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1438705770395364e-05,
+      "loss": 0.4646,
+      "step": 15917
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1438129019234652e-05,
+      "loss": 0.5226,
+      "step": 15918
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1437552268073941e-05,
+      "loss": 0.4799,
+      "step": 15919
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1436975516913229e-05,
+      "loss": 0.5579,
+      "step": 15920
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1436398765752517e-05,
+      "loss": 0.5332,
+      "step": 15921
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1435822014591806e-05,
+      "loss": 0.4285,
+      "step": 15922
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1435245263431094e-05,
+      "loss": 0.5306,
+      "step": 15923
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1434668512270381e-05,
+      "loss": 0.4202,
+      "step": 15924
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.143409176110967e-05,
+      "loss": 0.4932,
+      "step": 15925
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1433515009948959e-05,
+      "loss": 0.5145,
+      "step": 15926
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1432938258788246e-05,
+      "loss": 0.5303,
+      "step": 15927
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1432361507627534e-05,
+      "loss": 0.5816,
+      "step": 15928
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1431784756466823e-05,
+      "loss": 0.5585,
+      "step": 15929
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1431208005306111e-05,
+      "loss": 0.4261,
+      "step": 15930
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1430631254145399e-05,
+      "loss": 0.4881,
+      "step": 15931
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1430054502984688e-05,
+      "loss": 0.4787,
+      "step": 15932
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1429477751823976e-05,
+      "loss": 0.5424,
+      "step": 15933
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1428901000663263e-05,
+      "loss": 0.6219,
+      "step": 15934
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1428324249502554e-05,
+      "loss": 0.4448,
+      "step": 15935
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.142774749834184e-05,
+      "loss": 0.4725,
+      "step": 15936
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1427170747181128e-05,
+      "loss": 0.3906,
+      "step": 15937
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.142659399602042e-05,
+      "loss": 0.5067,
+      "step": 15938
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1426017244859707e-05,
+      "loss": 0.5721,
+      "step": 15939
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1425440493698995e-05,
+      "loss": 0.5869,
+      "step": 15940
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1424863742538284e-05,
+      "loss": 0.5866,
+      "step": 15941
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1424286991377572e-05,
+      "loss": 0.4903,
+      "step": 15942
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.142371024021686e-05,
+      "loss": 0.4801,
+      "step": 15943
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1423133489056149e-05,
+      "loss": 0.5958,
+      "step": 15944
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1422556737895436e-05,
+      "loss": 0.4449,
+      "step": 15945
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1421979986734724e-05,
+      "loss": 0.4574,
+      "step": 15946
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1421403235574013e-05,
+      "loss": 0.4099,
+      "step": 15947
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1420826484413301e-05,
+      "loss": 0.5971,
+      "step": 15948
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1420249733252589e-05,
+      "loss": 0.4064,
+      "step": 15949
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1419672982091878e-05,
+      "loss": 0.5083,
+      "step": 15950
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1419096230931166e-05,
+      "loss": 0.4775,
+      "step": 15951
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1418519479770454e-05,
+      "loss": 0.4029,
+      "step": 15952
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1417942728609743e-05,
+      "loss": 0.4232,
+      "step": 15953
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.141736597744903e-05,
+      "loss": 0.5085,
+      "step": 15954
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1416789226288318e-05,
+      "loss": 0.5076,
+      "step": 15955
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1416212475127608e-05,
+      "loss": 0.4868,
+      "step": 15956
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1415635723966895e-05,
+      "loss": 0.4747,
+      "step": 15957
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1415058972806183e-05,
+      "loss": 0.667,
+      "step": 15958
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.141448222164547e-05,
+      "loss": 0.5103,
+      "step": 15959
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.141390547048476e-05,
+      "loss": 0.5287,
+      "step": 15960
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1413328719324048e-05,
+      "loss": 0.4146,
+      "step": 15961
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1412751968163336e-05,
+      "loss": 0.5171,
+      "step": 15962
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1412175217002625e-05,
+      "loss": 0.554,
+      "step": 15963
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1411598465841913e-05,
+      "loss": 0.4716,
+      "step": 15964
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.14110217146812e-05,
+      "loss": 0.4357,
+      "step": 15965
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1410444963520491e-05,
+      "loss": 0.5416,
+      "step": 15966
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1409868212359777e-05,
+      "loss": 0.5037,
+      "step": 15967
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1409291461199065e-05,
+      "loss": 0.4425,
+      "step": 15968
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1408714710038356e-05,
+      "loss": 0.3717,
+      "step": 15969
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1408137958877644e-05,
+      "loss": 0.4897,
+      "step": 15970
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1407561207716931e-05,
+      "loss": 0.438,
+      "step": 15971
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1406984456556221e-05,
+      "loss": 0.5448,
+      "step": 15972
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1406407705395509e-05,
+      "loss": 0.5058,
+      "step": 15973
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1405830954234796e-05,
+      "loss": 0.5092,
+      "step": 15974
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1405254203074086e-05,
+      "loss": 0.4766,
+      "step": 15975
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1404677451913373e-05,
+      "loss": 0.4213,
+      "step": 15976
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1404100700752661e-05,
+      "loss": 0.5926,
+      "step": 15977
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.140352394959195e-05,
+      "loss": 0.5524,
+      "step": 15978
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1402947198431238e-05,
+      "loss": 0.4697,
+      "step": 15979
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1402370447270526e-05,
+      "loss": 0.5482,
+      "step": 15980
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1401793696109815e-05,
+      "loss": 0.5927,
+      "step": 15981
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1401216944949103e-05,
+      "loss": 0.3614,
+      "step": 15982
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.140064019378839e-05,
+      "loss": 0.5961,
+      "step": 15983
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.140006344262768e-05,
+      "loss": 0.48,
+      "step": 15984
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1399486691466968e-05,
+      "loss": 0.4329,
+      "step": 15985
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1398909940306255e-05,
+      "loss": 0.5257,
+      "step": 15986
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1398333189145545e-05,
+      "loss": 0.4224,
+      "step": 15987
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1397756437984832e-05,
+      "loss": 0.5066,
+      "step": 15988
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.139717968682412e-05,
+      "loss": 0.5199,
+      "step": 15989
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.139660293566341e-05,
+      "loss": 0.5786,
+      "step": 15990
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1396026184502697e-05,
+      "loss": 0.5803,
+      "step": 15991
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1395449433341985e-05,
+      "loss": 0.615,
+      "step": 15992
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1394872682181272e-05,
+      "loss": 0.4597,
+      "step": 15993
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1394295931020562e-05,
+      "loss": 0.5086,
+      "step": 15994
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.139371917985985e-05,
+      "loss": 0.5734,
+      "step": 15995
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1393142428699137e-05,
+      "loss": 0.5046,
+      "step": 15996
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1392565677538428e-05,
+      "loss": 0.5252,
+      "step": 15997
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 1.1391988926377714e-05,
+      "loss": 0.3733,
+      "step": 15998
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1391412175217002e-05,
+      "loss": 0.4405,
+      "step": 15999
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1390835424056293e-05,
+      "loss": 0.5781,
+      "step": 16000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.139025867289558e-05,
+      "loss": 0.4323,
+      "step": 16001
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1389681921734868e-05,
+      "loss": 0.5102,
+      "step": 16002
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1389105170574158e-05,
+      "loss": 0.5465,
+      "step": 16003
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1388528419413445e-05,
+      "loss": 0.4543,
+      "step": 16004
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1387951668252733e-05,
+      "loss": 0.4776,
+      "step": 16005
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1387374917092022e-05,
+      "loss": 0.4651,
+      "step": 16006
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.138679816593131e-05,
+      "loss": 0.4625,
+      "step": 16007
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1386221414770598e-05,
+      "loss": 0.4333,
+      "step": 16008
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1385644663609887e-05,
+      "loss": 0.4697,
+      "step": 16009
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1385067912449175e-05,
+      "loss": 0.4496,
+      "step": 16010
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1384491161288463e-05,
+      "loss": 0.466,
+      "step": 16011
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1383914410127752e-05,
+      "loss": 0.4636,
+      "step": 16012
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.138333765896704e-05,
+      "loss": 0.454,
+      "step": 16013
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1382760907806327e-05,
+      "loss": 0.4309,
+      "step": 16014
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1382184156645617e-05,
+      "loss": 0.5163,
+      "step": 16015
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1381607405484904e-05,
+      "loss": 0.3867,
+      "step": 16016
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1381030654324192e-05,
+      "loss": 0.3988,
+      "step": 16017
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1380453903163481e-05,
+      "loss": 0.6811,
+      "step": 16018
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.137987715200277e-05,
+      "loss": 0.4134,
+      "step": 16019
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1379300400842057e-05,
+      "loss": 0.4253,
+      "step": 16020
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1378723649681346e-05,
+      "loss": 0.4049,
+      "step": 16021
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1378146898520634e-05,
+      "loss": 0.5253,
+      "step": 16022
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1377570147359922e-05,
+      "loss": 0.4781,
+      "step": 16023
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1376993396199213e-05,
+      "loss": 0.4724,
+      "step": 16024
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1376416645038499e-05,
+      "loss": 0.4207,
+      "step": 16025
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1375839893877786e-05,
+      "loss": 0.3881,
+      "step": 16026
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1375263142717074e-05,
+      "loss": 0.4868,
+      "step": 16027
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1374686391556365e-05,
+      "loss": 0.525,
+      "step": 16028
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1374109640395651e-05,
+      "loss": 0.5386,
+      "step": 16029
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1373532889234939e-05,
+      "loss": 0.3512,
+      "step": 16030
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.137295613807423e-05,
+      "loss": 0.5291,
+      "step": 16031
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1372379386913518e-05,
+      "loss": 0.4347,
+      "step": 16032
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1371802635752805e-05,
+      "loss": 0.5912,
+      "step": 16033
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1371225884592095e-05,
+      "loss": 0.5141,
+      "step": 16034
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1370649133431382e-05,
+      "loss": 0.4546,
+      "step": 16035
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.137007238227067e-05,
+      "loss": 0.47,
+      "step": 16036
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.136949563110996e-05,
+      "loss": 0.5726,
+      "step": 16037
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1368918879949247e-05,
+      "loss": 0.4867,
+      "step": 16038
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1368342128788535e-05,
+      "loss": 0.4738,
+      "step": 16039
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1367765377627824e-05,
+      "loss": 0.5215,
+      "step": 16040
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1367188626467112e-05,
+      "loss": 0.4692,
+      "step": 16041
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.13666118753064e-05,
+      "loss": 0.4777,
+      "step": 16042
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1366035124145689e-05,
+      "loss": 0.4585,
+      "step": 16043
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1365458372984977e-05,
+      "loss": 0.4389,
+      "step": 16044
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1364881621824264e-05,
+      "loss": 0.4425,
+      "step": 16045
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1364304870663554e-05,
+      "loss": 0.4671,
+      "step": 16046
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1363728119502841e-05,
+      "loss": 0.4635,
+      "step": 16047
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1363151368342129e-05,
+      "loss": 0.4838,
+      "step": 16048
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1362574617181418e-05,
+      "loss": 0.4428,
+      "step": 16049
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1361997866020706e-05,
+      "loss": 0.6032,
+      "step": 16050
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1361421114859994e-05,
+      "loss": 0.5898,
+      "step": 16051
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1360844363699283e-05,
+      "loss": 0.5838,
+      "step": 16052
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.136026761253857e-05,
+      "loss": 0.3698,
+      "step": 16053
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1359690861377858e-05,
+      "loss": 0.5989,
+      "step": 16054
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.135911411021715e-05,
+      "loss": 0.4454,
+      "step": 16055
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1358537359056436e-05,
+      "loss": 0.5427,
+      "step": 16056
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1357960607895723e-05,
+      "loss": 0.5552,
+      "step": 16057
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1357383856735014e-05,
+      "loss": 0.4741,
+      "step": 16058
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1356807105574302e-05,
+      "loss": 0.4376,
+      "step": 16059
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1356230354413588e-05,
+      "loss": 0.3708,
+      "step": 16060
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1355653603252879e-05,
+      "loss": 0.5534,
+      "step": 16061
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1355076852092167e-05,
+      "loss": 0.5138,
+      "step": 16062
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1354500100931454e-05,
+      "loss": 0.566,
+      "step": 16063
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1353923349770742e-05,
+      "loss": 0.5209,
+      "step": 16064
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1353346598610031e-05,
+      "loss": 0.4408,
+      "step": 16065
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.135276984744932e-05,
+      "loss": 0.42,
+      "step": 16066
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1352193096288607e-05,
+      "loss": 0.4197,
+      "step": 16067
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1351616345127896e-05,
+      "loss": 0.4078,
+      "step": 16068
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1351039593967184e-05,
+      "loss": 0.4945,
+      "step": 16069
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1350462842806472e-05,
+      "loss": 0.4649,
+      "step": 16070
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1349886091645761e-05,
+      "loss": 0.4811,
+      "step": 16071
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1349309340485049e-05,
+      "loss": 0.568,
+      "step": 16072
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1348732589324336e-05,
+      "loss": 0.4749,
+      "step": 16073
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1348155838163626e-05,
+      "loss": 0.4263,
+      "step": 16074
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1347579087002913e-05,
+      "loss": 0.5689,
+      "step": 16075
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1347002335842201e-05,
+      "loss": 0.5285,
+      "step": 16076
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.134642558468149e-05,
+      "loss": 0.5322,
+      "step": 16077
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1345848833520778e-05,
+      "loss": 0.5189,
+      "step": 16078
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1345272082360066e-05,
+      "loss": 0.5615,
+      "step": 16079
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1344695331199355e-05,
+      "loss": 0.3943,
+      "step": 16080
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1344118580038643e-05,
+      "loss": 0.4507,
+      "step": 16081
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.134354182887793e-05,
+      "loss": 0.4403,
+      "step": 16082
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.134296507771722e-05,
+      "loss": 0.4698,
+      "step": 16083
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1342388326556508e-05,
+      "loss": 0.534,
+      "step": 16084
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1341811575395795e-05,
+      "loss": 0.5381,
+      "step": 16085
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1341234824235086e-05,
+      "loss": 0.5157,
+      "step": 16086
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1340658073074372e-05,
+      "loss": 0.5667,
+      "step": 16087
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.134008132191366e-05,
+      "loss": 0.5375,
+      "step": 16088
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1339504570752951e-05,
+      "loss": 0.4929,
+      "step": 16089
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1338927819592239e-05,
+      "loss": 0.5247,
+      "step": 16090
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1338351068431525e-05,
+      "loss": 0.5047,
+      "step": 16091
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1337774317270816e-05,
+      "loss": 0.4394,
+      "step": 16092
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1337197566110104e-05,
+      "loss": 0.4485,
+      "step": 16093
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1336620814949391e-05,
+      "loss": 0.4731,
+      "step": 16094
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.133604406378868e-05,
+      "loss": 0.4807,
+      "step": 16095
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1335467312627968e-05,
+      "loss": 0.5771,
+      "step": 16096
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1334890561467256e-05,
+      "loss": 0.4444,
+      "step": 16097
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1334313810306544e-05,
+      "loss": 0.4733,
+      "step": 16098
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1333737059145833e-05,
+      "loss": 0.4329,
+      "step": 16099
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.133316030798512e-05,
+      "loss": 0.4663,
+      "step": 16100
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1332583556824408e-05,
+      "loss": 0.5154,
+      "step": 16101
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1332006805663698e-05,
+      "loss": 0.5613,
+      "step": 16102
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1331430054502986e-05,
+      "loss": 0.4564,
+      "step": 16103
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1330853303342273e-05,
+      "loss": 0.4884,
+      "step": 16104
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1330276552181563e-05,
+      "loss": 0.6301,
+      "step": 16105
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.132969980102085e-05,
+      "loss": 0.422,
+      "step": 16106
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1329123049860138e-05,
+      "loss": 0.4924,
+      "step": 16107
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1328546298699427e-05,
+      "loss": 0.5137,
+      "step": 16108
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1327969547538715e-05,
+      "loss": 0.4494,
+      "step": 16109
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1327392796378003e-05,
+      "loss": 0.3963,
+      "step": 16110
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1326816045217292e-05,
+      "loss": 0.4339,
+      "step": 16111
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.132623929405658e-05,
+      "loss": 0.4563,
+      "step": 16112
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1325662542895867e-05,
+      "loss": 0.6048,
+      "step": 16113
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1325085791735157e-05,
+      "loss": 0.523,
+      "step": 16114
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1324509040574445e-05,
+      "loss": 0.3583,
+      "step": 16115
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1323932289413732e-05,
+      "loss": 0.5209,
+      "step": 16116
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1323355538253023e-05,
+      "loss": 0.4485,
+      "step": 16117
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.132277878709231e-05,
+      "loss": 0.4726,
+      "step": 16118
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1322202035931597e-05,
+      "loss": 0.5123,
+      "step": 16119
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1321625284770888e-05,
+      "loss": 0.5204,
+      "step": 16120
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1321048533610176e-05,
+      "loss": 0.4211,
+      "step": 16121
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1320471782449462e-05,
+      "loss": 0.3763,
+      "step": 16122
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1319895031288753e-05,
+      "loss": 0.4902,
+      "step": 16123
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.131931828012804e-05,
+      "loss": 0.4415,
+      "step": 16124
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1318741528967328e-05,
+      "loss": 0.5554,
+      "step": 16125
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1318164777806618e-05,
+      "loss": 0.5137,
+      "step": 16126
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1317588026645905e-05,
+      "loss": 0.4995,
+      "step": 16127
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1317011275485193e-05,
+      "loss": 0.6042,
+      "step": 16128
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1316434524324482e-05,
+      "loss": 0.4446,
+      "step": 16129
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.131585777316377e-05,
+      "loss": 0.3989,
+      "step": 16130
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1315281022003058e-05,
+      "loss": 0.5898,
+      "step": 16131
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1314704270842345e-05,
+      "loss": 0.4619,
+      "step": 16132
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1314127519681635e-05,
+      "loss": 0.4347,
+      "step": 16133
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1313550768520922e-05,
+      "loss": 0.5986,
+      "step": 16134
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.131297401736021e-05,
+      "loss": 0.3645,
+      "step": 16135
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.13123972661995e-05,
+      "loss": 0.4693,
+      "step": 16136
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1311820515038787e-05,
+      "loss": 0.5903,
+      "step": 16137
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1311243763878075e-05,
+      "loss": 0.4591,
+      "step": 16138
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1310667012717364e-05,
+      "loss": 0.5651,
+      "step": 16139
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1310090261556652e-05,
+      "loss": 0.5475,
+      "step": 16140
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.130951351039594e-05,
+      "loss": 0.4383,
+      "step": 16141
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1308936759235229e-05,
+      "loss": 0.6277,
+      "step": 16142
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1308360008074517e-05,
+      "loss": 0.618,
+      "step": 16143
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1307783256913804e-05,
+      "loss": 0.4581,
+      "step": 16144
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1307206505753094e-05,
+      "loss": 0.4883,
+      "step": 16145
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1306629754592381e-05,
+      "loss": 0.4762,
+      "step": 16146
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1306053003431669e-05,
+      "loss": 0.3824,
+      "step": 16147
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.130547625227096e-05,
+      "loss": 0.47,
+      "step": 16148
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1304899501110246e-05,
+      "loss": 0.3849,
+      "step": 16149
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1304322749949534e-05,
+      "loss": 0.4236,
+      "step": 16150
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1303745998788825e-05,
+      "loss": 0.5845,
+      "step": 16151
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1303169247628113e-05,
+      "loss": 0.325,
+      "step": 16152
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1302592496467399e-05,
+      "loss": 0.4157,
+      "step": 16153
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.130201574530669e-05,
+      "loss": 0.3846,
+      "step": 16154
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1301438994145977e-05,
+      "loss": 0.4862,
+      "step": 16155
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1300862242985265e-05,
+      "loss": 0.4653,
+      "step": 16156
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1300285491824554e-05,
+      "loss": 0.4199,
+      "step": 16157
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1299708740663842e-05,
+      "loss": 0.553,
+      "step": 16158
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.129913198950313e-05,
+      "loss": 0.4085,
+      "step": 16159
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.129855523834242e-05,
+      "loss": 0.477,
+      "step": 16160
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1297978487181707e-05,
+      "loss": 0.5791,
+      "step": 16161
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1297401736020995e-05,
+      "loss": 0.4606,
+      "step": 16162
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1296824984860284e-05,
+      "loss": 0.4149,
+      "step": 16163
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1296248233699572e-05,
+      "loss": 0.4612,
+      "step": 16164
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.129567148253886e-05,
+      "loss": 0.6832,
+      "step": 16165
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1295094731378147e-05,
+      "loss": 0.4414,
+      "step": 16166
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1294517980217436e-05,
+      "loss": 0.4149,
+      "step": 16167
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1293941229056724e-05,
+      "loss": 0.4996,
+      "step": 16168
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1293364477896012e-05,
+      "loss": 0.4466,
+      "step": 16169
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1292787726735301e-05,
+      "loss": 0.4816,
+      "step": 16170
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1292210975574589e-05,
+      "loss": 0.5631,
+      "step": 16171
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1291634224413876e-05,
+      "loss": 0.4767,
+      "step": 16172
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1291057473253166e-05,
+      "loss": 0.4189,
+      "step": 16173
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1290480722092454e-05,
+      "loss": 0.4946,
+      "step": 16174
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.1289903970931741e-05,
+      "loss": 0.5443,
+      "step": 16175
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.128932721977103e-05,
+      "loss": 0.4866,
+      "step": 16176
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1288750468610318e-05,
+      "loss": 0.3567,
+      "step": 16177
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1288173717449606e-05,
+      "loss": 0.4463,
+      "step": 16178
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1287596966288897e-05,
+      "loss": 0.5678,
+      "step": 16179
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1287020215128183e-05,
+      "loss": 0.4261,
+      "step": 16180
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.128644346396747e-05,
+      "loss": 0.4802,
+      "step": 16181
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1285866712806762e-05,
+      "loss": 0.4662,
+      "step": 16182
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.128528996164605e-05,
+      "loss": 0.4389,
+      "step": 16183
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1284713210485335e-05,
+      "loss": 0.5798,
+      "step": 16184
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1284136459324627e-05,
+      "loss": 0.5053,
+      "step": 16185
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1283559708163914e-05,
+      "loss": 0.5707,
+      "step": 16186
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1282982957003202e-05,
+      "loss": 0.5025,
+      "step": 16187
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1282406205842491e-05,
+      "loss": 0.3253,
+      "step": 16188
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1281829454681779e-05,
+      "loss": 0.3422,
+      "step": 16189
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1281252703521067e-05,
+      "loss": 0.4845,
+      "step": 16190
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1280675952360356e-05,
+      "loss": 0.463,
+      "step": 16191
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1280099201199644e-05,
+      "loss": 0.5701,
+      "step": 16192
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1279522450038931e-05,
+      "loss": 0.4681,
+      "step": 16193
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.127894569887822e-05,
+      "loss": 0.3989,
+      "step": 16194
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1278368947717508e-05,
+      "loss": 0.4276,
+      "step": 16195
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1277792196556796e-05,
+      "loss": 0.5232,
+      "step": 16196
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1277215445396086e-05,
+      "loss": 0.3964,
+      "step": 16197
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1276638694235373e-05,
+      "loss": 0.4949,
+      "step": 16198
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1276061943074661e-05,
+      "loss": 0.3637,
+      "step": 16199
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.127548519191395e-05,
+      "loss": 0.4975,
+      "step": 16200
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1274908440753238e-05,
+      "loss": 0.4918,
+      "step": 16201
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1274331689592526e-05,
+      "loss": 0.5362,
+      "step": 16202
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1273754938431813e-05,
+      "loss": 0.4608,
+      "step": 16203
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1273178187271103e-05,
+      "loss": 0.4145,
+      "step": 16204
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.127260143611039e-05,
+      "loss": 0.4152,
+      "step": 16205
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1272024684949678e-05,
+      "loss": 0.4536,
+      "step": 16206
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1271447933788967e-05,
+      "loss": 0.51,
+      "step": 16207
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1270871182628255e-05,
+      "loss": 0.5247,
+      "step": 16208
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1270294431467543e-05,
+      "loss": 0.427,
+      "step": 16209
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1269717680306834e-05,
+      "loss": 0.5623,
+      "step": 16210
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.126914092914612e-05,
+      "loss": 0.4215,
+      "step": 16211
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1268564177985408e-05,
+      "loss": 0.4295,
+      "step": 16212
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1267987426824699e-05,
+      "loss": 0.5896,
+      "step": 16213
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1267410675663986e-05,
+      "loss": 0.5606,
+      "step": 16214
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1266833924503272e-05,
+      "loss": 0.4821,
+      "step": 16215
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1266257173342563e-05,
+      "loss": 0.418,
+      "step": 16216
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1265680422181851e-05,
+      "loss": 0.5453,
+      "step": 16217
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1265103671021139e-05,
+      "loss": 0.3846,
+      "step": 16218
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1264526919860428e-05,
+      "loss": 0.5747,
+      "step": 16219
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1263950168699716e-05,
+      "loss": 0.3235,
+      "step": 16220
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1263373417539004e-05,
+      "loss": 0.4944,
+      "step": 16221
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1262796666378293e-05,
+      "loss": 0.4093,
+      "step": 16222
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.126221991521758e-05,
+      "loss": 0.5751,
+      "step": 16223
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1261643164056868e-05,
+      "loss": 0.4773,
+      "step": 16224
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1261066412896158e-05,
+      "loss": 0.5814,
+      "step": 16225
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1260489661735445e-05,
+      "loss": 0.4166,
+      "step": 16226
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1259912910574733e-05,
+      "loss": 0.5345,
+      "step": 16227
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1259336159414022e-05,
+      "loss": 0.4766,
+      "step": 16228
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.125875940825331e-05,
+      "loss": 0.4935,
+      "step": 16229
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1258182657092598e-05,
+      "loss": 0.4699,
+      "step": 16230
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1257605905931887e-05,
+      "loss": 0.4366,
+      "step": 16231
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1257029154771175e-05,
+      "loss": 0.4064,
+      "step": 16232
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1256452403610463e-05,
+      "loss": 0.5403,
+      "step": 16233
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1255875652449752e-05,
+      "loss": 0.5721,
+      "step": 16234
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.125529890128904e-05,
+      "loss": 0.6068,
+      "step": 16235
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1254722150128327e-05,
+      "loss": 0.452,
+      "step": 16236
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1254145398967615e-05,
+      "loss": 0.5336,
+      "step": 16237
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1253568647806904e-05,
+      "loss": 0.476,
+      "step": 16238
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1252991896646192e-05,
+      "loss": 0.5614,
+      "step": 16239
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.125241514548548e-05,
+      "loss": 0.5209,
+      "step": 16240
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.125183839432477e-05,
+      "loss": 0.4472,
+      "step": 16241
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1251261643164057e-05,
+      "loss": 0.5422,
+      "step": 16242
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1250684892003344e-05,
+      "loss": 0.4808,
+      "step": 16243
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1250108140842636e-05,
+      "loss": 0.4263,
+      "step": 16244
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1249531389681923e-05,
+      "loss": 0.3738,
+      "step": 16245
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.124895463852121e-05,
+      "loss": 0.4971,
+      "step": 16246
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.12483778873605e-05,
+      "loss": 0.5522,
+      "step": 16247
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1247801136199788e-05,
+      "loss": 0.5245,
+      "step": 16248
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1247224385039076e-05,
+      "loss": 0.4665,
+      "step": 16249
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1246647633878365e-05,
+      "loss": 0.5411,
+      "step": 16250
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1246070882717653e-05,
+      "loss": 0.4247,
+      "step": 16251
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.124549413155694e-05,
+      "loss": 0.4887,
+      "step": 16252
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.124491738039623e-05,
+      "loss": 0.5417,
+      "step": 16253
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1244340629235517e-05,
+      "loss": 0.4276,
+      "step": 16254
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1243763878074805e-05,
+      "loss": 0.2641,
+      "step": 16255
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1243187126914095e-05,
+      "loss": 0.4266,
+      "step": 16256
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1242610375753382e-05,
+      "loss": 0.4435,
+      "step": 16257
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.124203362459267e-05,
+      "loss": 0.5158,
+      "step": 16258
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.124145687343196e-05,
+      "loss": 0.5173,
+      "step": 16259
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1240880122271247e-05,
+      "loss": 0.5513,
+      "step": 16260
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1240303371110535e-05,
+      "loss": 0.4563,
+      "step": 16261
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1239726619949824e-05,
+      "loss": 0.4715,
+      "step": 16262
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1239149868789112e-05,
+      "loss": 0.6216,
+      "step": 16263
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.12385731176284e-05,
+      "loss": 0.486,
+      "step": 16264
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1237996366467689e-05,
+      "loss": 0.5374,
+      "step": 16265
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1237419615306976e-05,
+      "loss": 0.5016,
+      "step": 16266
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1236842864146264e-05,
+      "loss": 0.4429,
+      "step": 16267
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1236266112985554e-05,
+      "loss": 0.507,
+      "step": 16268
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1235689361824841e-05,
+      "loss": 0.4416,
+      "step": 16269
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1235112610664129e-05,
+      "loss": 0.4578,
+      "step": 16270
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1234535859503417e-05,
+      "loss": 0.3889,
+      "step": 16271
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1233959108342708e-05,
+      "loss": 0.4445,
+      "step": 16272
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1233382357181994e-05,
+      "loss": 0.4128,
+      "step": 16273
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1232805606021281e-05,
+      "loss": 0.5,
+      "step": 16274
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1232228854860572e-05,
+      "loss": 0.4465,
+      "step": 16275
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.123165210369986e-05,
+      "loss": 0.48,
+      "step": 16276
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1231075352539148e-05,
+      "loss": 0.5174,
+      "step": 16277
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1230498601378437e-05,
+      "loss": 0.4427,
+      "step": 16278
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1229921850217725e-05,
+      "loss": 0.5554,
+      "step": 16279
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1229345099057013e-05,
+      "loss": 0.4814,
+      "step": 16280
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1228768347896302e-05,
+      "loss": 0.5329,
+      "step": 16281
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.122819159673559e-05,
+      "loss": 0.4169,
+      "step": 16282
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1227614845574877e-05,
+      "loss": 0.5716,
+      "step": 16283
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1227038094414167e-05,
+      "loss": 0.6176,
+      "step": 16284
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1226461343253454e-05,
+      "loss": 0.4498,
+      "step": 16285
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1225884592092742e-05,
+      "loss": 0.4896,
+      "step": 16286
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1225307840932031e-05,
+      "loss": 0.5312,
+      "step": 16287
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1224731089771319e-05,
+      "loss": 0.4485,
+      "step": 16288
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1224154338610607e-05,
+      "loss": 0.3921,
+      "step": 16289
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1223577587449896e-05,
+      "loss": 0.4109,
+      "step": 16290
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1223000836289184e-05,
+      "loss": 0.4029,
+      "step": 16291
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1222424085128472e-05,
+      "loss": 0.5463,
+      "step": 16292
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1221847333967761e-05,
+      "loss": 0.5104,
+      "step": 16293
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1221270582807049e-05,
+      "loss": 0.6675,
+      "step": 16294
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1220693831646336e-05,
+      "loss": 0.4628,
+      "step": 16295
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1220117080485626e-05,
+      "loss": 0.5789,
+      "step": 16296
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1219540329324913e-05,
+      "loss": 0.4038,
+      "step": 16297
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1218963578164201e-05,
+      "loss": 0.5248,
+      "step": 16298
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.121838682700349e-05,
+      "loss": 0.3918,
+      "step": 16299
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1217810075842778e-05,
+      "loss": 0.4437,
+      "step": 16300
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1217233324682066e-05,
+      "loss": 0.5991,
+      "step": 16301
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1216656573521357e-05,
+      "loss": 0.5235,
+      "step": 16302
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1216079822360645e-05,
+      "loss": 0.5303,
+      "step": 16303
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.121550307119993e-05,
+      "loss": 0.5684,
+      "step": 16304
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1214926320039222e-05,
+      "loss": 0.4561,
+      "step": 16305
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.121434956887851e-05,
+      "loss": 0.5028,
+      "step": 16306
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1213772817717797e-05,
+      "loss": 0.4642,
+      "step": 16307
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1213196066557085e-05,
+      "loss": 0.3864,
+      "step": 16308
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1212619315396374e-05,
+      "loss": 0.3691,
+      "step": 16309
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1212042564235662e-05,
+      "loss": 0.3956,
+      "step": 16310
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.121146581307495e-05,
+      "loss": 0.5138,
+      "step": 16311
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1210889061914239e-05,
+      "loss": 0.4482,
+      "step": 16312
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1210312310753526e-05,
+      "loss": 0.4705,
+      "step": 16313
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1209735559592814e-05,
+      "loss": 0.4689,
+      "step": 16314
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1209158808432104e-05,
+      "loss": 0.5021,
+      "step": 16315
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1208582057271391e-05,
+      "loss": 0.4234,
+      "step": 16316
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1208005306110679e-05,
+      "loss": 0.5793,
+      "step": 16317
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1207428554949968e-05,
+      "loss": 0.4224,
+      "step": 16318
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1206851803789256e-05,
+      "loss": 0.4546,
+      "step": 16319
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1206275052628544e-05,
+      "loss": 0.3932,
+      "step": 16320
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1205698301467833e-05,
+      "loss": 0.4409,
+      "step": 16321
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.120512155030712e-05,
+      "loss": 0.4156,
+      "step": 16322
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1204544799146408e-05,
+      "loss": 0.4542,
+      "step": 16323
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1203968047985698e-05,
+      "loss": 0.5179,
+      "step": 16324
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1203391296824985e-05,
+      "loss": 0.4756,
+      "step": 16325
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1202814545664273e-05,
+      "loss": 0.3485,
+      "step": 16326
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1202237794503563e-05,
+      "loss": 0.5457,
+      "step": 16327
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.120166104334285e-05,
+      "loss": 0.4735,
+      "step": 16328
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1201084292182138e-05,
+      "loss": 0.4191,
+      "step": 16329
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1200507541021427e-05,
+      "loss": 0.5106,
+      "step": 16330
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1199930789860715e-05,
+      "loss": 0.4691,
+      "step": 16331
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1199354038700003e-05,
+      "loss": 0.5415,
+      "step": 16332
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1198777287539294e-05,
+      "loss": 0.4045,
+      "step": 16333
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1198200536378581e-05,
+      "loss": 0.487,
+      "step": 16334
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1197623785217867e-05,
+      "loss": 0.4059,
+      "step": 16335
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1197047034057158e-05,
+      "loss": 0.4195,
+      "step": 16336
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1196470282896446e-05,
+      "loss": 0.474,
+      "step": 16337
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1195893531735734e-05,
+      "loss": 0.4712,
+      "step": 16338
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1195316780575023e-05,
+      "loss": 0.3596,
+      "step": 16339
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1194740029414311e-05,
+      "loss": 0.609,
+      "step": 16340
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1194163278253599e-05,
+      "loss": 0.4416,
+      "step": 16341
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1193586527092886e-05,
+      "loss": 0.4682,
+      "step": 16342
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1193009775932176e-05,
+      "loss": 0.5684,
+      "step": 16343
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1192433024771463e-05,
+      "loss": 0.565,
+      "step": 16344
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1191856273610751e-05,
+      "loss": 0.4981,
+      "step": 16345
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.119127952245004e-05,
+      "loss": 0.4561,
+      "step": 16346
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1190702771289328e-05,
+      "loss": 0.5429,
+      "step": 16347
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1190126020128616e-05,
+      "loss": 0.5893,
+      "step": 16348
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1189549268967905e-05,
+      "loss": 0.5853,
+      "step": 16349
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1188972517807193e-05,
+      "loss": 0.5548,
+      "step": 16350
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.118839576664648e-05,
+      "loss": 0.4018,
+      "step": 16351
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.118781901548577e-05,
+      "loss": 0.4613,
+      "step": 16352
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1187242264325058e-05,
+      "loss": 0.5589,
+      "step": 16353
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1186665513164345e-05,
+      "loss": 0.6456,
+      "step": 16354
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 1.1186088762003635e-05,
+      "loss": 0.4028,
+      "step": 16355
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1185512010842922e-05,
+      "loss": 0.4681,
+      "step": 16356
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.118493525968221e-05,
+      "loss": 0.4188,
+      "step": 16357
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.11843585085215e-05,
+      "loss": 0.4436,
+      "step": 16358
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1183781757360787e-05,
+      "loss": 0.3544,
+      "step": 16359
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1183205006200075e-05,
+      "loss": 0.5444,
+      "step": 16360
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1182628255039366e-05,
+      "loss": 0.4561,
+      "step": 16361
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1182051503878652e-05,
+      "loss": 0.4466,
+      "step": 16362
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.118147475271794e-05,
+      "loss": 0.4885,
+      "step": 16363
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.118089800155723e-05,
+      "loss": 0.5844,
+      "step": 16364
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1180321250396518e-05,
+      "loss": 0.4436,
+      "step": 16365
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1179744499235804e-05,
+      "loss": 0.563,
+      "step": 16366
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1179167748075095e-05,
+      "loss": 0.5109,
+      "step": 16367
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1178590996914383e-05,
+      "loss": 0.407,
+      "step": 16368
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.117801424575367e-05,
+      "loss": 0.3903,
+      "step": 16369
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.117743749459296e-05,
+      "loss": 0.4396,
+      "step": 16370
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1176860743432248e-05,
+      "loss": 0.4833,
+      "step": 16371
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1176283992271535e-05,
+      "loss": 0.5332,
+      "step": 16372
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1175707241110825e-05,
+      "loss": 0.6595,
+      "step": 16373
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1175130489950113e-05,
+      "loss": 0.4393,
+      "step": 16374
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.11745537387894e-05,
+      "loss": 0.4979,
+      "step": 16375
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1173976987628688e-05,
+      "loss": 0.2962,
+      "step": 16376
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1173400236467977e-05,
+      "loss": 0.4655,
+      "step": 16377
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1172823485307265e-05,
+      "loss": 0.5397,
+      "step": 16378
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1172246734146553e-05,
+      "loss": 0.6428,
+      "step": 16379
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1171669982985842e-05,
+      "loss": 0.4452,
+      "step": 16380
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.117109323182513e-05,
+      "loss": 0.4459,
+      "step": 16381
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1170516480664417e-05,
+      "loss": 0.4484,
+      "step": 16382
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1169939729503707e-05,
+      "loss": 0.4847,
+      "step": 16383
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1169362978342994e-05,
+      "loss": 0.4432,
+      "step": 16384
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1168786227182282e-05,
+      "loss": 0.7437,
+      "step": 16385
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1168209476021572e-05,
+      "loss": 0.449,
+      "step": 16386
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.116763272486086e-05,
+      "loss": 0.4216,
+      "step": 16387
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1167055973700147e-05,
+      "loss": 0.4937,
+      "step": 16388
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1166479222539436e-05,
+      "loss": 0.463,
+      "step": 16389
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1165902471378724e-05,
+      "loss": 0.5868,
+      "step": 16390
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1165325720218012e-05,
+      "loss": 0.5921,
+      "step": 16391
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1164748969057303e-05,
+      "loss": 0.4187,
+      "step": 16392
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1164172217896589e-05,
+      "loss": 0.4775,
+      "step": 16393
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1163595466735876e-05,
+      "loss": 0.6383,
+      "step": 16394
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1163018715575167e-05,
+      "loss": 0.4224,
+      "step": 16395
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1162441964414455e-05,
+      "loss": 0.3424,
+      "step": 16396
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1161865213253741e-05,
+      "loss": 0.475,
+      "step": 16397
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1161288462093032e-05,
+      "loss": 0.4458,
+      "step": 16398
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.116071171093232e-05,
+      "loss": 0.6193,
+      "step": 16399
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1160134959771608e-05,
+      "loss": 0.5257,
+      "step": 16400
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1159558208610897e-05,
+      "loss": 0.4327,
+      "step": 16401
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1158981457450185e-05,
+      "loss": 0.3766,
+      "step": 16402
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1158404706289472e-05,
+      "loss": 0.5378,
+      "step": 16403
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1157827955128762e-05,
+      "loss": 0.5124,
+      "step": 16404
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.115725120396805e-05,
+      "loss": 0.5656,
+      "step": 16405
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1156674452807337e-05,
+      "loss": 0.3811,
+      "step": 16406
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1156097701646626e-05,
+      "loss": 0.4109,
+      "step": 16407
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1155520950485914e-05,
+      "loss": 0.5534,
+      "step": 16408
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1154944199325202e-05,
+      "loss": 0.4609,
+      "step": 16409
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.115436744816449e-05,
+      "loss": 0.4569,
+      "step": 16410
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1153790697003779e-05,
+      "loss": 0.4084,
+      "step": 16411
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1153213945843067e-05,
+      "loss": 0.4632,
+      "step": 16412
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1152637194682354e-05,
+      "loss": 0.4919,
+      "step": 16413
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1152060443521644e-05,
+      "loss": 0.6007,
+      "step": 16414
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1151483692360931e-05,
+      "loss": 0.5063,
+      "step": 16415
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1150906941200219e-05,
+      "loss": 0.4615,
+      "step": 16416
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1150330190039508e-05,
+      "loss": 0.508,
+      "step": 16417
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1149753438878796e-05,
+      "loss": 0.4811,
+      "step": 16418
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1149176687718084e-05,
+      "loss": 0.4369,
+      "step": 16419
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1148599936557373e-05,
+      "loss": 0.4848,
+      "step": 16420
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.114802318539666e-05,
+      "loss": 0.5124,
+      "step": 16421
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1147446434235949e-05,
+      "loss": 0.3904,
+      "step": 16422
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.114686968307524e-05,
+      "loss": 0.589,
+      "step": 16423
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1146292931914526e-05,
+      "loss": 0.4856,
+      "step": 16424
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1145716180753813e-05,
+      "loss": 0.4522,
+      "step": 16425
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1145139429593104e-05,
+      "loss": 0.5981,
+      "step": 16426
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1144562678432392e-05,
+      "loss": 0.4915,
+      "step": 16427
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1143985927271678e-05,
+      "loss": 0.5295,
+      "step": 16428
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1143409176110969e-05,
+      "loss": 0.6531,
+      "step": 16429
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1142832424950257e-05,
+      "loss": 0.4401,
+      "step": 16430
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1142255673789544e-05,
+      "loss": 0.4161,
+      "step": 16431
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1141678922628834e-05,
+      "loss": 0.4319,
+      "step": 16432
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1141102171468122e-05,
+      "loss": 0.4815,
+      "step": 16433
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.114052542030741e-05,
+      "loss": 0.3708,
+      "step": 16434
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1139948669146699e-05,
+      "loss": 0.4783,
+      "step": 16435
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1139371917985986e-05,
+      "loss": 0.5373,
+      "step": 16436
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1138795166825274e-05,
+      "loss": 0.4694,
+      "step": 16437
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1138218415664563e-05,
+      "loss": 0.4375,
+      "step": 16438
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1137641664503851e-05,
+      "loss": 0.4609,
+      "step": 16439
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1137064913343139e-05,
+      "loss": 0.4798,
+      "step": 16440
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1136488162182428e-05,
+      "loss": 0.4466,
+      "step": 16441
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1135911411021716e-05,
+      "loss": 0.5018,
+      "step": 16442
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1135334659861003e-05,
+      "loss": 0.5426,
+      "step": 16443
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1134757908700293e-05,
+      "loss": 0.4358,
+      "step": 16444
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.113418115753958e-05,
+      "loss": 0.5029,
+      "step": 16445
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1133604406378868e-05,
+      "loss": 0.4918,
+      "step": 16446
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1133027655218156e-05,
+      "loss": 0.4243,
+      "step": 16447
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1132450904057445e-05,
+      "loss": 0.4703,
+      "step": 16448
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1131874152896733e-05,
+      "loss": 0.4457,
+      "step": 16449
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.113129740173602e-05,
+      "loss": 0.5368,
+      "step": 16450
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.113072065057531e-05,
+      "loss": 0.5179,
+      "step": 16451
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1130143899414598e-05,
+      "loss": 0.519,
+      "step": 16452
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1129567148253885e-05,
+      "loss": 0.5317,
+      "step": 16453
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1128990397093176e-05,
+      "loss": 0.5903,
+      "step": 16454
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1128413645932462e-05,
+      "loss": 0.5266,
+      "step": 16455
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.112783689477175e-05,
+      "loss": 0.4117,
+      "step": 16456
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1127260143611041e-05,
+      "loss": 0.5518,
+      "step": 16457
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1126683392450329e-05,
+      "loss": 0.5465,
+      "step": 16458
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1126106641289615e-05,
+      "loss": 0.4813,
+      "step": 16459
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1125529890128906e-05,
+      "loss": 0.5018,
+      "step": 16460
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1124953138968194e-05,
+      "loss": 0.3841,
+      "step": 16461
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1124376387807481e-05,
+      "loss": 0.5002,
+      "step": 16462
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.112379963664677e-05,
+      "loss": 0.4492,
+      "step": 16463
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1123222885486058e-05,
+      "loss": 0.4749,
+      "step": 16464
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1122646134325346e-05,
+      "loss": 0.3347,
+      "step": 16465
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1122069383164635e-05,
+      "loss": 0.4546,
+      "step": 16466
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1121492632003923e-05,
+      "loss": 0.4243,
+      "step": 16467
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.112091588084321e-05,
+      "loss": 0.5432,
+      "step": 16468
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.11203391296825e-05,
+      "loss": 0.4459,
+      "step": 16469
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1119762378521788e-05,
+      "loss": 0.5312,
+      "step": 16470
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1119185627361076e-05,
+      "loss": 0.5554,
+      "step": 16471
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1118608876200365e-05,
+      "loss": 0.5152,
+      "step": 16472
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1118032125039653e-05,
+      "loss": 0.4216,
+      "step": 16473
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.111745537387894e-05,
+      "loss": 0.4285,
+      "step": 16474
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.111687862271823e-05,
+      "loss": 0.6491,
+      "step": 16475
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1116301871557517e-05,
+      "loss": 0.6037,
+      "step": 16476
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1115725120396805e-05,
+      "loss": 0.556,
+      "step": 16477
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1115148369236094e-05,
+      "loss": 0.4812,
+      "step": 16478
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1114571618075382e-05,
+      "loss": 0.4345,
+      "step": 16479
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.111399486691467e-05,
+      "loss": 0.5059,
+      "step": 16480
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1113418115753958e-05,
+      "loss": 0.5662,
+      "step": 16481
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1112841364593247e-05,
+      "loss": 0.4433,
+      "step": 16482
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1112264613432535e-05,
+      "loss": 0.5158,
+      "step": 16483
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1111687862271822e-05,
+      "loss": 0.4387,
+      "step": 16484
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1111111111111113e-05,
+      "loss": 0.4133,
+      "step": 16485
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.11105343599504e-05,
+      "loss": 0.448,
+      "step": 16486
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1109957608789687e-05,
+      "loss": 0.3814,
+      "step": 16487
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1109380857628978e-05,
+      "loss": 0.4603,
+      "step": 16488
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1108804106468266e-05,
+      "loss": 0.4986,
+      "step": 16489
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1108227355307552e-05,
+      "loss": 0.6094,
+      "step": 16490
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1107650604146843e-05,
+      "loss": 0.4587,
+      "step": 16491
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.110707385298613e-05,
+      "loss": 0.3965,
+      "step": 16492
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1106497101825418e-05,
+      "loss": 0.4657,
+      "step": 16493
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1105920350664708e-05,
+      "loss": 0.4336,
+      "step": 16494
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1105343599503995e-05,
+      "loss": 0.6639,
+      "step": 16495
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1104766848343283e-05,
+      "loss": 0.3499,
+      "step": 16496
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1104190097182572e-05,
+      "loss": 0.4775,
+      "step": 16497
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.110361334602186e-05,
+      "loss": 0.5161,
+      "step": 16498
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1103036594861148e-05,
+      "loss": 0.4619,
+      "step": 16499
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1102459843700437e-05,
+      "loss": 0.3638,
+      "step": 16500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1101883092539725e-05,
+      "loss": 0.4206,
+      "step": 16501
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1101306341379012e-05,
+      "loss": 0.579,
+      "step": 16502
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1100729590218302e-05,
+      "loss": 0.4934,
+      "step": 16503
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.110015283905759e-05,
+      "loss": 0.473,
+      "step": 16504
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1099576087896877e-05,
+      "loss": 0.4914,
+      "step": 16505
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1098999336736167e-05,
+      "loss": 0.543,
+      "step": 16506
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1098422585575454e-05,
+      "loss": 0.6758,
+      "step": 16507
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1097845834414742e-05,
+      "loss": 0.4924,
+      "step": 16508
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1097269083254031e-05,
+      "loss": 0.5531,
+      "step": 16509
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1096692332093319e-05,
+      "loss": 0.5017,
+      "step": 16510
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1096115580932607e-05,
+      "loss": 0.5456,
+      "step": 16511
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1095538829771896e-05,
+      "loss": 0.4339,
+      "step": 16512
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1094962078611184e-05,
+      "loss": 0.516,
+      "step": 16513
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1094385327450471e-05,
+      "loss": 0.3289,
+      "step": 16514
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1093808576289759e-05,
+      "loss": 0.4218,
+      "step": 16515
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.109323182512905e-05,
+      "loss": 0.5226,
+      "step": 16516
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1092655073968336e-05,
+      "loss": 0.4907,
+      "step": 16517
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1092078322807624e-05,
+      "loss": 0.4246,
+      "step": 16518
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1091501571646915e-05,
+      "loss": 0.4664,
+      "step": 16519
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1090924820486203e-05,
+      "loss": 0.4222,
+      "step": 16520
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1090348069325489e-05,
+      "loss": 0.601,
+      "step": 16521
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.108977131816478e-05,
+      "loss": 0.4199,
+      "step": 16522
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1089194567004067e-05,
+      "loss": 0.404,
+      "step": 16523
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1088617815843355e-05,
+      "loss": 0.5236,
+      "step": 16524
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1088041064682644e-05,
+      "loss": 0.429,
+      "step": 16525
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1087464313521932e-05,
+      "loss": 0.6656,
+      "step": 16526
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.108688756236122e-05,
+      "loss": 0.4331,
+      "step": 16527
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.108631081120051e-05,
+      "loss": 0.4607,
+      "step": 16528
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1085734060039797e-05,
+      "loss": 0.531,
+      "step": 16529
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1085157308879085e-05,
+      "loss": 0.613,
+      "step": 16530
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1084580557718374e-05,
+      "loss": 0.4556,
+      "step": 16531
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1084003806557662e-05,
+      "loss": 0.3755,
+      "step": 16532
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.108342705539695e-05,
+      "loss": 0.5361,
+      "step": 16533
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 1.1082850304236239e-05,
+      "loss": 0.5386,
+      "step": 16534
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1082273553075526e-05,
+      "loss": 0.4369,
+      "step": 16535
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1081696801914814e-05,
+      "loss": 0.6531,
+      "step": 16536
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1081120050754103e-05,
+      "loss": 0.3166,
+      "step": 16537
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1080543299593391e-05,
+      "loss": 0.4503,
+      "step": 16538
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1079966548432679e-05,
+      "loss": 0.4179,
+      "step": 16539
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1079389797271968e-05,
+      "loss": 0.5053,
+      "step": 16540
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1078813046111256e-05,
+      "loss": 0.5881,
+      "step": 16541
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1078236294950544e-05,
+      "loss": 0.5019,
+      "step": 16542
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1077659543789833e-05,
+      "loss": 0.4535,
+      "step": 16543
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.107708279262912e-05,
+      "loss": 0.5072,
+      "step": 16544
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1076506041468408e-05,
+      "loss": 0.4255,
+      "step": 16545
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.10759292903077e-05,
+      "loss": 0.6016,
+      "step": 16546
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1075352539146987e-05,
+      "loss": 0.3927,
+      "step": 16547
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1074775787986273e-05,
+      "loss": 0.5223,
+      "step": 16548
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.107419903682556e-05,
+      "loss": 0.4832,
+      "step": 16549
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1073622285664852e-05,
+      "loss": 0.4533,
+      "step": 16550
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.107304553450414e-05,
+      "loss": 0.4891,
+      "step": 16551
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1072468783343426e-05,
+      "loss": 0.5792,
+      "step": 16552
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1071892032182717e-05,
+      "loss": 0.5429,
+      "step": 16553
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1071315281022004e-05,
+      "loss": 0.4028,
+      "step": 16554
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1070738529861292e-05,
+      "loss": 0.5267,
+      "step": 16555
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1070161778700581e-05,
+      "loss": 0.4698,
+      "step": 16556
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1069585027539869e-05,
+      "loss": 0.3964,
+      "step": 16557
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1069008276379157e-05,
+      "loss": 0.5001,
+      "step": 16558
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1068431525218446e-05,
+      "loss": 0.4593,
+      "step": 16559
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1067854774057734e-05,
+      "loss": 0.5074,
+      "step": 16560
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1067278022897021e-05,
+      "loss": 0.6331,
+      "step": 16561
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.106670127173631e-05,
+      "loss": 0.4131,
+      "step": 16562
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1066124520575599e-05,
+      "loss": 0.4538,
+      "step": 16563
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1065547769414886e-05,
+      "loss": 0.4797,
+      "step": 16564
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1064971018254176e-05,
+      "loss": 0.426,
+      "step": 16565
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1064394267093463e-05,
+      "loss": 0.3963,
+      "step": 16566
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1063817515932751e-05,
+      "loss": 0.3841,
+      "step": 16567
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.106324076477204e-05,
+      "loss": 0.5535,
+      "step": 16568
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1062664013611328e-05,
+      "loss": 0.4513,
+      "step": 16569
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1062087262450616e-05,
+      "loss": 0.5513,
+      "step": 16570
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1061510511289905e-05,
+      "loss": 0.452,
+      "step": 16571
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1060933760129193e-05,
+      "loss": 0.4313,
+      "step": 16572
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.106035700896848e-05,
+      "loss": 0.462,
+      "step": 16573
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.105978025780777e-05,
+      "loss": 0.5833,
+      "step": 16574
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1059203506647058e-05,
+      "loss": 0.4805,
+      "step": 16575
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1058626755486345e-05,
+      "loss": 0.5208,
+      "step": 16576
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1058050004325636e-05,
+      "loss": 0.5102,
+      "step": 16577
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1057473253164924e-05,
+      "loss": 0.4354,
+      "step": 16578
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.105689650200421e-05,
+      "loss": 0.4106,
+      "step": 16579
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1056319750843501e-05,
+      "loss": 0.4143,
+      "step": 16580
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1055742999682789e-05,
+      "loss": 0.5208,
+      "step": 16581
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1055166248522076e-05,
+      "loss": 0.4941,
+      "step": 16582
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1054589497361366e-05,
+      "loss": 0.3279,
+      "step": 16583
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1054012746200653e-05,
+      "loss": 0.4195,
+      "step": 16584
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1053435995039941e-05,
+      "loss": 0.4619,
+      "step": 16585
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1052859243879229e-05,
+      "loss": 0.5148,
+      "step": 16586
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1052282492718518e-05,
+      "loss": 0.445,
+      "step": 16587
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1051705741557806e-05,
+      "loss": 0.6156,
+      "step": 16588
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1051128990397094e-05,
+      "loss": 0.4257,
+      "step": 16589
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1050552239236383e-05,
+      "loss": 0.5695,
+      "step": 16590
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.104997548807567e-05,
+      "loss": 0.4274,
+      "step": 16591
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1049398736914958e-05,
+      "loss": 0.5673,
+      "step": 16592
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1048821985754248e-05,
+      "loss": 0.4416,
+      "step": 16593
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1048245234593535e-05,
+      "loss": 0.4887,
+      "step": 16594
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1047668483432823e-05,
+      "loss": 0.5156,
+      "step": 16595
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1047091732272112e-05,
+      "loss": 0.5222,
+      "step": 16596
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.10465149811114e-05,
+      "loss": 0.4009,
+      "step": 16597
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1045938229950688e-05,
+      "loss": 0.4895,
+      "step": 16598
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1045361478789977e-05,
+      "loss": 0.4074,
+      "step": 16599
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1044784727629265e-05,
+      "loss": 0.5072,
+      "step": 16600
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1044207976468553e-05,
+      "loss": 0.5311,
+      "step": 16601
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1043631225307842e-05,
+      "loss": 0.5937,
+      "step": 16602
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.104305447414713e-05,
+      "loss": 0.4545,
+      "step": 16603
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1042477722986417e-05,
+      "loss": 0.4964,
+      "step": 16604
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1041900971825707e-05,
+      "loss": 0.5033,
+      "step": 16605
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1041324220664994e-05,
+      "loss": 0.419,
+      "step": 16606
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1040747469504282e-05,
+      "loss": 0.479,
+      "step": 16607
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1040170718343573e-05,
+      "loss": 0.4949,
+      "step": 16608
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.103959396718286e-05,
+      "loss": 0.4262,
+      "step": 16609
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1039017216022147e-05,
+      "loss": 0.4492,
+      "step": 16610
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1038440464861438e-05,
+      "loss": 0.5186,
+      "step": 16611
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1037863713700726e-05,
+      "loss": 0.436,
+      "step": 16612
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1037286962540013e-05,
+      "loss": 0.5591,
+      "step": 16613
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1036710211379303e-05,
+      "loss": 0.5348,
+      "step": 16614
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.103613346021859e-05,
+      "loss": 0.4499,
+      "step": 16615
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1035556709057878e-05,
+      "loss": 0.5681,
+      "step": 16616
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1034979957897167e-05,
+      "loss": 0.38,
+      "step": 16617
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1034403206736455e-05,
+      "loss": 0.6306,
+      "step": 16618
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1033826455575743e-05,
+      "loss": 0.4683,
+      "step": 16619
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.103324970441503e-05,
+      "loss": 0.4576,
+      "step": 16620
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.103267295325432e-05,
+      "loss": 0.5652,
+      "step": 16621
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1032096202093608e-05,
+      "loss": 0.5346,
+      "step": 16622
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1031519450932895e-05,
+      "loss": 0.3963,
+      "step": 16623
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1030942699772185e-05,
+      "loss": 0.4902,
+      "step": 16624
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1030365948611472e-05,
+      "loss": 0.4355,
+      "step": 16625
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.102978919745076e-05,
+      "loss": 0.4193,
+      "step": 16626
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.102921244629005e-05,
+      "loss": 0.4585,
+      "step": 16627
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1028635695129337e-05,
+      "loss": 0.5201,
+      "step": 16628
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1028058943968625e-05,
+      "loss": 0.5643,
+      "step": 16629
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1027482192807914e-05,
+      "loss": 0.4983,
+      "step": 16630
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1026905441647202e-05,
+      "loss": 0.4258,
+      "step": 16631
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.102632869048649e-05,
+      "loss": 0.4796,
+      "step": 16632
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1025751939325779e-05,
+      "loss": 0.5041,
+      "step": 16633
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1025175188165067e-05,
+      "loss": 0.5394,
+      "step": 16634
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1024598437004354e-05,
+      "loss": 0.4859,
+      "step": 16635
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1024021685843644e-05,
+      "loss": 0.4177,
+      "step": 16636
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1023444934682931e-05,
+      "loss": 0.5107,
+      "step": 16637
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1022868183522219e-05,
+      "loss": 0.5325,
+      "step": 16638
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.102229143236151e-05,
+      "loss": 0.5389,
+      "step": 16639
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1021714681200798e-05,
+      "loss": 0.5042,
+      "step": 16640
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1021137930040084e-05,
+      "loss": 0.4057,
+      "step": 16641
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1020561178879375e-05,
+      "loss": 0.5626,
+      "step": 16642
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1019984427718662e-05,
+      "loss": 0.4772,
+      "step": 16643
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.101940767655795e-05,
+      "loss": 0.4708,
+      "step": 16644
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.101883092539724e-05,
+      "loss": 0.4887,
+      "step": 16645
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1018254174236527e-05,
+      "loss": 0.4284,
+      "step": 16646
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1017677423075815e-05,
+      "loss": 0.3647,
+      "step": 16647
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1017100671915104e-05,
+      "loss": 0.624,
+      "step": 16648
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1016523920754392e-05,
+      "loss": 0.5328,
+      "step": 16649
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.101594716959368e-05,
+      "loss": 0.5884,
+      "step": 16650
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1015370418432969e-05,
+      "loss": 0.4461,
+      "step": 16651
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1014793667272257e-05,
+      "loss": 0.4765,
+      "step": 16652
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1014216916111544e-05,
+      "loss": 0.4067,
+      "step": 16653
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1013640164950832e-05,
+      "loss": 0.4409,
+      "step": 16654
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1013063413790121e-05,
+      "loss": 0.4543,
+      "step": 16655
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1012486662629409e-05,
+      "loss": 0.4526,
+      "step": 16656
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1011909911468697e-05,
+      "loss": 0.4728,
+      "step": 16657
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1011333160307986e-05,
+      "loss": 0.4087,
+      "step": 16658
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1010756409147274e-05,
+      "loss": 0.5704,
+      "step": 16659
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1010179657986562e-05,
+      "loss": 0.4634,
+      "step": 16660
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1009602906825851e-05,
+      "loss": 0.5746,
+      "step": 16661
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1009026155665139e-05,
+      "loss": 0.4632,
+      "step": 16662
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1008449404504426e-05,
+      "loss": 0.4046,
+      "step": 16663
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1007872653343716e-05,
+      "loss": 0.5109,
+      "step": 16664
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1007295902183003e-05,
+      "loss": 0.4469,
+      "step": 16665
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1006719151022291e-05,
+      "loss": 0.5124,
+      "step": 16666
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.100614239986158e-05,
+      "loss": 0.5451,
+      "step": 16667
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1005565648700868e-05,
+      "loss": 0.4871,
+      "step": 16668
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1004988897540156e-05,
+      "loss": 0.5217,
+      "step": 16669
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1004412146379447e-05,
+      "loss": 0.4215,
+      "step": 16670
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1003835395218735e-05,
+      "loss": 0.5566,
+      "step": 16671
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.100325864405802e-05,
+      "loss": 0.4896,
+      "step": 16672
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1002681892897312e-05,
+      "loss": 0.4098,
+      "step": 16673
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.10021051417366e-05,
+      "loss": 0.4374,
+      "step": 16674
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1001528390575887e-05,
+      "loss": 0.4526,
+      "step": 16675
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1000951639415176e-05,
+      "loss": 0.5667,
+      "step": 16676
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.1000374888254464e-05,
+      "loss": 0.4855,
+      "step": 16677
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0999798137093752e-05,
+      "loss": 0.5388,
+      "step": 16678
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0999221385933041e-05,
+      "loss": 0.5162,
+      "step": 16679
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0998644634772329e-05,
+      "loss": 0.5774,
+      "step": 16680
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0998067883611617e-05,
+      "loss": 0.4764,
+      "step": 16681
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0997491132450906e-05,
+      "loss": 0.5823,
+      "step": 16682
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0996914381290194e-05,
+      "loss": 0.4068,
+      "step": 16683
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0996337630129481e-05,
+      "loss": 0.4162,
+      "step": 16684
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.099576087896877e-05,
+      "loss": 0.5297,
+      "step": 16685
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0995184127808058e-05,
+      "loss": 0.5792,
+      "step": 16686
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0994607376647346e-05,
+      "loss": 0.476,
+      "step": 16687
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0994030625486635e-05,
+      "loss": 0.4911,
+      "step": 16688
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0993453874325923e-05,
+      "loss": 0.4394,
+      "step": 16689
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.099287712316521e-05,
+      "loss": 0.4607,
+      "step": 16690
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0992300372004498e-05,
+      "loss": 0.5889,
+      "step": 16691
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0991723620843788e-05,
+      "loss": 0.4881,
+      "step": 16692
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0991146869683076e-05,
+      "loss": 0.6137,
+      "step": 16693
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0990570118522363e-05,
+      "loss": 0.4904,
+      "step": 16694
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0989993367361653e-05,
+      "loss": 0.5611,
+      "step": 16695
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.098941661620094e-05,
+      "loss": 0.4341,
+      "step": 16696
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0988839865040228e-05,
+      "loss": 0.5076,
+      "step": 16697
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0988263113879517e-05,
+      "loss": 0.4758,
+      "step": 16698
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0987686362718805e-05,
+      "loss": 0.3872,
+      "step": 16699
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0987109611558093e-05,
+      "loss": 0.4155,
+      "step": 16700
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0986532860397384e-05,
+      "loss": 0.5189,
+      "step": 16701
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0985956109236671e-05,
+      "loss": 0.468,
+      "step": 16702
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0985379358075957e-05,
+      "loss": 0.528,
+      "step": 16703
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0984802606915249e-05,
+      "loss": 0.5534,
+      "step": 16704
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0984225855754536e-05,
+      "loss": 0.5929,
+      "step": 16705
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0983649104593824e-05,
+      "loss": 0.5435,
+      "step": 16706
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0983072353433113e-05,
+      "loss": 0.5218,
+      "step": 16707
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0982495602272401e-05,
+      "loss": 0.5286,
+      "step": 16708
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0981918851111689e-05,
+      "loss": 0.5997,
+      "step": 16709
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0981342099950978e-05,
+      "loss": 0.4881,
+      "step": 16710
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0980765348790266e-05,
+      "loss": 0.5484,
+      "step": 16711
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0980188597629553e-05,
+      "loss": 0.3751,
+      "step": 16712
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.0979611846468843e-05,
+      "loss": 0.5334,
+      "step": 16713
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.097903509530813e-05,
+      "loss": 0.4321,
+      "step": 16714
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0978458344147418e-05,
+      "loss": 0.4488,
+      "step": 16715
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0977881592986708e-05,
+      "loss": 0.4041,
+      "step": 16716
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0977304841825995e-05,
+      "loss": 0.4661,
+      "step": 16717
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0976728090665283e-05,
+      "loss": 0.4938,
+      "step": 16718
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0976151339504572e-05,
+      "loss": 0.4674,
+      "step": 16719
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.097557458834386e-05,
+      "loss": 0.3469,
+      "step": 16720
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0974997837183148e-05,
+      "loss": 0.6293,
+      "step": 16721
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0974421086022437e-05,
+      "loss": 0.4885,
+      "step": 16722
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0973844334861725e-05,
+      "loss": 0.523,
+      "step": 16723
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0973267583701012e-05,
+      "loss": 0.4447,
+      "step": 16724
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.09726908325403e-05,
+      "loss": 0.5502,
+      "step": 16725
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.097211408137959e-05,
+      "loss": 0.4292,
+      "step": 16726
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0971537330218877e-05,
+      "loss": 0.5129,
+      "step": 16727
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0970960579058165e-05,
+      "loss": 0.5279,
+      "step": 16728
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0970383827897456e-05,
+      "loss": 0.4049,
+      "step": 16729
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0969807076736742e-05,
+      "loss": 0.5052,
+      "step": 16730
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096923032557603e-05,
+      "loss": 0.4494,
+      "step": 16731
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096865357441532e-05,
+      "loss": 0.587,
+      "step": 16732
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0968076823254608e-05,
+      "loss": 0.6107,
+      "step": 16733
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0967500072093894e-05,
+      "loss": 0.4214,
+      "step": 16734
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0966923320933185e-05,
+      "loss": 0.5848,
+      "step": 16735
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0966346569772473e-05,
+      "loss": 0.5466,
+      "step": 16736
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096576981861176e-05,
+      "loss": 0.537,
+      "step": 16737
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096519306745105e-05,
+      "loss": 0.5469,
+      "step": 16738
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0964616316290338e-05,
+      "loss": 0.5871,
+      "step": 16739
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0964039565129626e-05,
+      "loss": 0.5536,
+      "step": 16740
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0963462813968915e-05,
+      "loss": 0.3726,
+      "step": 16741
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0962886062808203e-05,
+      "loss": 0.4996,
+      "step": 16742
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096230931164749e-05,
+      "loss": 0.2966,
+      "step": 16743
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.096173256048678e-05,
+      "loss": 0.4062,
+      "step": 16744
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0961155809326067e-05,
+      "loss": 0.5943,
+      "step": 16745
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0960579058165355e-05,
+      "loss": 0.5352,
+      "step": 16746
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0960002307004644e-05,
+      "loss": 0.5234,
+      "step": 16747
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0959425555843932e-05,
+      "loss": 0.4454,
+      "step": 16748
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.095884880468322e-05,
+      "loss": 0.4522,
+      "step": 16749
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0958272053522509e-05,
+      "loss": 0.3565,
+      "step": 16750
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0957695302361797e-05,
+      "loss": 0.3948,
+      "step": 16751
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0957118551201085e-05,
+      "loss": 0.421,
+      "step": 16752
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0956541800040374e-05,
+      "loss": 0.4644,
+      "step": 16753
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0955965048879662e-05,
+      "loss": 0.4489,
+      "step": 16754
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.095538829771895e-05,
+      "loss": 0.5749,
+      "step": 16755
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0954811546558239e-05,
+      "loss": 0.4123,
+      "step": 16756
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0954234795397526e-05,
+      "loss": 0.366,
+      "step": 16757
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0953658044236814e-05,
+      "loss": 0.4431,
+      "step": 16758
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0953081293076102e-05,
+      "loss": 0.499,
+      "step": 16759
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0952504541915393e-05,
+      "loss": 0.429,
+      "step": 16760
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0951927790754679e-05,
+      "loss": 0.5121,
+      "step": 16761
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0951351039593966e-05,
+      "loss": 0.482,
+      "step": 16762
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0950774288433258e-05,
+      "loss": 0.4764,
+      "step": 16763
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0950197537272545e-05,
+      "loss": 0.4653,
+      "step": 16764
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0949620786111831e-05,
+      "loss": 0.5122,
+      "step": 16765
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0949044034951122e-05,
+      "loss": 0.5542,
+      "step": 16766
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.094846728379041e-05,
+      "loss": 0.4797,
+      "step": 16767
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0947890532629698e-05,
+      "loss": 0.5029,
+      "step": 16768
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0947313781468987e-05,
+      "loss": 0.4918,
+      "step": 16769
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0946737030308275e-05,
+      "loss": 0.4185,
+      "step": 16770
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0946160279147562e-05,
+      "loss": 0.5095,
+      "step": 16771
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0945583527986852e-05,
+      "loss": 0.3618,
+      "step": 16772
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.094500677682614e-05,
+      "loss": 0.5318,
+      "step": 16773
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0944430025665427e-05,
+      "loss": 0.4683,
+      "step": 16774
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0943853274504717e-05,
+      "loss": 0.4416,
+      "step": 16775
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0943276523344004e-05,
+      "loss": 0.4552,
+      "step": 16776
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0942699772183292e-05,
+      "loss": 0.4477,
+      "step": 16777
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0942123021022581e-05,
+      "loss": 0.5895,
+      "step": 16778
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0941546269861869e-05,
+      "loss": 0.5644,
+      "step": 16779
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0940969518701157e-05,
+      "loss": 0.5718,
+      "step": 16780
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0940392767540446e-05,
+      "loss": 0.4363,
+      "step": 16781
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0939816016379734e-05,
+      "loss": 0.5264,
+      "step": 16782
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0939239265219021e-05,
+      "loss": 0.4136,
+      "step": 16783
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.093866251405831e-05,
+      "loss": 0.4059,
+      "step": 16784
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0938085762897598e-05,
+      "loss": 0.4999,
+      "step": 16785
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0937509011736886e-05,
+      "loss": 0.6797,
+      "step": 16786
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0936932260576176e-05,
+      "loss": 0.5018,
+      "step": 16787
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0936355509415463e-05,
+      "loss": 0.489,
+      "step": 16788
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0935778758254751e-05,
+      "loss": 0.4091,
+      "step": 16789
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0935202007094042e-05,
+      "loss": 0.4875,
+      "step": 16790
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.093462525593333e-05,
+      "loss": 0.3622,
+      "step": 16791
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0934048504772616e-05,
+      "loss": 0.4657,
+      "step": 16792
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0933471753611903e-05,
+      "loss": 0.4959,
+      "step": 16793
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0932895002451194e-05,
+      "loss": 0.4564,
+      "step": 16794
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0932318251290482e-05,
+      "loss": 0.3919,
+      "step": 16795
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0931741500129768e-05,
+      "loss": 0.3639,
+      "step": 16796
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0931164748969059e-05,
+      "loss": 0.5221,
+      "step": 16797
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0930587997808347e-05,
+      "loss": 0.46,
+      "step": 16798
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0930011246647635e-05,
+      "loss": 0.4332,
+      "step": 16799
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0929434495486924e-05,
+      "loss": 0.4901,
+      "step": 16800
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0928857744326212e-05,
+      "loss": 0.4305,
+      "step": 16801
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.09282809931655e-05,
+      "loss": 0.4962,
+      "step": 16802
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0927704242004789e-05,
+      "loss": 0.498,
+      "step": 16803
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0927127490844076e-05,
+      "loss": 0.5908,
+      "step": 16804
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0926550739683364e-05,
+      "loss": 0.5534,
+      "step": 16805
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0925973988522653e-05,
+      "loss": 0.378,
+      "step": 16806
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0925397237361941e-05,
+      "loss": 0.4866,
+      "step": 16807
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0924820486201229e-05,
+      "loss": 0.3959,
+      "step": 16808
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0924243735040518e-05,
+      "loss": 0.5114,
+      "step": 16809
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0923666983879806e-05,
+      "loss": 0.5739,
+      "step": 16810
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0923090232719093e-05,
+      "loss": 0.5068,
+      "step": 16811
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0922513481558383e-05,
+      "loss": 0.595,
+      "step": 16812
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.092193673039767e-05,
+      "loss": 0.5317,
+      "step": 16813
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0921359979236958e-05,
+      "loss": 0.4693,
+      "step": 16814
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0920783228076248e-05,
+      "loss": 0.4531,
+      "step": 16815
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0920206476915535e-05,
+      "loss": 0.4249,
+      "step": 16816
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0919629725754823e-05,
+      "loss": 0.5642,
+      "step": 16817
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0919052974594112e-05,
+      "loss": 0.5311,
+      "step": 16818
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.09184762234334e-05,
+      "loss": 0.5827,
+      "step": 16819
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0917899472272688e-05,
+      "loss": 0.4346,
+      "step": 16820
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0917322721111979e-05,
+      "loss": 0.4448,
+      "step": 16821
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0916745969951267e-05,
+      "loss": 0.5035,
+      "step": 16822
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0916169218790552e-05,
+      "loss": 0.506,
+      "step": 16823
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0915592467629844e-05,
+      "loss": 0.5231,
+      "step": 16824
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0915015716469131e-05,
+      "loss": 0.4244,
+      "step": 16825
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0914438965308419e-05,
+      "loss": 0.5379,
+      "step": 16826
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0913862214147708e-05,
+      "loss": 0.464,
+      "step": 16827
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0913285462986996e-05,
+      "loss": 0.4058,
+      "step": 16828
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0912708711826284e-05,
+      "loss": 0.4641,
+      "step": 16829
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0912131960665571e-05,
+      "loss": 0.4721,
+      "step": 16830
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.091155520950486e-05,
+      "loss": 0.5959,
+      "step": 16831
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0910978458344148e-05,
+      "loss": 0.5021,
+      "step": 16832
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0910401707183436e-05,
+      "loss": 0.4568,
+      "step": 16833
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0909824956022726e-05,
+      "loss": 0.5661,
+      "step": 16834
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0909248204862013e-05,
+      "loss": 0.4711,
+      "step": 16835
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0908671453701301e-05,
+      "loss": 0.5779,
+      "step": 16836
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090809470254059e-05,
+      "loss": 0.5836,
+      "step": 16837
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0907517951379878e-05,
+      "loss": 0.543,
+      "step": 16838
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0906941200219166e-05,
+      "loss": 0.4891,
+      "step": 16839
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0906364449058455e-05,
+      "loss": 0.5106,
+      "step": 16840
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0905787697897743e-05,
+      "loss": 0.5069,
+      "step": 16841
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090521094673703e-05,
+      "loss": 0.5406,
+      "step": 16842
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090463419557632e-05,
+      "loss": 0.471,
+      "step": 16843
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0904057444415607e-05,
+      "loss": 0.5144,
+      "step": 16844
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0903480693254895e-05,
+      "loss": 0.4529,
+      "step": 16845
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0902903942094185e-05,
+      "loss": 0.4372,
+      "step": 16846
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0902327190933472e-05,
+      "loss": 0.4793,
+      "step": 16847
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090175043977276e-05,
+      "loss": 0.5756,
+      "step": 16848
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.090117368861205e-05,
+      "loss": 0.5546,
+      "step": 16849
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0900596937451337e-05,
+      "loss": 0.5081,
+      "step": 16850
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0900020186290625e-05,
+      "loss": 0.4912,
+      "step": 16851
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0899443435129916e-05,
+      "loss": 0.4648,
+      "step": 16852
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0898866683969203e-05,
+      "loss": 0.5025,
+      "step": 16853
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.089828993280849e-05,
+      "loss": 0.446,
+      "step": 16854
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.089771318164778e-05,
+      "loss": 0.5417,
+      "step": 16855
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0897136430487068e-05,
+      "loss": 0.5383,
+      "step": 16856
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0896559679326356e-05,
+      "loss": 0.4734,
+      "step": 16857
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0895982928165645e-05,
+      "loss": 0.3773,
+      "step": 16858
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0895406177004933e-05,
+      "loss": 0.5172,
+      "step": 16859
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.089482942584422e-05,
+      "loss": 0.5356,
+      "step": 16860
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.089425267468351e-05,
+      "loss": 0.3653,
+      "step": 16861
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0893675923522798e-05,
+      "loss": 0.5297,
+      "step": 16862
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0893099172362085e-05,
+      "loss": 0.5954,
+      "step": 16863
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0892522421201373e-05,
+      "loss": 0.5072,
+      "step": 16864
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0891945670040662e-05,
+      "loss": 0.5191,
+      "step": 16865
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.089136891887995e-05,
+      "loss": 0.4898,
+      "step": 16866
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0890792167719238e-05,
+      "loss": 0.5355,
+      "step": 16867
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0890215416558527e-05,
+      "loss": 0.3856,
+      "step": 16868
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0889638665397815e-05,
+      "loss": 0.4831,
+      "step": 16869
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0889061914237102e-05,
+      "loss": 0.4439,
+      "step": 16870
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0888485163076392e-05,
+      "loss": 0.4722,
+      "step": 16871
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.088790841191568e-05,
+      "loss": 0.5449,
+      "step": 16872
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0887331660754967e-05,
+      "loss": 0.4448,
+      "step": 16873
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0886754909594257e-05,
+      "loss": 0.5853,
+      "step": 16874
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0886178158433544e-05,
+      "loss": 0.507,
+      "step": 16875
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0885601407272832e-05,
+      "loss": 0.4454,
+      "step": 16876
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0885024656112121e-05,
+      "loss": 0.5955,
+      "step": 16877
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0884447904951409e-05,
+      "loss": 0.4309,
+      "step": 16878
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0883871153790697e-05,
+      "loss": 0.4611,
+      "step": 16879
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0883294402629986e-05,
+      "loss": 0.4072,
+      "step": 16880
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0882717651469274e-05,
+      "loss": 0.5461,
+      "step": 16881
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0882140900308561e-05,
+      "loss": 0.472,
+      "step": 16882
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0881564149147853e-05,
+      "loss": 0.5558,
+      "step": 16883
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.088098739798714e-05,
+      "loss": 0.4126,
+      "step": 16884
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0880410646826426e-05,
+      "loss": 0.5039,
+      "step": 16885
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0879833895665717e-05,
+      "loss": 0.5133,
+      "step": 16886
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0879257144505005e-05,
+      "loss": 0.3946,
+      "step": 16887
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0878680393344293e-05,
+      "loss": 0.5621,
+      "step": 16888
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0878103642183582e-05,
+      "loss": 0.6073,
+      "step": 16889
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.087752689102287e-05,
+      "loss": 0.4959,
+      "step": 16890
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 1.0876950139862157e-05,
+      "loss": 0.4627,
+      "step": 16891
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0876373388701447e-05,
+      "loss": 0.5424,
+      "step": 16892
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0875796637540734e-05,
+      "loss": 0.3695,
+      "step": 16893
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0875219886380022e-05,
+      "loss": 0.5717,
+      "step": 16894
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0874643135219312e-05,
+      "loss": 0.6464,
+      "step": 16895
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.08740663840586e-05,
+      "loss": 0.4736,
+      "step": 16896
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0873489632897887e-05,
+      "loss": 0.5149,
+      "step": 16897
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0872912881737175e-05,
+      "loss": 0.5706,
+      "step": 16898
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0872336130576464e-05,
+      "loss": 0.4113,
+      "step": 16899
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0871759379415752e-05,
+      "loss": 0.4991,
+      "step": 16900
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.087118262825504e-05,
+      "loss": 0.4329,
+      "step": 16901
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0870605877094329e-05,
+      "loss": 0.5556,
+      "step": 16902
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0870029125933616e-05,
+      "loss": 0.3772,
+      "step": 16903
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0869452374772904e-05,
+      "loss": 0.459,
+      "step": 16904
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0868875623612193e-05,
+      "loss": 0.6666,
+      "step": 16905
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0868298872451481e-05,
+      "loss": 0.5013,
+      "step": 16906
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0867722121290769e-05,
+      "loss": 0.5937,
+      "step": 16907
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0867145370130058e-05,
+      "loss": 0.3703,
+      "step": 16908
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0866568618969346e-05,
+      "loss": 0.406,
+      "step": 16909
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0865991867808634e-05,
+      "loss": 0.5792,
+      "step": 16910
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0865415116647923e-05,
+      "loss": 0.3493,
+      "step": 16911
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.086483836548721e-05,
+      "loss": 0.5397,
+      "step": 16912
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0864261614326498e-05,
+      "loss": 0.5068,
+      "step": 16913
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.086368486316579e-05,
+      "loss": 0.4468,
+      "step": 16914
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0863108112005077e-05,
+      "loss": 0.4443,
+      "step": 16915
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0862531360844363e-05,
+      "loss": 0.4987,
+      "step": 16916
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0861954609683654e-05,
+      "loss": 0.4604,
+      "step": 16917
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0861377858522942e-05,
+      "loss": 0.4639,
+      "step": 16918
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.086080110736223e-05,
+      "loss": 0.439,
+      "step": 16919
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0860224356201519e-05,
+      "loss": 0.4854,
+      "step": 16920
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0859647605040807e-05,
+      "loss": 0.4708,
+      "step": 16921
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0859070853880094e-05,
+      "loss": 0.486,
+      "step": 16922
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0858494102719384e-05,
+      "loss": 0.4298,
+      "step": 16923
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0857917351558671e-05,
+      "loss": 0.498,
+      "step": 16924
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0857340600397959e-05,
+      "loss": 0.4534,
+      "step": 16925
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0856763849237248e-05,
+      "loss": 0.5417,
+      "step": 16926
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0856187098076536e-05,
+      "loss": 0.4474,
+      "step": 16927
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0855610346915824e-05,
+      "loss": 0.387,
+      "step": 16928
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0855033595755113e-05,
+      "loss": 0.5433,
+      "step": 16929
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0854456844594401e-05,
+      "loss": 0.4783,
+      "step": 16930
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0853880093433689e-05,
+      "loss": 0.5118,
+      "step": 16931
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0853303342272976e-05,
+      "loss": 0.3459,
+      "step": 16932
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0852726591112266e-05,
+      "loss": 0.3692,
+      "step": 16933
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0852149839951553e-05,
+      "loss": 0.6398,
+      "step": 16934
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0851573088790841e-05,
+      "loss": 0.4495,
+      "step": 16935
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.085099633763013e-05,
+      "loss": 0.5815,
+      "step": 16936
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0850419586469418e-05,
+      "loss": 0.4424,
+      "step": 16937
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0849842835308706e-05,
+      "loss": 0.5394,
+      "step": 16938
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0849266084147995e-05,
+      "loss": 0.4159,
+      "step": 16939
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0848689332987283e-05,
+      "loss": 0.4129,
+      "step": 16940
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.084811258182657e-05,
+      "loss": 0.412,
+      "step": 16941
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.084753583066586e-05,
+      "loss": 0.4463,
+      "step": 16942
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0846959079505148e-05,
+      "loss": 0.3719,
+      "step": 16943
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0846382328344435e-05,
+      "loss": 0.4822,
+      "step": 16944
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0845805577183726e-05,
+      "loss": 0.5268,
+      "step": 16945
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0845228826023014e-05,
+      "loss": 0.5128,
+      "step": 16946
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.08446520748623e-05,
+      "loss": 0.3953,
+      "step": 16947
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0844075323701591e-05,
+      "loss": 0.5317,
+      "step": 16948
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0843498572540879e-05,
+      "loss": 0.4766,
+      "step": 16949
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0842921821380166e-05,
+      "loss": 0.4656,
+      "step": 16950
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0842345070219456e-05,
+      "loss": 0.6128,
+      "step": 16951
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0841768319058743e-05,
+      "loss": 0.5591,
+      "step": 16952
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0841191567898031e-05,
+      "loss": 0.445,
+      "step": 16953
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.084061481673732e-05,
+      "loss": 0.3499,
+      "step": 16954
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0840038065576608e-05,
+      "loss": 0.5536,
+      "step": 16955
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0839461314415896e-05,
+      "loss": 0.3331,
+      "step": 16956
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0838884563255185e-05,
+      "loss": 0.6877,
+      "step": 16957
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0838307812094473e-05,
+      "loss": 0.5105,
+      "step": 16958
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.083773106093376e-05,
+      "loss": 0.5166,
+      "step": 16959
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.083715430977305e-05,
+      "loss": 0.5521,
+      "step": 16960
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0836577558612338e-05,
+      "loss": 0.4258,
+      "step": 16961
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0836000807451625e-05,
+      "loss": 0.4438,
+      "step": 16962
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0835424056290915e-05,
+      "loss": 0.4083,
+      "step": 16963
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0834847305130202e-05,
+      "loss": 0.4022,
+      "step": 16964
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.083427055396949e-05,
+      "loss": 0.4179,
+      "step": 16965
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.083369380280878e-05,
+      "loss": 0.4506,
+      "step": 16966
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0833117051648067e-05,
+      "loss": 0.4921,
+      "step": 16967
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0832540300487355e-05,
+      "loss": 0.4987,
+      "step": 16968
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0831963549326643e-05,
+      "loss": 0.4716,
+      "step": 16969
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0831386798165932e-05,
+      "loss": 0.4757,
+      "step": 16970
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.083081004700522e-05,
+      "loss": 0.4911,
+      "step": 16971
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0830233295844507e-05,
+      "loss": 0.5672,
+      "step": 16972
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0829656544683797e-05,
+      "loss": 0.5029,
+      "step": 16973
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0829079793523084e-05,
+      "loss": 0.5143,
+      "step": 16974
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0828503042362372e-05,
+      "loss": 0.4407,
+      "step": 16975
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0827926291201663e-05,
+      "loss": 0.5182,
+      "step": 16976
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0827349540040951e-05,
+      "loss": 0.4824,
+      "step": 16977
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0826772788880237e-05,
+      "loss": 0.4885,
+      "step": 16978
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0826196037719528e-05,
+      "loss": 0.512,
+      "step": 16979
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0825619286558816e-05,
+      "loss": 0.4705,
+      "step": 16980
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0825042535398103e-05,
+      "loss": 0.5065,
+      "step": 16981
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0824465784237393e-05,
+      "loss": 0.4495,
+      "step": 16982
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.082388903307668e-05,
+      "loss": 0.4219,
+      "step": 16983
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0823312281915968e-05,
+      "loss": 0.5348,
+      "step": 16984
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0822735530755257e-05,
+      "loss": 0.3902,
+      "step": 16985
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0822158779594545e-05,
+      "loss": 0.3821,
+      "step": 16986
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0821582028433833e-05,
+      "loss": 0.5178,
+      "step": 16987
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0821005277273122e-05,
+      "loss": 0.4589,
+      "step": 16988
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.082042852611241e-05,
+      "loss": 0.5372,
+      "step": 16989
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0819851774951698e-05,
+      "loss": 0.4013,
+      "step": 16990
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0819275023790987e-05,
+      "loss": 0.4056,
+      "step": 16991
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0818698272630275e-05,
+      "loss": 0.4499,
+      "step": 16992
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0818121521469562e-05,
+      "loss": 0.489,
+      "step": 16993
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0817544770308852e-05,
+      "loss": 0.4412,
+      "step": 16994
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.081696801914814e-05,
+      "loss": 0.4749,
+      "step": 16995
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0816391267987427e-05,
+      "loss": 0.4655,
+      "step": 16996
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0815814516826716e-05,
+      "loss": 0.4234,
+      "step": 16997
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0815237765666004e-05,
+      "loss": 0.4016,
+      "step": 16998
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0814661014505292e-05,
+      "loss": 0.5069,
+      "step": 16999
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0814084263344581e-05,
+      "loss": 0.5077,
+      "step": 17000
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0813507512183869e-05,
+      "loss": 0.4639,
+      "step": 17001
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0812930761023157e-05,
+      "loss": 0.5368,
+      "step": 17002
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0812354009862444e-05,
+      "loss": 0.4074,
+      "step": 17003
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0811777258701734e-05,
+      "loss": 0.5704,
+      "step": 17004
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0811200507541021e-05,
+      "loss": 0.5033,
+      "step": 17005
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0810623756380309e-05,
+      "loss": 0.3876,
+      "step": 17006
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.08100470052196e-05,
+      "loss": 0.5429,
+      "step": 17007
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0809470254058888e-05,
+      "loss": 0.3767,
+      "step": 17008
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0808893502898174e-05,
+      "loss": 0.4885,
+      "step": 17009
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0808316751737465e-05,
+      "loss": 0.3988,
+      "step": 17010
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0807740000576752e-05,
+      "loss": 0.3231,
+      "step": 17011
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.080716324941604e-05,
+      "loss": 0.5,
+      "step": 17012
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.080658649825533e-05,
+      "loss": 0.3695,
+      "step": 17013
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0806009747094617e-05,
+      "loss": 0.5279,
+      "step": 17014
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0805432995933905e-05,
+      "loss": 0.4159,
+      "step": 17015
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0804856244773194e-05,
+      "loss": 0.4255,
+      "step": 17016
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0804279493612482e-05,
+      "loss": 0.5271,
+      "step": 17017
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.080370274245177e-05,
+      "loss": 0.5282,
+      "step": 17018
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0803125991291059e-05,
+      "loss": 0.3495,
+      "step": 17019
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0802549240130347e-05,
+      "loss": 0.4475,
+      "step": 17020
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0801972488969634e-05,
+      "loss": 0.5105,
+      "step": 17021
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0801395737808924e-05,
+      "loss": 0.4943,
+      "step": 17022
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0800818986648211e-05,
+      "loss": 0.4183,
+      "step": 17023
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.08002422354875e-05,
+      "loss": 0.4717,
+      "step": 17024
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0799665484326789e-05,
+      "loss": 0.5333,
+      "step": 17025
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0799088733166076e-05,
+      "loss": 0.3348,
+      "step": 17026
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0798511982005364e-05,
+      "loss": 0.4154,
+      "step": 17027
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0797935230844653e-05,
+      "loss": 0.4022,
+      "step": 17028
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0797358479683941e-05,
+      "loss": 0.4365,
+      "step": 17029
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0796781728523229e-05,
+      "loss": 0.52,
+      "step": 17030
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0796204977362518e-05,
+      "loss": 0.5278,
+      "step": 17031
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0795628226201806e-05,
+      "loss": 0.5601,
+      "step": 17032
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0795051475041093e-05,
+      "loss": 0.4209,
+      "step": 17033
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0794474723880384e-05,
+      "loss": 0.5602,
+      "step": 17034
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.079389797271967e-05,
+      "loss": 0.4292,
+      "step": 17035
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0793321221558958e-05,
+      "loss": 0.705,
+      "step": 17036
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0792744470398246e-05,
+      "loss": 0.5474,
+      "step": 17037
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0792167719237537e-05,
+      "loss": 0.5026,
+      "step": 17038
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0791590968076825e-05,
+      "loss": 0.5999,
+      "step": 17039
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.079101421691611e-05,
+      "loss": 0.3882,
+      "step": 17040
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0790437465755402e-05,
+      "loss": 0.5879,
+      "step": 17041
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.078986071459469e-05,
+      "loss": 0.5106,
+      "step": 17042
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0789283963433977e-05,
+      "loss": 0.6967,
+      "step": 17043
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0788707212273266e-05,
+      "loss": 0.4023,
+      "step": 17044
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0788130461112554e-05,
+      "loss": 0.4842,
+      "step": 17045
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0787553709951842e-05,
+      "loss": 0.5184,
+      "step": 17046
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0786976958791131e-05,
+      "loss": 0.5978,
+      "step": 17047
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0786400207630419e-05,
+      "loss": 0.5481,
+      "step": 17048
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0785823456469707e-05,
+      "loss": 0.4977,
+      "step": 17049
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0785246705308996e-05,
+      "loss": 0.4983,
+      "step": 17050
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0784669954148284e-05,
+      "loss": 0.4142,
+      "step": 17051
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0784093202987571e-05,
+      "loss": 0.4754,
+      "step": 17052
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.078351645182686e-05,
+      "loss": 0.5568,
+      "step": 17053
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0782939700666148e-05,
+      "loss": 0.6025,
+      "step": 17054
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0782362949505436e-05,
+      "loss": 0.5153,
+      "step": 17055
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0781786198344725e-05,
+      "loss": 0.4857,
+      "step": 17056
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0781209447184013e-05,
+      "loss": 0.4846,
+      "step": 17057
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.07806326960233e-05,
+      "loss": 0.6376,
+      "step": 17058
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.078005594486259e-05,
+      "loss": 0.6276,
+      "step": 17059
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0779479193701878e-05,
+      "loss": 0.5299,
+      "step": 17060
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0778902442541166e-05,
+      "loss": 0.3408,
+      "step": 17061
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0778325691380455e-05,
+      "loss": 0.4975,
+      "step": 17062
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0777748940219743e-05,
+      "loss": 0.4045,
+      "step": 17063
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.077717218905903e-05,
+      "loss": 0.4553,
+      "step": 17064
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0776595437898321e-05,
+      "loss": 0.4865,
+      "step": 17065
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0776018686737607e-05,
+      "loss": 0.392,
+      "step": 17066
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0775441935576895e-05,
+      "loss": 0.5788,
+      "step": 17067
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0774865184416186e-05,
+      "loss": 0.4522,
+      "step": 17068
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0774288433255474e-05,
+      "loss": 0.5541,
+      "step": 17069
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.0773711682094761e-05,
+      "loss": 0.3671,
+      "step": 17070
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0773134930934051e-05,
+      "loss": 0.3972,
+      "step": 17071
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0772558179773339e-05,
+      "loss": 0.4635,
+      "step": 17072
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0771981428612626e-05,
+      "loss": 0.4914,
+      "step": 17073
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0771404677451914e-05,
+      "loss": 0.4629,
+      "step": 17074
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0770827926291203e-05,
+      "loss": 0.4581,
+      "step": 17075
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0770251175130491e-05,
+      "loss": 0.3894,
+      "step": 17076
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0769674423969779e-05,
+      "loss": 0.4713,
+      "step": 17077
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0769097672809068e-05,
+      "loss": 0.3838,
+      "step": 17078
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0768520921648356e-05,
+      "loss": 0.3679,
+      "step": 17079
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0767944170487643e-05,
+      "loss": 0.3592,
+      "step": 17080
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0767367419326933e-05,
+      "loss": 0.4504,
+      "step": 17081
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.076679066816622e-05,
+      "loss": 0.4761,
+      "step": 17082
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0766213917005508e-05,
+      "loss": 0.5119,
+      "step": 17083
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0765637165844798e-05,
+      "loss": 0.4528,
+      "step": 17084
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0765060414684085e-05,
+      "loss": 0.4283,
+      "step": 17085
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0764483663523373e-05,
+      "loss": 0.4958,
+      "step": 17086
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0763906912362662e-05,
+      "loss": 0.445,
+      "step": 17087
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.076333016120195e-05,
+      "loss": 0.4677,
+      "step": 17088
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0762753410041238e-05,
+      "loss": 0.3488,
+      "step": 17089
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0762176658880527e-05,
+      "loss": 0.4632,
+      "step": 17090
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0761599907719815e-05,
+      "loss": 0.6043,
+      "step": 17091
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0761023156559102e-05,
+      "loss": 0.5099,
+      "step": 17092
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0760446405398392e-05,
+      "loss": 0.3999,
+      "step": 17093
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.075986965423768e-05,
+      "loss": 0.5092,
+      "step": 17094
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0759292903076967e-05,
+      "loss": 0.6424,
+      "step": 17095
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0758716151916258e-05,
+      "loss": 0.4876,
+      "step": 17096
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0758139400755546e-05,
+      "loss": 0.4182,
+      "step": 17097
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0757562649594832e-05,
+      "loss": 0.5783,
+      "step": 17098
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0756985898434123e-05,
+      "loss": 0.4461,
+      "step": 17099
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.075640914727341e-05,
+      "loss": 0.4091,
+      "step": 17100
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0755832396112698e-05,
+      "loss": 0.4298,
+      "step": 17101
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0755255644951988e-05,
+      "loss": 0.4342,
+      "step": 17102
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0754678893791275e-05,
+      "loss": 0.3983,
+      "step": 17103
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0754102142630563e-05,
+      "loss": 0.4459,
+      "step": 17104
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0753525391469852e-05,
+      "loss": 0.5037,
+      "step": 17105
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.075294864030914e-05,
+      "loss": 0.5167,
+      "step": 17106
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0752371889148428e-05,
+      "loss": 0.4489,
+      "step": 17107
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0751795137987716e-05,
+      "loss": 0.4449,
+      "step": 17108
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0751218386827005e-05,
+      "loss": 0.4377,
+      "step": 17109
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0750641635666293e-05,
+      "loss": 0.4327,
+      "step": 17110
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.075006488450558e-05,
+      "loss": 0.5749,
+      "step": 17111
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.074948813334487e-05,
+      "loss": 0.4992,
+      "step": 17112
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0748911382184157e-05,
+      "loss": 0.5256,
+      "step": 17113
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0748334631023445e-05,
+      "loss": 0.5833,
+      "step": 17114
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0747757879862734e-05,
+      "loss": 0.4123,
+      "step": 17115
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0747181128702022e-05,
+      "loss": 0.4506,
+      "step": 17116
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.074660437754131e-05,
+      "loss": 0.4444,
+      "step": 17117
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.07460276263806e-05,
+      "loss": 0.4642,
+      "step": 17118
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0745450875219887e-05,
+      "loss": 0.4645,
+      "step": 17119
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0744874124059175e-05,
+      "loss": 0.4226,
+      "step": 17120
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0744297372898464e-05,
+      "loss": 0.5385,
+      "step": 17121
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0743720621737752e-05,
+      "loss": 0.5379,
+      "step": 17122
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.074314387057704e-05,
+      "loss": 0.4426,
+      "step": 17123
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0742567119416329e-05,
+      "loss": 0.6733,
+      "step": 17124
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0741990368255616e-05,
+      "loss": 0.5558,
+      "step": 17125
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0741413617094904e-05,
+      "loss": 0.3965,
+      "step": 17126
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0740836865934195e-05,
+      "loss": 0.4873,
+      "step": 17127
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0740260114773483e-05,
+      "loss": 0.6031,
+      "step": 17128
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0739683363612769e-05,
+      "loss": 0.3821,
+      "step": 17129
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.073910661245206e-05,
+      "loss": 0.627,
+      "step": 17130
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0738529861291348e-05,
+      "loss": 0.5242,
+      "step": 17131
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0737953110130635e-05,
+      "loss": 0.6057,
+      "step": 17132
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0737376358969925e-05,
+      "loss": 0.4813,
+      "step": 17133
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0736799607809212e-05,
+      "loss": 0.348,
+      "step": 17134
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.07362228566485e-05,
+      "loss": 0.4434,
+      "step": 17135
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.073564610548779e-05,
+      "loss": 0.5173,
+      "step": 17136
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0735069354327077e-05,
+      "loss": 0.4605,
+      "step": 17137
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0734492603166365e-05,
+      "loss": 0.5485,
+      "step": 17138
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0733915852005654e-05,
+      "loss": 0.463,
+      "step": 17139
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0733339100844942e-05,
+      "loss": 0.4868,
+      "step": 17140
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.073276234968423e-05,
+      "loss": 0.5384,
+      "step": 17141
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0732185598523517e-05,
+      "loss": 0.5062,
+      "step": 17142
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0731608847362807e-05,
+      "loss": 0.4383,
+      "step": 17143
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0731032096202094e-05,
+      "loss": 0.5268,
+      "step": 17144
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0730455345041382e-05,
+      "loss": 0.3666,
+      "step": 17145
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0729878593880671e-05,
+      "loss": 0.5041,
+      "step": 17146
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0729301842719959e-05,
+      "loss": 0.5628,
+      "step": 17147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0728725091559247e-05,
+      "loss": 0.5024,
+      "step": 17148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0728148340398536e-05,
+      "loss": 0.4074,
+      "step": 17149
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0727571589237824e-05,
+      "loss": 0.4041,
+      "step": 17150
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0726994838077111e-05,
+      "loss": 0.4801,
+      "step": 17151
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.07264180869164e-05,
+      "loss": 0.4056,
+      "step": 17152
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0725841335755688e-05,
+      "loss": 0.4285,
+      "step": 17153
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0725264584594976e-05,
+      "loss": 0.4658,
+      "step": 17154
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0724687833434266e-05,
+      "loss": 0.4351,
+      "step": 17155
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0724111082273553e-05,
+      "loss": 0.569,
+      "step": 17156
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0723534331112841e-05,
+      "loss": 0.4032,
+      "step": 17157
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0722957579952132e-05,
+      "loss": 0.5377,
+      "step": 17158
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.072238082879142e-05,
+      "loss": 0.5641,
+      "step": 17159
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0721804077630706e-05,
+      "loss": 0.3851,
+      "step": 17160
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0721227326469997e-05,
+      "loss": 0.4692,
+      "step": 17161
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0720650575309284e-05,
+      "loss": 0.4718,
+      "step": 17162
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0720073824148572e-05,
+      "loss": 0.4349,
+      "step": 17163
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0719497072987861e-05,
+      "loss": 0.5786,
+      "step": 17164
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.071892032182715e-05,
+      "loss": 0.43,
+      "step": 17165
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0718343570666437e-05,
+      "loss": 0.5694,
+      "step": 17166
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0717766819505726e-05,
+      "loss": 0.5527,
+      "step": 17167
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0717190068345014e-05,
+      "loss": 0.3847,
+      "step": 17168
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0716613317184302e-05,
+      "loss": 0.5618,
+      "step": 17169
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0716036566023591e-05,
+      "loss": 0.4636,
+      "step": 17170
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0715459814862879e-05,
+      "loss": 0.4565,
+      "step": 17171
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0714883063702166e-05,
+      "loss": 0.585,
+      "step": 17172
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0714306312541456e-05,
+      "loss": 0.3768,
+      "step": 17173
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0713729561380743e-05,
+      "loss": 0.3695,
+      "step": 17174
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0713152810220031e-05,
+      "loss": 0.4989,
+      "step": 17175
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0712576059059319e-05,
+      "loss": 0.5209,
+      "step": 17176
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0711999307898608e-05,
+      "loss": 0.6029,
+      "step": 17177
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0711422556737896e-05,
+      "loss": 0.2499,
+      "step": 17178
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0710845805577184e-05,
+      "loss": 0.591,
+      "step": 17179
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0710269054416473e-05,
+      "loss": 0.5292,
+      "step": 17180
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.070969230325576e-05,
+      "loss": 0.5688,
+      "step": 17181
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0709115552095048e-05,
+      "loss": 0.534,
+      "step": 17182
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0708538800934338e-05,
+      "loss": 0.5614,
+      "step": 17183
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0707962049773625e-05,
+      "loss": 0.4394,
+      "step": 17184
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0707385298612913e-05,
+      "loss": 0.4471,
+      "step": 17185
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0706808547452202e-05,
+      "loss": 0.5539,
+      "step": 17186
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.070623179629149e-05,
+      "loss": 0.5107,
+      "step": 17187
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0705655045130778e-05,
+      "loss": 0.5614,
+      "step": 17188
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0705078293970069e-05,
+      "loss": 0.5634,
+      "step": 17189
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0704501542809357e-05,
+      "loss": 0.477,
+      "step": 17190
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0703924791648643e-05,
+      "loss": 0.4895,
+      "step": 17191
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0703348040487934e-05,
+      "loss": 0.5592,
+      "step": 17192
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0702771289327221e-05,
+      "loss": 0.5414,
+      "step": 17193
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0702194538166509e-05,
+      "loss": 0.4124,
+      "step": 17194
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0701617787005798e-05,
+      "loss": 0.5522,
+      "step": 17195
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0701041035845086e-05,
+      "loss": 0.4323,
+      "step": 17196
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0700464284684374e-05,
+      "loss": 0.5613,
+      "step": 17197
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0699887533523663e-05,
+      "loss": 0.4136,
+      "step": 17198
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.069931078236295e-05,
+      "loss": 0.4347,
+      "step": 17199
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0698734031202238e-05,
+      "loss": 0.4545,
+      "step": 17200
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0698157280041528e-05,
+      "loss": 0.4185,
+      "step": 17201
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0697580528880816e-05,
+      "loss": 0.5403,
+      "step": 17202
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0697003777720103e-05,
+      "loss": 0.4311,
+      "step": 17203
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0696427026559393e-05,
+      "loss": 0.4904,
+      "step": 17204
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.069585027539868e-05,
+      "loss": 0.3708,
+      "step": 17205
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0695273524237968e-05,
+      "loss": 0.5287,
+      "step": 17206
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0694696773077257e-05,
+      "loss": 0.4668,
+      "step": 17207
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0694120021916545e-05,
+      "loss": 0.3825,
+      "step": 17208
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0693543270755833e-05,
+      "loss": 0.4528,
+      "step": 17209
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0692966519595122e-05,
+      "loss": 0.4861,
+      "step": 17210
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.069238976843441e-05,
+      "loss": 0.4739,
+      "step": 17211
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0691813017273697e-05,
+      "loss": 0.4949,
+      "step": 17212
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0691236266112985e-05,
+      "loss": 0.4846,
+      "step": 17213
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0690659514952275e-05,
+      "loss": 0.4942,
+      "step": 17214
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0690082763791562e-05,
+      "loss": 0.4468,
+      "step": 17215
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.068950601263085e-05,
+      "loss": 0.5211,
+      "step": 17216
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.068892926147014e-05,
+      "loss": 0.4633,
+      "step": 17217
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0688352510309427e-05,
+      "loss": 0.5549,
+      "step": 17218
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0687775759148715e-05,
+      "loss": 0.4232,
+      "step": 17219
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0687199007988006e-05,
+      "loss": 0.487,
+      "step": 17220
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0686622256827293e-05,
+      "loss": 0.4801,
+      "step": 17221
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.068604550566658e-05,
+      "loss": 0.435,
+      "step": 17222
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.068546875450587e-05,
+      "loss": 0.4433,
+      "step": 17223
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0684892003345158e-05,
+      "loss": 0.4151,
+      "step": 17224
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0684315252184446e-05,
+      "loss": 0.3266,
+      "step": 17225
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0683738501023735e-05,
+      "loss": 0.4847,
+      "step": 17226
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0683161749863023e-05,
+      "loss": 0.4572,
+      "step": 17227
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.068258499870231e-05,
+      "loss": 0.5158,
+      "step": 17228
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.06820082475416e-05,
+      "loss": 0.4264,
+      "step": 17229
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0681431496380888e-05,
+      "loss": 0.4306,
+      "step": 17230
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0680854745220175e-05,
+      "loss": 0.463,
+      "step": 17231
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0680277994059465e-05,
+      "loss": 0.4657,
+      "step": 17232
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0679701242898752e-05,
+      "loss": 0.5261,
+      "step": 17233
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.067912449173804e-05,
+      "loss": 0.3691,
+      "step": 17234
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.067854774057733e-05,
+      "loss": 0.5975,
+      "step": 17235
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0677970989416617e-05,
+      "loss": 0.5229,
+      "step": 17236
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0677394238255905e-05,
+      "loss": 0.4815,
+      "step": 17237
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0676817487095194e-05,
+      "loss": 0.4064,
+      "step": 17238
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0676240735934482e-05,
+      "loss": 0.4359,
+      "step": 17239
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.067566398477377e-05,
+      "loss": 0.6096,
+      "step": 17240
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0675087233613059e-05,
+      "loss": 0.4588,
+      "step": 17241
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0674510482452347e-05,
+      "loss": 0.3997,
+      "step": 17242
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0673933731291634e-05,
+      "loss": 0.4365,
+      "step": 17243
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0673356980130924e-05,
+      "loss": 0.5381,
+      "step": 17244
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0672780228970211e-05,
+      "loss": 0.455,
+      "step": 17245
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0672203477809499e-05,
+      "loss": 0.4314,
+      "step": 17246
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0671626726648787e-05,
+      "loss": 0.5544,
+      "step": 17247
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0671049975488076e-05,
+      "loss": 0.5766,
+      "step": 17248
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 1.0670473224327364e-05,
+      "loss": 0.4454,
+      "step": 17249
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0669896473166652e-05,
+      "loss": 0.441,
+      "step": 17250
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0669319722005943e-05,
+      "loss": 0.4736,
+      "step": 17251
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.066874297084523e-05,
+      "loss": 0.5899,
+      "step": 17252
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0668166219684516e-05,
+      "loss": 0.4495,
+      "step": 17253
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0667589468523807e-05,
+      "loss": 0.4614,
+      "step": 17254
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0667012717363095e-05,
+      "loss": 0.3631,
+      "step": 17255
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0666435966202383e-05,
+      "loss": 0.4211,
+      "step": 17256
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0665859215041672e-05,
+      "loss": 0.4931,
+      "step": 17257
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.066528246388096e-05,
+      "loss": 0.4824,
+      "step": 17258
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0664705712720247e-05,
+      "loss": 0.4145,
+      "step": 17259
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0664128961559537e-05,
+      "loss": 0.4015,
+      "step": 17260
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0663552210398825e-05,
+      "loss": 0.5199,
+      "step": 17261
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0662975459238112e-05,
+      "loss": 0.4744,
+      "step": 17262
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0662398708077402e-05,
+      "loss": 0.4865,
+      "step": 17263
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.066182195691669e-05,
+      "loss": 0.3264,
+      "step": 17264
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0661245205755977e-05,
+      "loss": 0.3938,
+      "step": 17265
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0660668454595266e-05,
+      "loss": 0.4565,
+      "step": 17266
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0660091703434554e-05,
+      "loss": 0.515,
+      "step": 17267
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0659514952273842e-05,
+      "loss": 0.3288,
+      "step": 17268
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0658938201113131e-05,
+      "loss": 0.5321,
+      "step": 17269
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0658361449952419e-05,
+      "loss": 0.5518,
+      "step": 17270
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0657784698791706e-05,
+      "loss": 0.4685,
+      "step": 17271
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0657207947630996e-05,
+      "loss": 0.5601,
+      "step": 17272
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0656631196470284e-05,
+      "loss": 0.4586,
+      "step": 17273
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0656054445309571e-05,
+      "loss": 0.4791,
+      "step": 17274
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.065547769414886e-05,
+      "loss": 0.5561,
+      "step": 17275
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0654900942988148e-05,
+      "loss": 0.634,
+      "step": 17276
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0654324191827436e-05,
+      "loss": 0.6665,
+      "step": 17277
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0653747440666727e-05,
+      "loss": 0.5265,
+      "step": 17278
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0653170689506013e-05,
+      "loss": 0.5284,
+      "step": 17279
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.06525939383453e-05,
+      "loss": 0.497,
+      "step": 17280
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0652017187184588e-05,
+      "loss": 0.5838,
+      "step": 17281
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.065144043602388e-05,
+      "loss": 0.5176,
+      "step": 17282
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0650863684863167e-05,
+      "loss": 0.3662,
+      "step": 17283
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0650286933702453e-05,
+      "loss": 0.5134,
+      "step": 17284
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0649710182541744e-05,
+      "loss": 0.4999,
+      "step": 17285
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0649133431381032e-05,
+      "loss": 0.5032,
+      "step": 17286
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.064855668022032e-05,
+      "loss": 0.5409,
+      "step": 17287
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0647979929059609e-05,
+      "loss": 0.4597,
+      "step": 17288
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0647403177898897e-05,
+      "loss": 0.4788,
+      "step": 17289
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0646826426738184e-05,
+      "loss": 0.4525,
+      "step": 17290
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0646249675577474e-05,
+      "loss": 0.4218,
+      "step": 17291
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0645672924416761e-05,
+      "loss": 0.4201,
+      "step": 17292
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0645096173256049e-05,
+      "loss": 0.3355,
+      "step": 17293
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0644519422095338e-05,
+      "loss": 0.5139,
+      "step": 17294
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0643942670934626e-05,
+      "loss": 0.4644,
+      "step": 17295
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0643365919773914e-05,
+      "loss": 0.3755,
+      "step": 17296
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0642789168613203e-05,
+      "loss": 0.4951,
+      "step": 17297
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0642212417452491e-05,
+      "loss": 0.45,
+      "step": 17298
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0641635666291779e-05,
+      "loss": 0.364,
+      "step": 17299
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0641058915131068e-05,
+      "loss": 0.4148,
+      "step": 17300
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0640482163970356e-05,
+      "loss": 0.5637,
+      "step": 17301
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0639905412809643e-05,
+      "loss": 0.4201,
+      "step": 17302
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0639328661648933e-05,
+      "loss": 0.4768,
+      "step": 17303
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.063875191048822e-05,
+      "loss": 0.5211,
+      "step": 17304
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0638175159327508e-05,
+      "loss": 0.402,
+      "step": 17305
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0637598408166797e-05,
+      "loss": 0.4725,
+      "step": 17306
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0637021657006085e-05,
+      "loss": 0.4577,
+      "step": 17307
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0636444905845373e-05,
+      "loss": 0.3637,
+      "step": 17308
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0635868154684664e-05,
+      "loss": 0.6186,
+      "step": 17309
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.063529140352395e-05,
+      "loss": 0.4635,
+      "step": 17310
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0634714652363238e-05,
+      "loss": 0.416,
+      "step": 17311
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0634137901202529e-05,
+      "loss": 0.5669,
+      "step": 17312
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0633561150041816e-05,
+      "loss": 0.4091,
+      "step": 17313
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0632984398881104e-05,
+      "loss": 0.4489,
+      "step": 17314
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.063240764772039e-05,
+      "loss": 0.5708,
+      "step": 17315
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0631830896559681e-05,
+      "loss": 0.5235,
+      "step": 17316
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0631254145398969e-05,
+      "loss": 0.5029,
+      "step": 17317
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0630677394238256e-05,
+      "loss": 0.4435,
+      "step": 17318
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0630100643077546e-05,
+      "loss": 0.6592,
+      "step": 17319
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0629523891916834e-05,
+      "loss": 0.3319,
+      "step": 17320
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0628947140756121e-05,
+      "loss": 0.4645,
+      "step": 17321
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.062837038959541e-05,
+      "loss": 0.5462,
+      "step": 17322
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0627793638434698e-05,
+      "loss": 0.4225,
+      "step": 17323
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0627216887273986e-05,
+      "loss": 0.632,
+      "step": 17324
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0626640136113275e-05,
+      "loss": 0.5518,
+      "step": 17325
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0626063384952563e-05,
+      "loss": 0.4142,
+      "step": 17326
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.062548663379185e-05,
+      "loss": 0.4719,
+      "step": 17327
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.062490988263114e-05,
+      "loss": 0.348,
+      "step": 17328
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0624333131470428e-05,
+      "loss": 0.4734,
+      "step": 17329
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0623756380309715e-05,
+      "loss": 0.512,
+      "step": 17330
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0623179629149005e-05,
+      "loss": 0.4649,
+      "step": 17331
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0622602877988293e-05,
+      "loss": 0.6011,
+      "step": 17332
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.062202612682758e-05,
+      "loss": 0.5511,
+      "step": 17333
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.062144937566687e-05,
+      "loss": 0.2807,
+      "step": 17334
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0620872624506157e-05,
+      "loss": 0.4212,
+      "step": 17335
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0620295873345445e-05,
+      "loss": 0.4975,
+      "step": 17336
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0619719122184734e-05,
+      "loss": 0.5338,
+      "step": 17337
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0619142371024022e-05,
+      "loss": 0.4657,
+      "step": 17338
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.061856561986331e-05,
+      "loss": 0.5192,
+      "step": 17339
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.06179888687026e-05,
+      "loss": 0.3046,
+      "step": 17340
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0617412117541887e-05,
+      "loss": 0.5567,
+      "step": 17341
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0616835366381174e-05,
+      "loss": 0.3913,
+      "step": 17342
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0616258615220466e-05,
+      "loss": 0.4787,
+      "step": 17343
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0615681864059753e-05,
+      "loss": 0.4674,
+      "step": 17344
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0615105112899041e-05,
+      "loss": 0.5396,
+      "step": 17345
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.061452836173833e-05,
+      "loss": 0.546,
+      "step": 17346
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0613951610577618e-05,
+      "loss": 0.4926,
+      "step": 17347
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0613374859416906e-05,
+      "loss": 0.4529,
+      "step": 17348
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0612798108256195e-05,
+      "loss": 0.5676,
+      "step": 17349
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0612221357095483e-05,
+      "loss": 0.5074,
+      "step": 17350
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.061164460593477e-05,
+      "loss": 0.4978,
+      "step": 17351
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0611067854774058e-05,
+      "loss": 0.4751,
+      "step": 17352
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0610491103613347e-05,
+      "loss": 0.5645,
+      "step": 17353
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0609914352452635e-05,
+      "loss": 0.5339,
+      "step": 17354
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0609337601291923e-05,
+      "loss": 0.4841,
+      "step": 17355
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0608760850131212e-05,
+      "loss": 0.553,
+      "step": 17356
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.06081840989705e-05,
+      "loss": 0.4189,
+      "step": 17357
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0607607347809788e-05,
+      "loss": 0.4317,
+      "step": 17358
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0607030596649077e-05,
+      "loss": 0.5252,
+      "step": 17359
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0606453845488365e-05,
+      "loss": 0.5689,
+      "step": 17360
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0605877094327652e-05,
+      "loss": 0.436,
+      "step": 17361
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0605300343166942e-05,
+      "loss": 0.4038,
+      "step": 17362
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.060472359200623e-05,
+      "loss": 0.536,
+      "step": 17363
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0604146840845517e-05,
+      "loss": 0.496,
+      "step": 17364
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0603570089684806e-05,
+      "loss": 0.5321,
+      "step": 17365
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0602993338524094e-05,
+      "loss": 0.4972,
+      "step": 17366
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0602416587363382e-05,
+      "loss": 0.4341,
+      "step": 17367
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0601839836202671e-05,
+      "loss": 0.4434,
+      "step": 17368
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0601263085041959e-05,
+      "loss": 0.4743,
+      "step": 17369
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0600686333881247e-05,
+      "loss": 0.5941,
+      "step": 17370
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0600109582720538e-05,
+      "loss": 0.4262,
+      "step": 17371
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0599532831559824e-05,
+      "loss": 0.386,
+      "step": 17372
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0598956080399111e-05,
+      "loss": 0.5128,
+      "step": 17373
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0598379329238402e-05,
+      "loss": 0.4605,
+      "step": 17374
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.059780257807769e-05,
+      "loss": 0.4879,
+      "step": 17375
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0597225826916978e-05,
+      "loss": 0.4175,
+      "step": 17376
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0596649075756267e-05,
+      "loss": 0.4963,
+      "step": 17377
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0596072324595555e-05,
+      "loss": 0.5421,
+      "step": 17378
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0595495573434843e-05,
+      "loss": 0.5486,
+      "step": 17379
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0594918822274132e-05,
+      "loss": 0.4203,
+      "step": 17380
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.059434207111342e-05,
+      "loss": 0.4136,
+      "step": 17381
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0593765319952707e-05,
+      "loss": 0.5158,
+      "step": 17382
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0593188568791997e-05,
+      "loss": 0.5404,
+      "step": 17383
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0592611817631284e-05,
+      "loss": 0.4642,
+      "step": 17384
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0592035066470572e-05,
+      "loss": 0.4677,
+      "step": 17385
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.059145831530986e-05,
+      "loss": 0.4997,
+      "step": 17386
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0590881564149149e-05,
+      "loss": 0.4414,
+      "step": 17387
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0590304812988437e-05,
+      "loss": 0.5232,
+      "step": 17388
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0589728061827724e-05,
+      "loss": 0.4411,
+      "step": 17389
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0589151310667014e-05,
+      "loss": 0.4501,
+      "step": 17390
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0588574559506302e-05,
+      "loss": 0.4748,
+      "step": 17391
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.058799780834559e-05,
+      "loss": 0.4616,
+      "step": 17392
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0587421057184879e-05,
+      "loss": 0.3501,
+      "step": 17393
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0586844306024166e-05,
+      "loss": 0.4655,
+      "step": 17394
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0586267554863454e-05,
+      "loss": 0.4587,
+      "step": 17395
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0585690803702743e-05,
+      "loss": 0.4749,
+      "step": 17396
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0585114052542031e-05,
+      "loss": 0.4312,
+      "step": 17397
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0584537301381319e-05,
+      "loss": 0.5252,
+      "step": 17398
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0583960550220608e-05,
+      "loss": 0.5389,
+      "step": 17399
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0583383799059896e-05,
+      "loss": 0.4718,
+      "step": 17400
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0582807047899183e-05,
+      "loss": 0.5112,
+      "step": 17401
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0582230296738475e-05,
+      "loss": 0.5488,
+      "step": 17402
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.058165354557776e-05,
+      "loss": 0.5038,
+      "step": 17403
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0581076794417048e-05,
+      "loss": 0.5137,
+      "step": 17404
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.058050004325634e-05,
+      "loss": 0.5733,
+      "step": 17405
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0579923292095627e-05,
+      "loss": 0.4854,
+      "step": 17406
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0579346540934915e-05,
+      "loss": 0.5238,
+      "step": 17407
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0578769789774204e-05,
+      "loss": 0.51,
+      "step": 17408
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0578193038613492e-05,
+      "loss": 0.4725,
+      "step": 17409
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.057761628745278e-05,
+      "loss": 0.4003,
+      "step": 17410
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0577039536292069e-05,
+      "loss": 0.4949,
+      "step": 17411
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0576462785131356e-05,
+      "loss": 0.356,
+      "step": 17412
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0575886033970644e-05,
+      "loss": 0.428,
+      "step": 17413
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0575309282809934e-05,
+      "loss": 0.5054,
+      "step": 17414
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0574732531649221e-05,
+      "loss": 0.4975,
+      "step": 17415
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0574155780488509e-05,
+      "loss": 0.4563,
+      "step": 17416
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0573579029327798e-05,
+      "loss": 0.4863,
+      "step": 17417
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0573002278167086e-05,
+      "loss": 0.548,
+      "step": 17418
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0572425527006374e-05,
+      "loss": 0.3852,
+      "step": 17419
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0571848775845661e-05,
+      "loss": 0.454,
+      "step": 17420
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.057127202468495e-05,
+      "loss": 0.4178,
+      "step": 17421
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0570695273524238e-05,
+      "loss": 0.5506,
+      "step": 17422
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0570118522363526e-05,
+      "loss": 0.471,
+      "step": 17423
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0569541771202815e-05,
+      "loss": 0.5321,
+      "step": 17424
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0568965020042103e-05,
+      "loss": 0.4443,
+      "step": 17425
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.056838826888139e-05,
+      "loss": 0.4804,
+      "step": 17426
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.056781151772068e-05,
+      "loss": 0.3575,
+      "step": 17427
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 1.0567234766559968e-05,
+      "loss": 0.6877,
+      "step": 17428
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0566658015399256e-05,
+      "loss": 0.4884,
+      "step": 17429
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0566081264238545e-05,
+      "loss": 0.5855,
+      "step": 17430
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0565504513077833e-05,
+      "loss": 0.5631,
+      "step": 17431
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.056492776191712e-05,
+      "loss": 0.4776,
+      "step": 17432
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0564351010756411e-05,
+      "loss": 0.5387,
+      "step": 17433
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0563774259595697e-05,
+      "loss": 0.4616,
+      "step": 17434
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0563197508434985e-05,
+      "loss": 0.4815,
+      "step": 17435
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0562620757274276e-05,
+      "loss": 0.5595,
+      "step": 17436
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0562044006113564e-05,
+      "loss": 0.3451,
+      "step": 17437
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0561467254952852e-05,
+      "loss": 0.4709,
+      "step": 17438
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0560890503792141e-05,
+      "loss": 0.5197,
+      "step": 17439
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0560313752631429e-05,
+      "loss": 0.4284,
+      "step": 17440
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0559737001470716e-05,
+      "loss": 0.3491,
+      "step": 17441
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0559160250310006e-05,
+      "loss": 0.5133,
+      "step": 17442
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0558583499149293e-05,
+      "loss": 0.5661,
+      "step": 17443
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0558006747988581e-05,
+      "loss": 0.4104,
+      "step": 17444
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.055742999682787e-05,
+      "loss": 0.4489,
+      "step": 17445
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0556853245667158e-05,
+      "loss": 0.4377,
+      "step": 17446
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0556276494506446e-05,
+      "loss": 0.4113,
+      "step": 17447
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0555699743345735e-05,
+      "loss": 0.4351,
+      "step": 17448
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0555122992185023e-05,
+      "loss": 0.4382,
+      "step": 17449
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.055454624102431e-05,
+      "loss": 0.3408,
+      "step": 17450
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.05539694898636e-05,
+      "loss": 0.458,
+      "step": 17451
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0553392738702888e-05,
+      "loss": 0.55,
+      "step": 17452
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0552815987542175e-05,
+      "loss": 0.5474,
+      "step": 17453
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0552239236381465e-05,
+      "loss": 0.5515,
+      "step": 17454
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0551662485220752e-05,
+      "loss": 0.4644,
+      "step": 17455
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.055108573406004e-05,
+      "loss": 0.4689,
+      "step": 17456
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0550508982899328e-05,
+      "loss": 0.5104,
+      "step": 17457
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0549932231738617e-05,
+      "loss": 0.459,
+      "step": 17458
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0549355480577905e-05,
+      "loss": 0.4661,
+      "step": 17459
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0548778729417192e-05,
+      "loss": 0.5418,
+      "step": 17460
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0548201978256482e-05,
+      "loss": 0.537,
+      "step": 17461
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.054762522709577e-05,
+      "loss": 0.4917,
+      "step": 17462
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0547048475935057e-05,
+      "loss": 0.5187,
+      "step": 17463
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0546471724774348e-05,
+      "loss": 0.4743,
+      "step": 17464
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0545894973613636e-05,
+      "loss": 0.4554,
+      "step": 17465
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0545318222452922e-05,
+      "loss": 0.5148,
+      "step": 17466
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0544741471292213e-05,
+      "loss": 0.4721,
+      "step": 17467
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.05441647201315e-05,
+      "loss": 0.3958,
+      "step": 17468
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0543587968970788e-05,
+      "loss": 0.5217,
+      "step": 17469
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0543011217810078e-05,
+      "loss": 0.5097,
+      "step": 17470
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0542434466649365e-05,
+      "loss": 0.4078,
+      "step": 17471
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0541857715488653e-05,
+      "loss": 0.4383,
+      "step": 17472
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0541280964327943e-05,
+      "loss": 0.5213,
+      "step": 17473
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.054070421316723e-05,
+      "loss": 0.4047,
+      "step": 17474
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0540127462006518e-05,
+      "loss": 0.5009,
+      "step": 17475
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0539550710845807e-05,
+      "loss": 0.4914,
+      "step": 17476
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0538973959685095e-05,
+      "loss": 0.3843,
+      "step": 17477
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0538397208524383e-05,
+      "loss": 0.4624,
+      "step": 17478
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0537820457363672e-05,
+      "loss": 0.4547,
+      "step": 17479
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.053724370620296e-05,
+      "loss": 0.415,
+      "step": 17480
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0536666955042247e-05,
+      "loss": 0.4991,
+      "step": 17481
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0536090203881537e-05,
+      "loss": 0.4685,
+      "step": 17482
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0535513452720824e-05,
+      "loss": 0.582,
+      "step": 17483
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0534936701560112e-05,
+      "loss": 0.475,
+      "step": 17484
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0534359950399402e-05,
+      "loss": 0.3984,
+      "step": 17485
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.053378319923869e-05,
+      "loss": 0.4209,
+      "step": 17486
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0533206448077977e-05,
+      "loss": 0.572,
+      "step": 17487
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0532629696917266e-05,
+      "loss": 0.4607,
+      "step": 17488
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0532052945756554e-05,
+      "loss": 0.5592,
+      "step": 17489
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0531476194595842e-05,
+      "loss": 0.4851,
+      "step": 17490
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.053089944343513e-05,
+      "loss": 0.5173,
+      "step": 17491
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0530322692274419e-05,
+      "loss": 0.6183,
+      "step": 17492
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0529745941113706e-05,
+      "loss": 0.4369,
+      "step": 17493
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0529169189952994e-05,
+      "loss": 0.3703,
+      "step": 17494
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0528592438792285e-05,
+      "loss": 0.326,
+      "step": 17495
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0528015687631573e-05,
+      "loss": 0.3433,
+      "step": 17496
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0527438936470859e-05,
+      "loss": 0.4519,
+      "step": 17497
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.052686218531015e-05,
+      "loss": 0.4545,
+      "step": 17498
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0526285434149438e-05,
+      "loss": 0.3959,
+      "step": 17499
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0525708682988725e-05,
+      "loss": 0.3773,
+      "step": 17500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0525131931828015e-05,
+      "loss": 0.4234,
+      "step": 17501
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0524555180667302e-05,
+      "loss": 0.5976,
+      "step": 17502
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.052397842950659e-05,
+      "loss": 0.6013,
+      "step": 17503
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.052340167834588e-05,
+      "loss": 0.482,
+      "step": 17504
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0522824927185167e-05,
+      "loss": 0.5637,
+      "step": 17505
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0522248176024455e-05,
+      "loss": 0.4056,
+      "step": 17506
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0521671424863744e-05,
+      "loss": 0.4975,
+      "step": 17507
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0521094673703032e-05,
+      "loss": 0.5088,
+      "step": 17508
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.052051792254232e-05,
+      "loss": 0.5015,
+      "step": 17509
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0519941171381609e-05,
+      "loss": 0.4141,
+      "step": 17510
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0519364420220897e-05,
+      "loss": 0.392,
+      "step": 17511
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0518787669060184e-05,
+      "loss": 0.5176,
+      "step": 17512
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0518210917899474e-05,
+      "loss": 0.4586,
+      "step": 17513
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0517634166738761e-05,
+      "loss": 0.4818,
+      "step": 17514
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0517057415578049e-05,
+      "loss": 0.6353,
+      "step": 17515
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0516480664417338e-05,
+      "loss": 0.4716,
+      "step": 17516
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0515903913256626e-05,
+      "loss": 0.436,
+      "step": 17517
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0515327162095914e-05,
+      "loss": 0.4702,
+      "step": 17518
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0514750410935203e-05,
+      "loss": 0.4206,
+      "step": 17519
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.051417365977449e-05,
+      "loss": 0.4657,
+      "step": 17520
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0513596908613779e-05,
+      "loss": 0.4939,
+      "step": 17521
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.051302015745307e-05,
+      "loss": 0.4222,
+      "step": 17522
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0512443406292356e-05,
+      "loss": 0.4519,
+      "step": 17523
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0511866655131643e-05,
+      "loss": 0.4662,
+      "step": 17524
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0511289903970931e-05,
+      "loss": 0.314,
+      "step": 17525
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0510713152810222e-05,
+      "loss": 0.4717,
+      "step": 17526
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.051013640164951e-05,
+      "loss": 0.401,
+      "step": 17527
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0509559650488796e-05,
+      "loss": 0.3713,
+      "step": 17528
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0508982899328087e-05,
+      "loss": 0.4919,
+      "step": 17529
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0508406148167374e-05,
+      "loss": 0.4464,
+      "step": 17530
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0507829397006662e-05,
+      "loss": 0.4238,
+      "step": 17531
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0507252645845952e-05,
+      "loss": 0.3744,
+      "step": 17532
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.050667589468524e-05,
+      "loss": 0.4387,
+      "step": 17533
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0506099143524527e-05,
+      "loss": 0.461,
+      "step": 17534
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0505522392363816e-05,
+      "loss": 0.3721,
+      "step": 17535
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0504945641203104e-05,
+      "loss": 0.4282,
+      "step": 17536
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0504368890042392e-05,
+      "loss": 0.4211,
+      "step": 17537
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0503792138881681e-05,
+      "loss": 0.4261,
+      "step": 17538
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0503215387720969e-05,
+      "loss": 0.4621,
+      "step": 17539
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0502638636560256e-05,
+      "loss": 0.622,
+      "step": 17540
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0502061885399546e-05,
+      "loss": 0.5303,
+      "step": 17541
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0501485134238833e-05,
+      "loss": 0.5023,
+      "step": 17542
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0500908383078121e-05,
+      "loss": 0.3583,
+      "step": 17543
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.050033163191741e-05,
+      "loss": 0.5201,
+      "step": 17544
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0499754880756698e-05,
+      "loss": 0.6045,
+      "step": 17545
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0499178129595986e-05,
+      "loss": 0.5023,
+      "step": 17546
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0498601378435275e-05,
+      "loss": 0.489,
+      "step": 17547
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0498024627274563e-05,
+      "loss": 0.4901,
+      "step": 17548
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.049744787611385e-05,
+      "loss": 0.5389,
+      "step": 17549
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.049687112495314e-05,
+      "loss": 0.393,
+      "step": 17550
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0496294373792428e-05,
+      "loss": 0.3775,
+      "step": 17551
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0495717622631715e-05,
+      "loss": 0.4329,
+      "step": 17552
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0495140871471006e-05,
+      "loss": 0.4425,
+      "step": 17553
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0494564120310292e-05,
+      "loss": 0.487,
+      "step": 17554
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.049398736914958e-05,
+      "loss": 0.4253,
+      "step": 17555
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0493410617988871e-05,
+      "loss": 0.4467,
+      "step": 17556
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0492833866828159e-05,
+      "loss": 0.5032,
+      "step": 17557
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0492257115667447e-05,
+      "loss": 0.4436,
+      "step": 17558
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0491680364506733e-05,
+      "loss": 0.3367,
+      "step": 17559
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0491103613346024e-05,
+      "loss": 0.525,
+      "step": 17560
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0490526862185311e-05,
+      "loss": 0.4438,
+      "step": 17561
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0489950111024599e-05,
+      "loss": 0.5461,
+      "step": 17562
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0489373359863888e-05,
+      "loss": 0.3784,
+      "step": 17563
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0488796608703176e-05,
+      "loss": 0.4937,
+      "step": 17564
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0488219857542464e-05,
+      "loss": 0.4741,
+      "step": 17565
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0487643106381753e-05,
+      "loss": 0.3361,
+      "step": 17566
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.048706635522104e-05,
+      "loss": 0.4228,
+      "step": 17567
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0486489604060329e-05,
+      "loss": 0.5787,
+      "step": 17568
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0485912852899618e-05,
+      "loss": 0.4908,
+      "step": 17569
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0485336101738906e-05,
+      "loss": 0.3917,
+      "step": 17570
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0484759350578193e-05,
+      "loss": 0.4303,
+      "step": 17571
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0484182599417483e-05,
+      "loss": 0.3901,
+      "step": 17572
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.048360584825677e-05,
+      "loss": 0.4836,
+      "step": 17573
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0483029097096058e-05,
+      "loss": 0.4132,
+      "step": 17574
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0482452345935347e-05,
+      "loss": 0.5303,
+      "step": 17575
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0481875594774635e-05,
+      "loss": 0.4655,
+      "step": 17576
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0481298843613923e-05,
+      "loss": 0.4358,
+      "step": 17577
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0480722092453212e-05,
+      "loss": 0.4613,
+      "step": 17578
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.04801453412925e-05,
+      "loss": 0.5863,
+      "step": 17579
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0479568590131788e-05,
+      "loss": 0.5414,
+      "step": 17580
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0478991838971077e-05,
+      "loss": 0.4415,
+      "step": 17581
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0478415087810365e-05,
+      "loss": 0.4402,
+      "step": 17582
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0477838336649652e-05,
+      "loss": 0.5654,
+      "step": 17583
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0477261585488943e-05,
+      "loss": 0.464,
+      "step": 17584
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.047668483432823e-05,
+      "loss": 0.4403,
+      "step": 17585
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0476108083167517e-05,
+      "loss": 0.5025,
+      "step": 17586
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0475531332006808e-05,
+      "loss": 0.6287,
+      "step": 17587
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0474954580846096e-05,
+      "loss": 0.5795,
+      "step": 17588
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0474377829685383e-05,
+      "loss": 0.4767,
+      "step": 17589
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0473801078524673e-05,
+      "loss": 0.4661,
+      "step": 17590
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.047322432736396e-05,
+      "loss": 0.4493,
+      "step": 17591
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0472647576203248e-05,
+      "loss": 0.4722,
+      "step": 17592
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0472070825042538e-05,
+      "loss": 0.397,
+      "step": 17593
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0471494073881825e-05,
+      "loss": 0.3618,
+      "step": 17594
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0470917322721113e-05,
+      "loss": 0.5557,
+      "step": 17595
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.04703405715604e-05,
+      "loss": 0.4487,
+      "step": 17596
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.046976382039969e-05,
+      "loss": 0.4817,
+      "step": 17597
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0469187069238978e-05,
+      "loss": 0.4139,
+      "step": 17598
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0468610318078265e-05,
+      "loss": 0.5383,
+      "step": 17599
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0468033566917555e-05,
+      "loss": 0.5233,
+      "step": 17600
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0467456815756842e-05,
+      "loss": 0.404,
+      "step": 17601
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.046688006459613e-05,
+      "loss": 0.4119,
+      "step": 17602
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.046630331343542e-05,
+      "loss": 0.4439,
+      "step": 17603
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0465726562274707e-05,
+      "loss": 0.4357,
+      "step": 17604
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0465149811113995e-05,
+      "loss": 0.487,
+      "step": 17605
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.0464573059953284e-05,
+      "loss": 0.4243,
+      "step": 17606
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0463996308792572e-05,
+      "loss": 0.4894,
+      "step": 17607
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.046341955763186e-05,
+      "loss": 0.4719,
+      "step": 17608
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0462842806471149e-05,
+      "loss": 0.4439,
+      "step": 17609
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0462266055310437e-05,
+      "loss": 0.4632,
+      "step": 17610
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0461689304149724e-05,
+      "loss": 0.4026,
+      "step": 17611
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0461112552989014e-05,
+      "loss": 0.5318,
+      "step": 17612
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0460535801828301e-05,
+      "loss": 0.4796,
+      "step": 17613
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0459959050667589e-05,
+      "loss": 0.4626,
+      "step": 17614
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045938229950688e-05,
+      "loss": 0.495,
+      "step": 17615
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0458805548346166e-05,
+      "loss": 0.3742,
+      "step": 17616
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0458228797185454e-05,
+      "loss": 0.4449,
+      "step": 17617
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0457652046024745e-05,
+      "loss": 0.5077,
+      "step": 17618
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0457075294864033e-05,
+      "loss": 0.494,
+      "step": 17619
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045649854370332e-05,
+      "loss": 0.4415,
+      "step": 17620
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045592179254261e-05,
+      "loss": 0.3459,
+      "step": 17621
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0455345041381897e-05,
+      "loss": 0.5568,
+      "step": 17622
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0454768290221185e-05,
+      "loss": 0.3466,
+      "step": 17623
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0454191539060474e-05,
+      "loss": 0.4199,
+      "step": 17624
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0453614787899762e-05,
+      "loss": 0.5418,
+      "step": 17625
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045303803673905e-05,
+      "loss": 0.3968,
+      "step": 17626
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.045246128557834e-05,
+      "loss": 0.4479,
+      "step": 17627
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0451884534417627e-05,
+      "loss": 0.4434,
+      "step": 17628
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0451307783256915e-05,
+      "loss": 0.5368,
+      "step": 17629
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0450731032096202e-05,
+      "loss": 0.3654,
+      "step": 17630
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0450154280935492e-05,
+      "loss": 0.4441,
+      "step": 17631
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.044957752977478e-05,
+      "loss": 0.3566,
+      "step": 17632
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0449000778614067e-05,
+      "loss": 0.424,
+      "step": 17633
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0448424027453356e-05,
+      "loss": 0.4052,
+      "step": 17634
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0447847276292644e-05,
+      "loss": 0.3785,
+      "step": 17635
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0447270525131932e-05,
+      "loss": 0.4273,
+      "step": 17636
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0446693773971221e-05,
+      "loss": 0.4638,
+      "step": 17637
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0446117022810509e-05,
+      "loss": 0.5936,
+      "step": 17638
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0445540271649797e-05,
+      "loss": 0.4449,
+      "step": 17639
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0444963520489086e-05,
+      "loss": 0.4504,
+      "step": 17640
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0444386769328374e-05,
+      "loss": 0.5686,
+      "step": 17641
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0443810018167661e-05,
+      "loss": 0.3937,
+      "step": 17642
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.044323326700695e-05,
+      "loss": 0.696,
+      "step": 17643
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0442656515846238e-05,
+      "loss": 0.4864,
+      "step": 17644
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0442079764685526e-05,
+      "loss": 0.4398,
+      "step": 17645
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0441503013524817e-05,
+      "loss": 0.472,
+      "step": 17646
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0440926262364103e-05,
+      "loss": 0.4488,
+      "step": 17647
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.044034951120339e-05,
+      "loss": 0.4043,
+      "step": 17648
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0439772760042682e-05,
+      "loss": 0.4729,
+      "step": 17649
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.043919600888197e-05,
+      "loss": 0.5968,
+      "step": 17650
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0438619257721257e-05,
+      "loss": 0.4113,
+      "step": 17651
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0438042506560547e-05,
+      "loss": 0.5988,
+      "step": 17652
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0437465755399834e-05,
+      "loss": 0.5259,
+      "step": 17653
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0436889004239122e-05,
+      "loss": 0.5741,
+      "step": 17654
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0436312253078411e-05,
+      "loss": 0.4936,
+      "step": 17655
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0435735501917699e-05,
+      "loss": 0.4317,
+      "step": 17656
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0435158750756987e-05,
+      "loss": 0.4976,
+      "step": 17657
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0434581999596276e-05,
+      "loss": 0.4456,
+      "step": 17658
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0434005248435564e-05,
+      "loss": 0.5464,
+      "step": 17659
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0433428497274851e-05,
+      "loss": 0.4524,
+      "step": 17660
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.043285174611414e-05,
+      "loss": 0.3992,
+      "step": 17661
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0432274994953429e-05,
+      "loss": 0.478,
+      "step": 17662
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0431698243792716e-05,
+      "loss": 0.6889,
+      "step": 17663
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0431121492632004e-05,
+      "loss": 0.6247,
+      "step": 17664
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0430544741471293e-05,
+      "loss": 0.4218,
+      "step": 17665
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0429967990310581e-05,
+      "loss": 0.416,
+      "step": 17666
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0429391239149869e-05,
+      "loss": 0.4262,
+      "step": 17667
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0428814487989158e-05,
+      "loss": 0.2877,
+      "step": 17668
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0428237736828446e-05,
+      "loss": 0.3881,
+      "step": 17669
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0427660985667733e-05,
+      "loss": 0.5208,
+      "step": 17670
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0427084234507023e-05,
+      "loss": 0.4077,
+      "step": 17671
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.042650748334631e-05,
+      "loss": 0.4699,
+      "step": 17672
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0425930732185598e-05,
+      "loss": 0.5294,
+      "step": 17673
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0425353981024888e-05,
+      "loss": 0.5058,
+      "step": 17674
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0424777229864175e-05,
+      "loss": 0.5098,
+      "step": 17675
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0424200478703463e-05,
+      "loss": 0.4894,
+      "step": 17676
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0423623727542754e-05,
+      "loss": 0.4917,
+      "step": 17677
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.042304697638204e-05,
+      "loss": 0.4959,
+      "step": 17678
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0422470225221328e-05,
+      "loss": 0.4732,
+      "step": 17679
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0421893474060619e-05,
+      "loss": 0.4616,
+      "step": 17680
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0421316722899906e-05,
+      "loss": 0.3846,
+      "step": 17681
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0420739971739194e-05,
+      "loss": 0.606,
+      "step": 17682
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0420163220578483e-05,
+      "loss": 0.3536,
+      "step": 17683
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0419586469417771e-05,
+      "loss": 0.4493,
+      "step": 17684
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0419009718257059e-05,
+      "loss": 0.4609,
+      "step": 17685
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0418432967096348e-05,
+      "loss": 0.4582,
+      "step": 17686
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0417856215935636e-05,
+      "loss": 0.5412,
+      "step": 17687
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0417279464774924e-05,
+      "loss": 0.5246,
+      "step": 17688
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0416702713614213e-05,
+      "loss": 0.5231,
+      "step": 17689
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.04161259624535e-05,
+      "loss": 0.5158,
+      "step": 17690
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0415549211292788e-05,
+      "loss": 0.4899,
+      "step": 17691
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0414972460132078e-05,
+      "loss": 0.593,
+      "step": 17692
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0414395708971365e-05,
+      "loss": 0.5702,
+      "step": 17693
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0413818957810653e-05,
+      "loss": 0.361,
+      "step": 17694
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0413242206649942e-05,
+      "loss": 0.4293,
+      "step": 17695
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.041266545548923e-05,
+      "loss": 0.6807,
+      "step": 17696
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0412088704328518e-05,
+      "loss": 0.5731,
+      "step": 17697
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0411511953167806e-05,
+      "loss": 0.5477,
+      "step": 17698
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0410935202007095e-05,
+      "loss": 0.3982,
+      "step": 17699
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0410358450846383e-05,
+      "loss": 0.4866,
+      "step": 17700
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.040978169968567e-05,
+      "loss": 0.4208,
+      "step": 17701
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.040920494852496e-05,
+      "loss": 0.4333,
+      "step": 17702
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0408628197364247e-05,
+      "loss": 0.4766,
+      "step": 17703
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0408051446203535e-05,
+      "loss": 0.5687,
+      "step": 17704
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0407474695042824e-05,
+      "loss": 0.3816,
+      "step": 17705
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0406897943882112e-05,
+      "loss": 0.5126,
+      "step": 17706
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.04063211927214e-05,
+      "loss": 0.5088,
+      "step": 17707
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.040574444156069e-05,
+      "loss": 0.5264,
+      "step": 17708
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0405167690399977e-05,
+      "loss": 0.4246,
+      "step": 17709
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0404590939239265e-05,
+      "loss": 0.4061,
+      "step": 17710
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0404014188078556e-05,
+      "loss": 0.4135,
+      "step": 17711
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0403437436917843e-05,
+      "loss": 0.5208,
+      "step": 17712
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0402860685757131e-05,
+      "loss": 0.5254,
+      "step": 17713
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.040228393459642e-05,
+      "loss": 0.4937,
+      "step": 17714
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0401707183435708e-05,
+      "loss": 0.5793,
+      "step": 17715
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0401130432274996e-05,
+      "loss": 0.4991,
+      "step": 17716
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0400553681114285e-05,
+      "loss": 0.5111,
+      "step": 17717
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0399976929953573e-05,
+      "loss": 0.5452,
+      "step": 17718
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.039940017879286e-05,
+      "loss": 0.5473,
+      "step": 17719
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.039882342763215e-05,
+      "loss": 0.5258,
+      "step": 17720
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0398246676471438e-05,
+      "loss": 0.4309,
+      "step": 17721
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0397669925310725e-05,
+      "loss": 0.3949,
+      "step": 17722
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0397093174150015e-05,
+      "loss": 0.4377,
+      "step": 17723
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0396516422989302e-05,
+      "loss": 0.5014,
+      "step": 17724
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.039593967182859e-05,
+      "loss": 0.3264,
+      "step": 17725
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.039536292066788e-05,
+      "loss": 0.3782,
+      "step": 17726
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0394786169507167e-05,
+      "loss": 0.4626,
+      "step": 17727
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0394209418346455e-05,
+      "loss": 0.5245,
+      "step": 17728
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0393632667185744e-05,
+      "loss": 0.4104,
+      "step": 17729
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0393055916025032e-05,
+      "loss": 0.5663,
+      "step": 17730
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.039247916486432e-05,
+      "loss": 0.4786,
+      "step": 17731
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0391902413703609e-05,
+      "loss": 0.4105,
+      "step": 17732
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0391325662542897e-05,
+      "loss": 0.5359,
+      "step": 17733
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0390748911382184e-05,
+      "loss": 0.639,
+      "step": 17734
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0390172160221472e-05,
+      "loss": 0.4475,
+      "step": 17735
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0389595409060761e-05,
+      "loss": 0.4951,
+      "step": 17736
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0389018657900049e-05,
+      "loss": 0.4616,
+      "step": 17737
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0388441906739337e-05,
+      "loss": 0.5121,
+      "step": 17738
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0387865155578628e-05,
+      "loss": 0.5038,
+      "step": 17739
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0387288404417914e-05,
+      "loss": 0.4776,
+      "step": 17740
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0386711653257201e-05,
+      "loss": 0.388,
+      "step": 17741
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0386134902096492e-05,
+      "loss": 0.458,
+      "step": 17742
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.038555815093578e-05,
+      "loss": 0.4572,
+      "step": 17743
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0384981399775068e-05,
+      "loss": 0.3713,
+      "step": 17744
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0384404648614357e-05,
+      "loss": 0.5351,
+      "step": 17745
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0383827897453645e-05,
+      "loss": 0.4964,
+      "step": 17746
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0383251146292933e-05,
+      "loss": 0.4519,
+      "step": 17747
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0382674395132222e-05,
+      "loss": 0.5005,
+      "step": 17748
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.038209764397151e-05,
+      "loss": 0.5814,
+      "step": 17749
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0381520892810797e-05,
+      "loss": 0.5832,
+      "step": 17750
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0380944141650087e-05,
+      "loss": 0.4811,
+      "step": 17751
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0380367390489374e-05,
+      "loss": 0.4251,
+      "step": 17752
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0379790639328662e-05,
+      "loss": 0.4441,
+      "step": 17753
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0379213888167951e-05,
+      "loss": 0.5166,
+      "step": 17754
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0378637137007239e-05,
+      "loss": 0.3615,
+      "step": 17755
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0378060385846527e-05,
+      "loss": 0.3914,
+      "step": 17756
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0377483634685816e-05,
+      "loss": 0.3692,
+      "step": 17757
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0376906883525104e-05,
+      "loss": 0.3689,
+      "step": 17758
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0376330132364392e-05,
+      "loss": 0.396,
+      "step": 17759
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0375753381203681e-05,
+      "loss": 0.4383,
+      "step": 17760
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0375176630042969e-05,
+      "loss": 0.3871,
+      "step": 17761
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0374599878882256e-05,
+      "loss": 0.4606,
+      "step": 17762
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0374023127721546e-05,
+      "loss": 0.5009,
+      "step": 17763
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0373446376560833e-05,
+      "loss": 0.4874,
+      "step": 17764
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0372869625400121e-05,
+      "loss": 0.4089,
+      "step": 17765
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0372292874239412e-05,
+      "loss": 0.5915,
+      "step": 17766
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0371716123078698e-05,
+      "loss": 0.3712,
+      "step": 17767
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0371139371917986e-05,
+      "loss": 0.5084,
+      "step": 17768
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0370562620757274e-05,
+      "loss": 0.4961,
+      "step": 17769
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0369985869596565e-05,
+      "loss": 0.4824,
+      "step": 17770
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.036940911843585e-05,
+      "loss": 0.5993,
+      "step": 17771
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0368832367275138e-05,
+      "loss": 0.412,
+      "step": 17772
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.036825561611443e-05,
+      "loss": 0.4358,
+      "step": 17773
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0367678864953717e-05,
+      "loss": 0.4742,
+      "step": 17774
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0367102113793005e-05,
+      "loss": 0.6497,
+      "step": 17775
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0366525362632294e-05,
+      "loss": 0.4064,
+      "step": 17776
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0365948611471582e-05,
+      "loss": 0.3584,
+      "step": 17777
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.036537186031087e-05,
+      "loss": 0.4442,
+      "step": 17778
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0364795109150159e-05,
+      "loss": 0.5808,
+      "step": 17779
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0364218357989447e-05,
+      "loss": 0.4658,
+      "step": 17780
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0363641606828734e-05,
+      "loss": 0.4924,
+      "step": 17781
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0363064855668024e-05,
+      "loss": 0.4646,
+      "step": 17782
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0362488104507311e-05,
+      "loss": 0.4807,
+      "step": 17783
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0361911353346599e-05,
+      "loss": 0.5159,
+      "step": 17784
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 1.0361334602185888e-05,
+      "loss": 0.4791,
+      "step": 17785
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0360757851025176e-05,
+      "loss": 0.4204,
+      "step": 17786
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0360181099864464e-05,
+      "loss": 0.4334,
+      "step": 17787
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0359604348703753e-05,
+      "loss": 0.5794,
+      "step": 17788
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.035902759754304e-05,
+      "loss": 0.4952,
+      "step": 17789
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0358450846382328e-05,
+      "loss": 0.4999,
+      "step": 17790
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0357874095221618e-05,
+      "loss": 0.4682,
+      "step": 17791
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0357297344060906e-05,
+      "loss": 0.4357,
+      "step": 17792
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0356720592900193e-05,
+      "loss": 0.5149,
+      "step": 17793
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0356143841739483e-05,
+      "loss": 0.3899,
+      "step": 17794
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.035556709057877e-05,
+      "loss": 0.4062,
+      "step": 17795
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0354990339418058e-05,
+      "loss": 0.4952,
+      "step": 17796
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0354413588257349e-05,
+      "loss": 0.4874,
+      "step": 17797
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0353836837096635e-05,
+      "loss": 0.4186,
+      "step": 17798
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0353260085935923e-05,
+      "loss": 0.5431,
+      "step": 17799
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0352683334775214e-05,
+      "loss": 0.4716,
+      "step": 17800
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0352106583614501e-05,
+      "loss": 0.4377,
+      "step": 17801
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0351529832453787e-05,
+      "loss": 0.5091,
+      "step": 17802
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0350953081293075e-05,
+      "loss": 0.4895,
+      "step": 17803
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0350376330132366e-05,
+      "loss": 0.5288,
+      "step": 17804
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0349799578971654e-05,
+      "loss": 0.57,
+      "step": 17805
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0349222827810942e-05,
+      "loss": 0.5011,
+      "step": 17806
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0348646076650231e-05,
+      "loss": 0.4616,
+      "step": 17807
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0348069325489519e-05,
+      "loss": 0.4622,
+      "step": 17808
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0347492574328806e-05,
+      "loss": 0.4627,
+      "step": 17809
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0346915823168096e-05,
+      "loss": 0.4718,
+      "step": 17810
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0346339072007383e-05,
+      "loss": 0.5503,
+      "step": 17811
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0345762320846671e-05,
+      "loss": 0.5584,
+      "step": 17812
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.034518556968596e-05,
+      "loss": 0.3654,
+      "step": 17813
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0344608818525248e-05,
+      "loss": 0.5382,
+      "step": 17814
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0344032067364536e-05,
+      "loss": 0.4587,
+      "step": 17815
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0343455316203825e-05,
+      "loss": 0.4454,
+      "step": 17816
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0342878565043113e-05,
+      "loss": 0.5305,
+      "step": 17817
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.03423018138824e-05,
+      "loss": 0.4637,
+      "step": 17818
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.034172506272169e-05,
+      "loss": 0.42,
+      "step": 17819
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0341148311560978e-05,
+      "loss": 0.4502,
+      "step": 17820
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0340571560400265e-05,
+      "loss": 0.5305,
+      "step": 17821
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0339994809239555e-05,
+      "loss": 0.5699,
+      "step": 17822
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0339418058078842e-05,
+      "loss": 0.6164,
+      "step": 17823
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033884130691813e-05,
+      "loss": 0.4475,
+      "step": 17824
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033826455575742e-05,
+      "loss": 0.4196,
+      "step": 17825
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0337687804596707e-05,
+      "loss": 0.5138,
+      "step": 17826
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0337111053435995e-05,
+      "loss": 0.4845,
+      "step": 17827
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0336534302275286e-05,
+      "loss": 0.4154,
+      "step": 17828
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0335957551114572e-05,
+      "loss": 0.4423,
+      "step": 17829
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033538079995386e-05,
+      "loss": 0.3788,
+      "step": 17830
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033480404879315e-05,
+      "loss": 0.4613,
+      "step": 17831
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0334227297632438e-05,
+      "loss": 0.4368,
+      "step": 17832
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0333650546471726e-05,
+      "loss": 0.5068,
+      "step": 17833
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0333073795311015e-05,
+      "loss": 0.4839,
+      "step": 17834
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0332497044150303e-05,
+      "loss": 0.5063,
+      "step": 17835
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033192029298959e-05,
+      "loss": 0.3841,
+      "step": 17836
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.033134354182888e-05,
+      "loss": 0.5474,
+      "step": 17837
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0330766790668168e-05,
+      "loss": 0.5078,
+      "step": 17838
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0330190039507456e-05,
+      "loss": 0.6095,
+      "step": 17839
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0329613288346743e-05,
+      "loss": 0.4681,
+      "step": 17840
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0329036537186033e-05,
+      "loss": 0.572,
+      "step": 17841
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.032845978602532e-05,
+      "loss": 0.3727,
+      "step": 17842
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0327883034864608e-05,
+      "loss": 0.3736,
+      "step": 17843
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0327306283703897e-05,
+      "loss": 0.3893,
+      "step": 17844
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0326729532543185e-05,
+      "loss": 0.4777,
+      "step": 17845
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0326152781382473e-05,
+      "loss": 0.3474,
+      "step": 17846
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0325576030221762e-05,
+      "loss": 0.5392,
+      "step": 17847
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.032499927906105e-05,
+      "loss": 0.428,
+      "step": 17848
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0324422527900337e-05,
+      "loss": 0.4185,
+      "step": 17849
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0323845776739627e-05,
+      "loss": 0.3869,
+      "step": 17850
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0323269025578915e-05,
+      "loss": 0.4285,
+      "step": 17851
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0322692274418202e-05,
+      "loss": 0.4527,
+      "step": 17852
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0322115523257492e-05,
+      "loss": 0.4264,
+      "step": 17853
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.032153877209678e-05,
+      "loss": 0.4579,
+      "step": 17854
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0320962020936067e-05,
+      "loss": 0.3243,
+      "step": 17855
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0320385269775356e-05,
+      "loss": 0.4499,
+      "step": 17856
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0319808518614644e-05,
+      "loss": 0.543,
+      "step": 17857
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0319231767453932e-05,
+      "loss": 0.4511,
+      "step": 17858
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0318655016293223e-05,
+      "loss": 0.4052,
+      "step": 17859
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0318078265132509e-05,
+      "loss": 0.4787,
+      "step": 17860
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0317501513971796e-05,
+      "loss": 0.6304,
+      "step": 17861
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0316924762811088e-05,
+      "loss": 0.4664,
+      "step": 17862
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0316348011650375e-05,
+      "loss": 0.4849,
+      "step": 17863
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0315771260489663e-05,
+      "loss": 0.5217,
+      "step": 17864
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0315194509328952e-05,
+      "loss": 0.5428,
+      "step": 17865
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.031461775816824e-05,
+      "loss": 0.5629,
+      "step": 17866
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0314041007007528e-05,
+      "loss": 0.4107,
+      "step": 17867
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0313464255846817e-05,
+      "loss": 0.4134,
+      "step": 17868
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0312887504686105e-05,
+      "loss": 0.6209,
+      "step": 17869
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0312310753525392e-05,
+      "loss": 0.5012,
+      "step": 17870
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0311734002364682e-05,
+      "loss": 0.438,
+      "step": 17871
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.031115725120397e-05,
+      "loss": 0.4985,
+      "step": 17872
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0310580500043257e-05,
+      "loss": 0.4306,
+      "step": 17873
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0310003748882545e-05,
+      "loss": 0.4339,
+      "step": 17874
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0309426997721834e-05,
+      "loss": 0.4883,
+      "step": 17875
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0308850246561122e-05,
+      "loss": 0.3585,
+      "step": 17876
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.030827349540041e-05,
+      "loss": 0.5991,
+      "step": 17877
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0307696744239699e-05,
+      "loss": 0.3988,
+      "step": 17878
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0307119993078987e-05,
+      "loss": 0.4867,
+      "step": 17879
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0306543241918274e-05,
+      "loss": 0.3858,
+      "step": 17880
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0305966490757564e-05,
+      "loss": 0.433,
+      "step": 17881
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0305389739596851e-05,
+      "loss": 0.4481,
+      "step": 17882
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0304812988436139e-05,
+      "loss": 0.3958,
+      "step": 17883
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0304236237275428e-05,
+      "loss": 0.4532,
+      "step": 17884
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0303659486114716e-05,
+      "loss": 0.3713,
+      "step": 17885
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0303082734954004e-05,
+      "loss": 0.4133,
+      "step": 17886
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0302505983793293e-05,
+      "loss": 0.5882,
+      "step": 17887
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0301929232632581e-05,
+      "loss": 0.3107,
+      "step": 17888
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0301352481471869e-05,
+      "loss": 0.4389,
+      "step": 17889
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.030077573031116e-05,
+      "loss": 0.6327,
+      "step": 17890
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0300198979150446e-05,
+      "loss": 0.4507,
+      "step": 17891
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0299622227989733e-05,
+      "loss": 0.6295,
+      "step": 17892
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0299045476829024e-05,
+      "loss": 0.4683,
+      "step": 17893
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0298468725668312e-05,
+      "loss": 0.4764,
+      "step": 17894
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.02978919745076e-05,
+      "loss": 0.4451,
+      "step": 17895
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0297315223346889e-05,
+      "loss": 0.3664,
+      "step": 17896
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0296738472186177e-05,
+      "loss": 0.4313,
+      "step": 17897
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0296161721025465e-05,
+      "loss": 0.5568,
+      "step": 17898
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0295584969864754e-05,
+      "loss": 0.4489,
+      "step": 17899
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0295008218704042e-05,
+      "loss": 0.4241,
+      "step": 17900
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.029443146754333e-05,
+      "loss": 0.5012,
+      "step": 17901
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0293854716382619e-05,
+      "loss": 0.3212,
+      "step": 17902
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0293277965221906e-05,
+      "loss": 0.4612,
+      "step": 17903
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0292701214061194e-05,
+      "loss": 0.4338,
+      "step": 17904
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0292124462900483e-05,
+      "loss": 0.3906,
+      "step": 17905
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0291547711739771e-05,
+      "loss": 0.3775,
+      "step": 17906
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0290970960579059e-05,
+      "loss": 0.6221,
+      "step": 17907
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0290394209418346e-05,
+      "loss": 0.4063,
+      "step": 17908
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0289817458257636e-05,
+      "loss": 0.3758,
+      "step": 17909
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0289240707096924e-05,
+      "loss": 0.4449,
+      "step": 17910
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0288663955936211e-05,
+      "loss": 0.4211,
+      "step": 17911
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.02880872047755e-05,
+      "loss": 0.3228,
+      "step": 17912
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0287510453614788e-05,
+      "loss": 0.4854,
+      "step": 17913
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0286933702454076e-05,
+      "loss": 0.4512,
+      "step": 17914
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0286356951293365e-05,
+      "loss": 0.4804,
+      "step": 17915
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0285780200132653e-05,
+      "loss": 0.4373,
+      "step": 17916
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.028520344897194e-05,
+      "loss": 0.4684,
+      "step": 17917
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.028462669781123e-05,
+      "loss": 0.4675,
+      "step": 17918
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0284049946650518e-05,
+      "loss": 0.5045,
+      "step": 17919
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0283473195489805e-05,
+      "loss": 0.5192,
+      "step": 17920
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0282896444329097e-05,
+      "loss": 0.5036,
+      "step": 17921
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0282319693168382e-05,
+      "loss": 0.491,
+      "step": 17922
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.028174294200767e-05,
+      "loss": 0.3788,
+      "step": 17923
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0281166190846961e-05,
+      "loss": 0.4346,
+      "step": 17924
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0280589439686249e-05,
+      "loss": 0.3777,
+      "step": 17925
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0280012688525537e-05,
+      "loss": 0.4676,
+      "step": 17926
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0279435937364826e-05,
+      "loss": 0.4032,
+      "step": 17927
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0278859186204114e-05,
+      "loss": 0.3246,
+      "step": 17928
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0278282435043401e-05,
+      "loss": 0.5282,
+      "step": 17929
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.027770568388269e-05,
+      "loss": 0.3882,
+      "step": 17930
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0277128932721978e-05,
+      "loss": 0.3538,
+      "step": 17931
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0276552181561266e-05,
+      "loss": 0.4388,
+      "step": 17932
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0275975430400556e-05,
+      "loss": 0.5223,
+      "step": 17933
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0275398679239843e-05,
+      "loss": 0.4284,
+      "step": 17934
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0274821928079131e-05,
+      "loss": 0.537,
+      "step": 17935
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.027424517691842e-05,
+      "loss": 0.4145,
+      "step": 17936
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0273668425757708e-05,
+      "loss": 0.3348,
+      "step": 17937
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0273091674596996e-05,
+      "loss": 0.4879,
+      "step": 17938
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0272514923436285e-05,
+      "loss": 0.4899,
+      "step": 17939
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0271938172275573e-05,
+      "loss": 0.4794,
+      "step": 17940
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.027136142111486e-05,
+      "loss": 0.4573,
+      "step": 17941
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0270784669954148e-05,
+      "loss": 0.432,
+      "step": 17942
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0270207918793437e-05,
+      "loss": 0.4802,
+      "step": 17943
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0269631167632725e-05,
+      "loss": 0.4507,
+      "step": 17944
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0269054416472013e-05,
+      "loss": 0.4855,
+      "step": 17945
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0268477665311302e-05,
+      "loss": 0.4611,
+      "step": 17946
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.026790091415059e-05,
+      "loss": 0.4475,
+      "step": 17947
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0267324162989878e-05,
+      "loss": 0.3622,
+      "step": 17948
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0266747411829167e-05,
+      "loss": 0.5897,
+      "step": 17949
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0266170660668455e-05,
+      "loss": 0.397,
+      "step": 17950
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0265593909507742e-05,
+      "loss": 0.3627,
+      "step": 17951
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0265017158347033e-05,
+      "loss": 0.3084,
+      "step": 17952
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.026444040718632e-05,
+      "loss": 0.486,
+      "step": 17953
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0263863656025607e-05,
+      "loss": 0.5593,
+      "step": 17954
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0263286904864898e-05,
+      "loss": 0.3994,
+      "step": 17955
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0262710153704186e-05,
+      "loss": 0.589,
+      "step": 17956
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0262133402543473e-05,
+      "loss": 0.5657,
+      "step": 17957
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0261556651382763e-05,
+      "loss": 0.3612,
+      "step": 17958
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.026097990022205e-05,
+      "loss": 0.6366,
+      "step": 17959
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0260403149061338e-05,
+      "loss": 0.401,
+      "step": 17960
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0259826397900628e-05,
+      "loss": 0.3957,
+      "step": 17961
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0259249646739915e-05,
+      "loss": 0.3837,
+      "step": 17962
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0258672895579203e-05,
+      "loss": 0.4389,
+      "step": 17963
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 1.0258096144418492e-05,
+      "loss": 0.4334,
+      "step": 17964
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.025751939325778e-05,
+      "loss": 0.5133,
+      "step": 17965
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0256942642097068e-05,
+      "loss": 0.3382,
+      "step": 17966
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0256365890936357e-05,
+      "loss": 0.4459,
+      "step": 17967
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0255789139775645e-05,
+      "loss": 0.4585,
+      "step": 17968
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0255212388614932e-05,
+      "loss": 0.4465,
+      "step": 17969
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0254635637454222e-05,
+      "loss": 0.422,
+      "step": 17970
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.025405888629351e-05,
+      "loss": 0.5804,
+      "step": 17971
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0253482135132797e-05,
+      "loss": 0.4332,
+      "step": 17972
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0252905383972087e-05,
+      "loss": 0.4757,
+      "step": 17973
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0252328632811374e-05,
+      "loss": 0.3642,
+      "step": 17974
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0251751881650662e-05,
+      "loss": 0.4526,
+      "step": 17975
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0251175130489951e-05,
+      "loss": 0.5437,
+      "step": 17976
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0250598379329239e-05,
+      "loss": 0.4594,
+      "step": 17977
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0250021628168527e-05,
+      "loss": 0.4741,
+      "step": 17978
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0249444877007814e-05,
+      "loss": 0.4327,
+      "step": 17979
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0248868125847104e-05,
+      "loss": 0.4404,
+      "step": 17980
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0248291374686391e-05,
+      "loss": 0.4965,
+      "step": 17981
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.024771462352568e-05,
+      "loss": 0.5482,
+      "step": 17982
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.024713787236497e-05,
+      "loss": 0.3621,
+      "step": 17983
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0246561121204256e-05,
+      "loss": 0.3873,
+      "step": 17984
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0245984370043544e-05,
+      "loss": 0.4961,
+      "step": 17985
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0245407618882835e-05,
+      "loss": 0.3305,
+      "step": 17986
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0244830867722123e-05,
+      "loss": 0.4233,
+      "step": 17987
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.024425411656141e-05,
+      "loss": 0.4285,
+      "step": 17988
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.02436773654007e-05,
+      "loss": 0.4853,
+      "step": 17989
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0243100614239987e-05,
+      "loss": 0.3531,
+      "step": 17990
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0242523863079275e-05,
+      "loss": 0.3276,
+      "step": 17991
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0241947111918565e-05,
+      "loss": 0.4549,
+      "step": 17992
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0241370360757852e-05,
+      "loss": 0.5301,
+      "step": 17993
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.024079360959714e-05,
+      "loss": 0.4071,
+      "step": 17994
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.024021685843643e-05,
+      "loss": 0.4622,
+      "step": 17995
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0239640107275717e-05,
+      "loss": 0.4761,
+      "step": 17996
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0239063356115005e-05,
+      "loss": 0.4641,
+      "step": 17997
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0238486604954294e-05,
+      "loss": 0.5046,
+      "step": 17998
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0237909853793582e-05,
+      "loss": 0.5802,
+      "step": 17999
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.023733310263287e-05,
+      "loss": 0.4225,
+      "step": 18000
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0236756351472159e-05,
+      "loss": 0.4519,
+      "step": 18001
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0236179600311446e-05,
+      "loss": 0.5233,
+      "step": 18002
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0235602849150734e-05,
+      "loss": 0.5407,
+      "step": 18003
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0235026097990023e-05,
+      "loss": 0.3932,
+      "step": 18004
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0234449346829311e-05,
+      "loss": 0.391,
+      "step": 18005
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0233872595668599e-05,
+      "loss": 0.3595,
+      "step": 18006
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0233295844507888e-05,
+      "loss": 0.4812,
+      "step": 18007
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0232719093347176e-05,
+      "loss": 0.4428,
+      "step": 18008
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0232142342186464e-05,
+      "loss": 0.5247,
+      "step": 18009
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0231565591025755e-05,
+      "loss": 0.4696,
+      "step": 18010
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.023098883986504e-05,
+      "loss": 0.4618,
+      "step": 18011
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0230412088704328e-05,
+      "loss": 0.4543,
+      "step": 18012
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0229835337543616e-05,
+      "loss": 0.3848,
+      "step": 18013
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0229258586382907e-05,
+      "loss": 0.4962,
+      "step": 18014
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0228681835222193e-05,
+      "loss": 0.507,
+      "step": 18015
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.022810508406148e-05,
+      "loss": 0.4675,
+      "step": 18016
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0227528332900772e-05,
+      "loss": 0.4884,
+      "step": 18017
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.022695158174006e-05,
+      "loss": 0.4481,
+      "step": 18018
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0226374830579347e-05,
+      "loss": 0.5322,
+      "step": 18019
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0225798079418637e-05,
+      "loss": 0.34,
+      "step": 18020
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0225221328257924e-05,
+      "loss": 0.4549,
+      "step": 18021
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0224644577097212e-05,
+      "loss": 0.5114,
+      "step": 18022
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0224067825936501e-05,
+      "loss": 0.4703,
+      "step": 18023
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0223491074775789e-05,
+      "loss": 0.4034,
+      "step": 18024
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0222914323615077e-05,
+      "loss": 0.4521,
+      "step": 18025
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0222337572454366e-05,
+      "loss": 0.5097,
+      "step": 18026
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0221760821293654e-05,
+      "loss": 0.6151,
+      "step": 18027
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0221184070132941e-05,
+      "loss": 0.4483,
+      "step": 18028
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0220607318972231e-05,
+      "loss": 0.4795,
+      "step": 18029
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0220030567811519e-05,
+      "loss": 0.5246,
+      "step": 18030
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0219453816650806e-05,
+      "loss": 0.3885,
+      "step": 18031
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0218877065490096e-05,
+      "loss": 0.4858,
+      "step": 18032
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0218300314329383e-05,
+      "loss": 0.5054,
+      "step": 18033
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0217723563168671e-05,
+      "loss": 0.4176,
+      "step": 18034
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.021714681200796e-05,
+      "loss": 0.4888,
+      "step": 18035
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0216570060847248e-05,
+      "loss": 0.3892,
+      "step": 18036
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0215993309686536e-05,
+      "loss": 0.4214,
+      "step": 18037
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0215416558525825e-05,
+      "loss": 0.4187,
+      "step": 18038
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0214839807365113e-05,
+      "loss": 0.3455,
+      "step": 18039
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.02142630562044e-05,
+      "loss": 0.3917,
+      "step": 18040
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0213686305043692e-05,
+      "loss": 0.4989,
+      "step": 18041
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0213109553882978e-05,
+      "loss": 0.4006,
+      "step": 18042
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0212532802722265e-05,
+      "loss": 0.5659,
+      "step": 18043
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0211956051561556e-05,
+      "loss": 0.3923,
+      "step": 18044
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0211379300400844e-05,
+      "loss": 0.4479,
+      "step": 18045
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.021080254924013e-05,
+      "loss": 0.3389,
+      "step": 18046
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0210225798079418e-05,
+      "loss": 0.3667,
+      "step": 18047
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0209649046918709e-05,
+      "loss": 0.4291,
+      "step": 18048
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0209072295757996e-05,
+      "loss": 0.4855,
+      "step": 18049
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0208495544597284e-05,
+      "loss": 0.4521,
+      "step": 18050
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0207918793436573e-05,
+      "loss": 0.4589,
+      "step": 18051
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0207342042275861e-05,
+      "loss": 0.6232,
+      "step": 18052
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0206765291115149e-05,
+      "loss": 0.4065,
+      "step": 18053
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0206188539954438e-05,
+      "loss": 0.3465,
+      "step": 18054
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0205611788793726e-05,
+      "loss": 0.3799,
+      "step": 18055
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0205035037633014e-05,
+      "loss": 0.5387,
+      "step": 18056
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0204458286472303e-05,
+      "loss": 0.2624,
+      "step": 18057
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.020388153531159e-05,
+      "loss": 0.4155,
+      "step": 18058
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0203304784150878e-05,
+      "loss": 0.5194,
+      "step": 18059
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0202728032990168e-05,
+      "loss": 0.4612,
+      "step": 18060
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0202151281829455e-05,
+      "loss": 0.4635,
+      "step": 18061
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0201574530668743e-05,
+      "loss": 0.5086,
+      "step": 18062
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0200997779508032e-05,
+      "loss": 0.3941,
+      "step": 18063
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.020042102834732e-05,
+      "loss": 0.407,
+      "step": 18064
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0199844277186608e-05,
+      "loss": 0.4431,
+      "step": 18065
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0199267526025897e-05,
+      "loss": 0.4225,
+      "step": 18066
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0198690774865185e-05,
+      "loss": 0.4865,
+      "step": 18067
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0198114023704473e-05,
+      "loss": 0.3261,
+      "step": 18068
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0197537272543762e-05,
+      "loss": 0.4351,
+      "step": 18069
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.019696052138305e-05,
+      "loss": 0.5469,
+      "step": 18070
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0196383770222337e-05,
+      "loss": 0.3991,
+      "step": 18071
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0195807019061628e-05,
+      "loss": 0.3638,
+      "step": 18072
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0195230267900914e-05,
+      "loss": 0.5022,
+      "step": 18073
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0194653516740202e-05,
+      "loss": 0.4571,
+      "step": 18074
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0194076765579493e-05,
+      "loss": 0.4933,
+      "step": 18075
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0193500014418781e-05,
+      "loss": 0.5721,
+      "step": 18076
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0192923263258067e-05,
+      "loss": 0.4157,
+      "step": 18077
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0192346512097358e-05,
+      "loss": 0.4279,
+      "step": 18078
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0191769760936646e-05,
+      "loss": 0.438,
+      "step": 18079
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0191193009775933e-05,
+      "loss": 0.4564,
+      "step": 18080
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0190616258615221e-05,
+      "loss": 0.5793,
+      "step": 18081
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.019003950745451e-05,
+      "loss": 0.4066,
+      "step": 18082
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0189462756293798e-05,
+      "loss": 0.4113,
+      "step": 18083
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0188886005133086e-05,
+      "loss": 0.3918,
+      "step": 18084
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0188309253972375e-05,
+      "loss": 0.4068,
+      "step": 18085
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0187732502811663e-05,
+      "loss": 0.385,
+      "step": 18086
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.018715575165095e-05,
+      "loss": 0.4028,
+      "step": 18087
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.018657900049024e-05,
+      "loss": 0.4522,
+      "step": 18088
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0186002249329528e-05,
+      "loss": 0.5564,
+      "step": 18089
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0185425498168815e-05,
+      "loss": 0.4212,
+      "step": 18090
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0184848747008105e-05,
+      "loss": 0.5026,
+      "step": 18091
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0184271995847392e-05,
+      "loss": 0.3557,
+      "step": 18092
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.018369524468668e-05,
+      "loss": 0.3999,
+      "step": 18093
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.018311849352597e-05,
+      "loss": 0.4483,
+      "step": 18094
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0182541742365257e-05,
+      "loss": 0.3013,
+      "step": 18095
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0181964991204545e-05,
+      "loss": 0.3973,
+      "step": 18096
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0181388240043834e-05,
+      "loss": 0.4097,
+      "step": 18097
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0180811488883122e-05,
+      "loss": 0.4669,
+      "step": 18098
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.018023473772241e-05,
+      "loss": 0.3869,
+      "step": 18099
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0179657986561699e-05,
+      "loss": 0.5016,
+      "step": 18100
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0179081235400987e-05,
+      "loss": 0.4202,
+      "step": 18101
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0178504484240274e-05,
+      "loss": 0.3615,
+      "step": 18102
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0177927733079565e-05,
+      "loss": 0.4706,
+      "step": 18103
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0177350981918851e-05,
+      "loss": 0.4687,
+      "step": 18104
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0176774230758139e-05,
+      "loss": 0.4753,
+      "step": 18105
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.017619747959743e-05,
+      "loss": 0.6027,
+      "step": 18106
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0175620728436718e-05,
+      "loss": 0.4387,
+      "step": 18107
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0175043977276004e-05,
+      "loss": 0.3859,
+      "step": 18108
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0174467226115295e-05,
+      "loss": 0.4991,
+      "step": 18109
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0173890474954582e-05,
+      "loss": 0.5,
+      "step": 18110
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.017331372379387e-05,
+      "loss": 0.5239,
+      "step": 18111
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.017273697263316e-05,
+      "loss": 0.3644,
+      "step": 18112
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0172160221472447e-05,
+      "loss": 0.4824,
+      "step": 18113
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0171583470311735e-05,
+      "loss": 0.363,
+      "step": 18114
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0171006719151024e-05,
+      "loss": 0.2944,
+      "step": 18115
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0170429967990312e-05,
+      "loss": 0.4218,
+      "step": 18116
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.01698532168296e-05,
+      "loss": 0.4016,
+      "step": 18117
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0169276465668887e-05,
+      "loss": 0.4032,
+      "step": 18118
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0168699714508177e-05,
+      "loss": 0.508,
+      "step": 18119
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0168122963347464e-05,
+      "loss": 0.5709,
+      "step": 18120
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0167546212186752e-05,
+      "loss": 0.303,
+      "step": 18121
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0166969461026041e-05,
+      "loss": 0.393,
+      "step": 18122
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.016639270986533e-05,
+      "loss": 0.3605,
+      "step": 18123
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0165815958704617e-05,
+      "loss": 0.4517,
+      "step": 18124
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0165239207543906e-05,
+      "loss": 0.395,
+      "step": 18125
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0164662456383194e-05,
+      "loss": 0.4577,
+      "step": 18126
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0164085705222482e-05,
+      "loss": 0.4476,
+      "step": 18127
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0163508954061771e-05,
+      "loss": 0.3895,
+      "step": 18128
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0162932202901059e-05,
+      "loss": 0.4198,
+      "step": 18129
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0162355451740346e-05,
+      "loss": 0.5592,
+      "step": 18130
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0161778700579636e-05,
+      "loss": 0.4755,
+      "step": 18131
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0161201949418923e-05,
+      "loss": 0.3642,
+      "step": 18132
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0160625198258211e-05,
+      "loss": 0.4752,
+      "step": 18133
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0160048447097502e-05,
+      "loss": 0.543,
+      "step": 18134
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0159471695936788e-05,
+      "loss": 0.3925,
+      "step": 18135
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0158894944776076e-05,
+      "loss": 0.4677,
+      "step": 18136
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0158318193615367e-05,
+      "loss": 0.4268,
+      "step": 18137
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0157741442454655e-05,
+      "loss": 0.5045,
+      "step": 18138
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.015716469129394e-05,
+      "loss": 0.559,
+      "step": 18139
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0156587940133232e-05,
+      "loss": 0.399,
+      "step": 18140
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.015601118897252e-05,
+      "loss": 0.4632,
+      "step": 18141
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0155434437811807e-05,
+      "loss": 0.3428,
+      "step": 18142
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.0154857686651096e-05,
+      "loss": 0.4629,
+      "step": 18143
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0154280935490384e-05,
+      "loss": 0.5546,
+      "step": 18144
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0153704184329672e-05,
+      "loss": 0.3395,
+      "step": 18145
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0153127433168961e-05,
+      "loss": 0.332,
+      "step": 18146
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0152550682008249e-05,
+      "loss": 0.5294,
+      "step": 18147
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0151973930847537e-05,
+      "loss": 0.4202,
+      "step": 18148
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0151397179686826e-05,
+      "loss": 0.521,
+      "step": 18149
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0150820428526114e-05,
+      "loss": 0.3871,
+      "step": 18150
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0150243677365401e-05,
+      "loss": 0.4808,
+      "step": 18151
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0149666926204689e-05,
+      "loss": 0.4497,
+      "step": 18152
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0149090175043978e-05,
+      "loss": 0.4567,
+      "step": 18153
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0148513423883266e-05,
+      "loss": 0.4486,
+      "step": 18154
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0147936672722554e-05,
+      "loss": 0.5024,
+      "step": 18155
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0147359921561843e-05,
+      "loss": 0.5084,
+      "step": 18156
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.014678317040113e-05,
+      "loss": 0.4161,
+      "step": 18157
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0146206419240418e-05,
+      "loss": 0.4809,
+      "step": 18158
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0145629668079708e-05,
+      "loss": 0.4709,
+      "step": 18159
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0145052916918996e-05,
+      "loss": 0.4436,
+      "step": 18160
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0144476165758283e-05,
+      "loss": 0.3962,
+      "step": 18161
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0143899414597573e-05,
+      "loss": 0.443,
+      "step": 18162
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.014332266343686e-05,
+      "loss": 0.5237,
+      "step": 18163
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0142745912276148e-05,
+      "loss": 0.3444,
+      "step": 18164
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0142169161115439e-05,
+      "loss": 0.4036,
+      "step": 18165
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0141592409954725e-05,
+      "loss": 0.5012,
+      "step": 18166
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0141015658794013e-05,
+      "loss": 0.4688,
+      "step": 18167
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0140438907633304e-05,
+      "loss": 0.4931,
+      "step": 18168
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0139862156472591e-05,
+      "loss": 0.5456,
+      "step": 18169
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0139285405311877e-05,
+      "loss": 0.4943,
+      "step": 18170
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0138708654151169e-05,
+      "loss": 0.4505,
+      "step": 18171
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0138131902990456e-05,
+      "loss": 0.343,
+      "step": 18172
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0137555151829744e-05,
+      "loss": 0.3706,
+      "step": 18173
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0136978400669033e-05,
+      "loss": 0.42,
+      "step": 18174
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0136401649508321e-05,
+      "loss": 0.3556,
+      "step": 18175
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0135824898347609e-05,
+      "loss": 0.361,
+      "step": 18176
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0135248147186898e-05,
+      "loss": 0.4808,
+      "step": 18177
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0134671396026186e-05,
+      "loss": 0.4832,
+      "step": 18178
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0134094644865473e-05,
+      "loss": 0.443,
+      "step": 18179
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0133517893704763e-05,
+      "loss": 0.4341,
+      "step": 18180
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.013294114254405e-05,
+      "loss": 0.4705,
+      "step": 18181
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0132364391383338e-05,
+      "loss": 0.4594,
+      "step": 18182
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0131787640222628e-05,
+      "loss": 0.4376,
+      "step": 18183
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0131210889061915e-05,
+      "loss": 0.5061,
+      "step": 18184
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0130634137901203e-05,
+      "loss": 0.4669,
+      "step": 18185
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.013005738674049e-05,
+      "loss": 0.486,
+      "step": 18186
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.012948063557978e-05,
+      "loss": 0.5101,
+      "step": 18187
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0128903884419068e-05,
+      "loss": 0.4651,
+      "step": 18188
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0128327133258355e-05,
+      "loss": 0.4632,
+      "step": 18189
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0127750382097645e-05,
+      "loss": 0.412,
+      "step": 18190
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0127173630936932e-05,
+      "loss": 0.5873,
+      "step": 18191
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.012659687977622e-05,
+      "loss": 0.4001,
+      "step": 18192
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.012602012861551e-05,
+      "loss": 0.531,
+      "step": 18193
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0125443377454797e-05,
+      "loss": 0.4528,
+      "step": 18194
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0124866626294085e-05,
+      "loss": 0.4341,
+      "step": 18195
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0124289875133376e-05,
+      "loss": 0.5107,
+      "step": 18196
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0123713123972662e-05,
+      "loss": 0.3868,
+      "step": 18197
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.012313637281195e-05,
+      "loss": 0.4903,
+      "step": 18198
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.012255962165124e-05,
+      "loss": 0.5076,
+      "step": 18199
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0121982870490528e-05,
+      "loss": 0.3635,
+      "step": 18200
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0121406119329816e-05,
+      "loss": 0.5321,
+      "step": 18201
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0120829368169105e-05,
+      "loss": 0.5017,
+      "step": 18202
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0120252617008393e-05,
+      "loss": 0.4406,
+      "step": 18203
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.011967586584768e-05,
+      "loss": 0.4018,
+      "step": 18204
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.011909911468697e-05,
+      "loss": 0.4385,
+      "step": 18205
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0118522363526258e-05,
+      "loss": 0.4428,
+      "step": 18206
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0117945612365546e-05,
+      "loss": 0.5286,
+      "step": 18207
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0117368861204835e-05,
+      "loss": 0.5057,
+      "step": 18208
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0116792110044123e-05,
+      "loss": 0.5439,
+      "step": 18209
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.011621535888341e-05,
+      "loss": 0.4776,
+      "step": 18210
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.01156386077227e-05,
+      "loss": 0.4502,
+      "step": 18211
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0115061856561987e-05,
+      "loss": 0.4685,
+      "step": 18212
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0114485105401275e-05,
+      "loss": 0.3979,
+      "step": 18213
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0113908354240564e-05,
+      "loss": 0.4485,
+      "step": 18214
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0113331603079852e-05,
+      "loss": 0.4788,
+      "step": 18215
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.011275485191914e-05,
+      "loss": 0.4428,
+      "step": 18216
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.011217810075843e-05,
+      "loss": 0.3646,
+      "step": 18217
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0111601349597717e-05,
+      "loss": 0.3827,
+      "step": 18218
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0111024598437005e-05,
+      "loss": 0.437,
+      "step": 18219
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0110447847276294e-05,
+      "loss": 0.4607,
+      "step": 18220
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0109871096115582e-05,
+      "loss": 0.4765,
+      "step": 18221
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.010929434495487e-05,
+      "loss": 0.5161,
+      "step": 18222
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0108717593794157e-05,
+      "loss": 0.3715,
+      "step": 18223
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0108140842633446e-05,
+      "loss": 0.4339,
+      "step": 18224
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0107564091472734e-05,
+      "loss": 0.4812,
+      "step": 18225
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0106987340312022e-05,
+      "loss": 0.4565,
+      "step": 18226
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0106410589151313e-05,
+      "loss": 0.3683,
+      "step": 18227
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0105833837990599e-05,
+      "loss": 0.4438,
+      "step": 18228
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0105257086829886e-05,
+      "loss": 0.3941,
+      "step": 18229
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0104680335669178e-05,
+      "loss": 0.3644,
+      "step": 18230
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0104103584508465e-05,
+      "loss": 0.472,
+      "step": 18231
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0103526833347753e-05,
+      "loss": 0.4682,
+      "step": 18232
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0102950082187042e-05,
+      "loss": 0.5221,
+      "step": 18233
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.010237333102633e-05,
+      "loss": 0.5334,
+      "step": 18234
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0101796579865618e-05,
+      "loss": 0.4742,
+      "step": 18235
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0101219828704907e-05,
+      "loss": 0.4393,
+      "step": 18236
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0100643077544195e-05,
+      "loss": 0.4186,
+      "step": 18237
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0100066326383482e-05,
+      "loss": 0.5097,
+      "step": 18238
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0099489575222772e-05,
+      "loss": 0.4265,
+      "step": 18239
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.009891282406206e-05,
+      "loss": 0.4711,
+      "step": 18240
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0098336072901347e-05,
+      "loss": 0.4626,
+      "step": 18241
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0097759321740637e-05,
+      "loss": 0.3954,
+      "step": 18242
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0097182570579924e-05,
+      "loss": 0.5523,
+      "step": 18243
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0096605819419212e-05,
+      "loss": 0.5016,
+      "step": 18244
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0096029068258501e-05,
+      "loss": 0.3623,
+      "step": 18245
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0095452317097789e-05,
+      "loss": 0.3509,
+      "step": 18246
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0094875565937077e-05,
+      "loss": 0.4703,
+      "step": 18247
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0094298814776366e-05,
+      "loss": 0.4352,
+      "step": 18248
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0093722063615654e-05,
+      "loss": 0.5029,
+      "step": 18249
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0093145312454941e-05,
+      "loss": 0.565,
+      "step": 18250
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.009256856129423e-05,
+      "loss": 0.3888,
+      "step": 18251
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0091991810133518e-05,
+      "loss": 0.5978,
+      "step": 18252
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0091415058972806e-05,
+      "loss": 0.3838,
+      "step": 18253
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0090838307812096e-05,
+      "loss": 0.4968,
+      "step": 18254
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0090261556651383e-05,
+      "loss": 0.404,
+      "step": 18255
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0089684805490671e-05,
+      "loss": 0.3536,
+      "step": 18256
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0089108054329959e-05,
+      "loss": 0.3885,
+      "step": 18257
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.008853130316925e-05,
+      "loss": 0.3876,
+      "step": 18258
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0087954552008536e-05,
+      "loss": 0.4124,
+      "step": 18259
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0087377800847823e-05,
+      "loss": 0.528,
+      "step": 18260
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0086801049687114e-05,
+      "loss": 0.4777,
+      "step": 18261
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0086224298526402e-05,
+      "loss": 0.4299,
+      "step": 18262
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.008564754736569e-05,
+      "loss": 0.4763,
+      "step": 18263
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.008507079620498e-05,
+      "loss": 0.4944,
+      "step": 18264
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0084494045044267e-05,
+      "loss": 0.4023,
+      "step": 18265
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0083917293883555e-05,
+      "loss": 0.5194,
+      "step": 18266
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0083340542722844e-05,
+      "loss": 0.6088,
+      "step": 18267
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0082763791562132e-05,
+      "loss": 0.544,
+      "step": 18268
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.008218704040142e-05,
+      "loss": 0.5677,
+      "step": 18269
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0081610289240709e-05,
+      "loss": 0.4144,
+      "step": 18270
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0081033538079996e-05,
+      "loss": 0.4472,
+      "step": 18271
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0080456786919284e-05,
+      "loss": 0.4383,
+      "step": 18272
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0079880035758573e-05,
+      "loss": 0.5664,
+      "step": 18273
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0079303284597861e-05,
+      "loss": 0.4425,
+      "step": 18274
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0078726533437149e-05,
+      "loss": 0.5186,
+      "step": 18275
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0078149782276438e-05,
+      "loss": 0.4535,
+      "step": 18276
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0077573031115726e-05,
+      "loss": 0.4948,
+      "step": 18277
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0076996279955014e-05,
+      "loss": 0.3442,
+      "step": 18278
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0076419528794303e-05,
+      "loss": 0.4451,
+      "step": 18279
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.007584277763359e-05,
+      "loss": 0.5289,
+      "step": 18280
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0075266026472878e-05,
+      "loss": 0.419,
+      "step": 18281
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0074689275312168e-05,
+      "loss": 0.5068,
+      "step": 18282
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0074112524151455e-05,
+      "loss": 0.341,
+      "step": 18283
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0073535772990743e-05,
+      "loss": 0.4572,
+      "step": 18284
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0072959021830034e-05,
+      "loss": 0.5182,
+      "step": 18285
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.007238227066932e-05,
+      "loss": 0.3429,
+      "step": 18286
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0071805519508608e-05,
+      "loss": 0.4546,
+      "step": 18287
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0071228768347899e-05,
+      "loss": 0.3608,
+      "step": 18288
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0070652017187187e-05,
+      "loss": 0.4635,
+      "step": 18289
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0070075266026473e-05,
+      "loss": 0.5146,
+      "step": 18290
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.006949851486576e-05,
+      "loss": 0.4966,
+      "step": 18291
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0068921763705051e-05,
+      "loss": 0.5281,
+      "step": 18292
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0068345012544339e-05,
+      "loss": 0.4531,
+      "step": 18293
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0067768261383627e-05,
+      "loss": 0.478,
+      "step": 18294
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0067191510222916e-05,
+      "loss": 0.3921,
+      "step": 18295
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0066614759062204e-05,
+      "loss": 0.3526,
+      "step": 18296
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0066038007901491e-05,
+      "loss": 0.4668,
+      "step": 18297
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.006546125674078e-05,
+      "loss": 0.4464,
+      "step": 18298
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0064884505580068e-05,
+      "loss": 0.3749,
+      "step": 18299
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0064307754419356e-05,
+      "loss": 0.5179,
+      "step": 18300
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0063731003258646e-05,
+      "loss": 0.3957,
+      "step": 18301
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0063154252097933e-05,
+      "loss": 0.3946,
+      "step": 18302
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0062577500937221e-05,
+      "loss": 0.2838,
+      "step": 18303
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.006200074977651e-05,
+      "loss": 0.476,
+      "step": 18304
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0061423998615798e-05,
+      "loss": 0.4635,
+      "step": 18305
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0060847247455086e-05,
+      "loss": 0.3529,
+      "step": 18306
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0060270496294375e-05,
+      "loss": 0.5159,
+      "step": 18307
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0059693745133663e-05,
+      "loss": 0.4586,
+      "step": 18308
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.005911699397295e-05,
+      "loss": 0.308,
+      "step": 18309
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.005854024281224e-05,
+      "loss": 0.6137,
+      "step": 18310
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0057963491651527e-05,
+      "loss": 0.6273,
+      "step": 18311
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0057386740490815e-05,
+      "loss": 0.5472,
+      "step": 18312
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0056809989330105e-05,
+      "loss": 0.4383,
+      "step": 18313
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0056233238169392e-05,
+      "loss": 0.4535,
+      "step": 18314
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.005565648700868e-05,
+      "loss": 0.4908,
+      "step": 18315
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0055079735847971e-05,
+      "loss": 0.4975,
+      "step": 18316
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0054502984687257e-05,
+      "loss": 0.4754,
+      "step": 18317
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0053926233526545e-05,
+      "loss": 0.4672,
+      "step": 18318
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0053349482365836e-05,
+      "loss": 0.4011,
+      "step": 18319
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.0052772731205123e-05,
+      "loss": 0.3962,
+      "step": 18320
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 1.005219598004441e-05,
+      "loss": 0.5022,
+      "step": 18321
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.00516192288837e-05,
+      "loss": 0.4475,
+      "step": 18322
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0051042477722988e-05,
+      "loss": 0.549,
+      "step": 18323
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0050465726562276e-05,
+      "loss": 0.5038,
+      "step": 18324
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0049888975401564e-05,
+      "loss": 0.5132,
+      "step": 18325
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0049312224240853e-05,
+      "loss": 0.4746,
+      "step": 18326
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.004873547308014e-05,
+      "loss": 0.4085,
+      "step": 18327
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0048158721919428e-05,
+      "loss": 0.5812,
+      "step": 18328
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0047581970758718e-05,
+      "loss": 0.4703,
+      "step": 18329
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0047005219598005e-05,
+      "loss": 0.4865,
+      "step": 18330
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0046428468437293e-05,
+      "loss": 0.433,
+      "step": 18331
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0045851717276582e-05,
+      "loss": 0.4453,
+      "step": 18332
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.004527496611587e-05,
+      "loss": 0.5573,
+      "step": 18333
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0044698214955158e-05,
+      "loss": 0.5324,
+      "step": 18334
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0044121463794447e-05,
+      "loss": 0.4684,
+      "step": 18335
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0043544712633735e-05,
+      "loss": 0.4737,
+      "step": 18336
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0042967961473023e-05,
+      "loss": 0.3645,
+      "step": 18337
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0042391210312312e-05,
+      "loss": 0.411,
+      "step": 18338
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.00418144591516e-05,
+      "loss": 0.4631,
+      "step": 18339
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0041237707990887e-05,
+      "loss": 0.3608,
+      "step": 18340
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0040660956830177e-05,
+      "loss": 0.3994,
+      "step": 18341
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0040084205669464e-05,
+      "loss": 0.5192,
+      "step": 18342
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0039507454508752e-05,
+      "loss": 0.3732,
+      "step": 18343
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0038930703348041e-05,
+      "loss": 0.4471,
+      "step": 18344
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0038353952187329e-05,
+      "loss": 0.4254,
+      "step": 18345
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0037777201026617e-05,
+      "loss": 0.4646,
+      "step": 18346
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0037200449865908e-05,
+      "loss": 0.3929,
+      "step": 18347
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0036623698705194e-05,
+      "loss": 0.502,
+      "step": 18348
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0036046947544482e-05,
+      "loss": 0.3701,
+      "step": 18349
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0035470196383773e-05,
+      "loss": 0.4303,
+      "step": 18350
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.003489344522306e-05,
+      "loss": 0.412,
+      "step": 18351
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0034316694062346e-05,
+      "loss": 0.4216,
+      "step": 18352
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0033739942901637e-05,
+      "loss": 0.4702,
+      "step": 18353
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0033163191740925e-05,
+      "loss": 0.3545,
+      "step": 18354
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0032586440580213e-05,
+      "loss": 0.4198,
+      "step": 18355
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0032009689419502e-05,
+      "loss": 0.3646,
+      "step": 18356
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.003143293825879e-05,
+      "loss": 0.4606,
+      "step": 18357
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0030856187098077e-05,
+      "loss": 0.4425,
+      "step": 18358
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0030279435937367e-05,
+      "loss": 0.5339,
+      "step": 18359
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0029702684776655e-05,
+      "loss": 0.5142,
+      "step": 18360
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0029125933615942e-05,
+      "loss": 0.4949,
+      "step": 18361
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.002854918245523e-05,
+      "loss": 0.4114,
+      "step": 18362
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.002797243129452e-05,
+      "loss": 0.4029,
+      "step": 18363
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0027395680133807e-05,
+      "loss": 0.5248,
+      "step": 18364
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0026818928973095e-05,
+      "loss": 0.403,
+      "step": 18365
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0026242177812384e-05,
+      "loss": 0.4449,
+      "step": 18366
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0025665426651672e-05,
+      "loss": 0.3361,
+      "step": 18367
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.002508867549096e-05,
+      "loss": 0.446,
+      "step": 18368
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0024511924330249e-05,
+      "loss": 0.6124,
+      "step": 18369
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0023935173169536e-05,
+      "loss": 0.5353,
+      "step": 18370
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0023358422008824e-05,
+      "loss": 0.4574,
+      "step": 18371
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0022781670848114e-05,
+      "loss": 0.4306,
+      "step": 18372
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0022204919687401e-05,
+      "loss": 0.4056,
+      "step": 18373
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0021628168526689e-05,
+      "loss": 0.4903,
+      "step": 18374
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0021051417365978e-05,
+      "loss": 0.5197,
+      "step": 18375
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0020474666205266e-05,
+      "loss": 0.4829,
+      "step": 18376
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0019897915044554e-05,
+      "loss": 0.336,
+      "step": 18377
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0019321163883845e-05,
+      "loss": 0.3954,
+      "step": 18378
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.001874441272313e-05,
+      "loss": 0.3854,
+      "step": 18379
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0018167661562418e-05,
+      "loss": 0.5153,
+      "step": 18380
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.001759091040171e-05,
+      "loss": 0.3479,
+      "step": 18381
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0017014159240997e-05,
+      "loss": 0.4955,
+      "step": 18382
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0016437408080283e-05,
+      "loss": 0.4652,
+      "step": 18383
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0015860656919574e-05,
+      "loss": 0.5606,
+      "step": 18384
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0015283905758862e-05,
+      "loss": 0.4652,
+      "step": 18385
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.001470715459815e-05,
+      "loss": 0.5444,
+      "step": 18386
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0014130403437439e-05,
+      "loss": 0.4532,
+      "step": 18387
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0013553652276727e-05,
+      "loss": 0.3944,
+      "step": 18388
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0012976901116014e-05,
+      "loss": 0.5237,
+      "step": 18389
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0012400149955304e-05,
+      "loss": 0.5476,
+      "step": 18390
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0011823398794591e-05,
+      "loss": 0.4377,
+      "step": 18391
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0011246647633879e-05,
+      "loss": 0.4197,
+      "step": 18392
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0010669896473168e-05,
+      "loss": 0.4236,
+      "step": 18393
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0010093145312456e-05,
+      "loss": 0.4952,
+      "step": 18394
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0009516394151744e-05,
+      "loss": 0.4508,
+      "step": 18395
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0008939642991032e-05,
+      "loss": 0.4184,
+      "step": 18396
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0008362891830321e-05,
+      "loss": 0.4998,
+      "step": 18397
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0007786140669609e-05,
+      "loss": 0.6732,
+      "step": 18398
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0007209389508896e-05,
+      "loss": 0.4993,
+      "step": 18399
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0006632638348186e-05,
+      "loss": 0.3115,
+      "step": 18400
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0006055887187473e-05,
+      "loss": 0.4782,
+      "step": 18401
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0005479136026761e-05,
+      "loss": 0.6009,
+      "step": 18402
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.000490238486605e-05,
+      "loss": 0.388,
+      "step": 18403
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0004325633705338e-05,
+      "loss": 0.4759,
+      "step": 18404
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0003748882544626e-05,
+      "loss": 0.5167,
+      "step": 18405
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0003172131383915e-05,
+      "loss": 0.3484,
+      "step": 18406
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0002595380223203e-05,
+      "loss": 0.4691,
+      "step": 18407
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.000201862906249e-05,
+      "loss": 0.4409,
+      "step": 18408
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0001441877901782e-05,
+      "loss": 0.5596,
+      "step": 18409
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0000865126741068e-05,
+      "loss": 0.3947,
+      "step": 18410
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.0000288375580355e-05,
+      "loss": 0.3369,
+      "step": 18411
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.999711624419645e-06,
+      "loss": 0.424,
+      "step": 18412
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.999134873258934e-06,
+      "loss": 0.5086,
+      "step": 18413
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.998558122098222e-06,
+      "loss": 0.481,
+      "step": 18414
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.99798137093751e-06,
+      "loss": 0.423,
+      "step": 18415
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.997404619776799e-06,
+      "loss": 0.4465,
+      "step": 18416
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.996827868616086e-06,
+      "loss": 0.4449,
+      "step": 18417
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.996251117455374e-06,
+      "loss": 0.4257,
+      "step": 18418
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.995674366294664e-06,
+      "loss": 0.5136,
+      "step": 18419
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.995097615133951e-06,
+      "loss": 0.4484,
+      "step": 18420
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.994520863973239e-06,
+      "loss": 0.5097,
+      "step": 18421
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.993944112812528e-06,
+      "loss": 0.4427,
+      "step": 18422
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.993367361651816e-06,
+      "loss": 0.5221,
+      "step": 18423
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.992790610491104e-06,
+      "loss": 0.3998,
+      "step": 18424
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.992213859330393e-06,
+      "loss": 0.3776,
+      "step": 18425
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.99163710816968e-06,
+      "loss": 0.4828,
+      "step": 18426
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.99106035700897e-06,
+      "loss": 0.481,
+      "step": 18427
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.990483605848258e-06,
+      "loss": 0.4639,
+      "step": 18428
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.989906854687545e-06,
+      "loss": 0.4716,
+      "step": 18429
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.989330103526835e-06,
+      "loss": 0.4215,
+      "step": 18430
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.988753352366123e-06,
+      "loss": 0.4528,
+      "step": 18431
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.98817660120541e-06,
+      "loss": 0.4703,
+      "step": 18432
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.9875998500447e-06,
+      "loss": 0.4679,
+      "step": 18433
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.987023098883987e-06,
+      "loss": 0.5387,
+      "step": 18434
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.986446347723275e-06,
+      "loss": 0.5605,
+      "step": 18435
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.985869596562564e-06,
+      "loss": 0.4235,
+      "step": 18436
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.985292845401852e-06,
+      "loss": 0.3531,
+      "step": 18437
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.98471609424114e-06,
+      "loss": 0.4256,
+      "step": 18438
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.984139343080429e-06,
+      "loss": 0.4197,
+      "step": 18439
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.983562591919717e-06,
+      "loss": 0.4241,
+      "step": 18440
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.982985840759004e-06,
+      "loss": 0.4492,
+      "step": 18441
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.982409089598294e-06,
+      "loss": 0.3865,
+      "step": 18442
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.981832338437582e-06,
+      "loss": 0.4394,
+      "step": 18443
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.981255587276871e-06,
+      "loss": 0.4352,
+      "step": 18444
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.980678836116159e-06,
+      "loss": 0.5156,
+      "step": 18445
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.980102084955446e-06,
+      "loss": 0.35,
+      "step": 18446
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.979525333794736e-06,
+      "loss": 0.4032,
+      "step": 18447
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.978948582634023e-06,
+      "loss": 0.4116,
+      "step": 18448
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.978371831473311e-06,
+      "loss": 0.2928,
+      "step": 18449
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.9777950803126e-06,
+      "loss": 0.5058,
+      "step": 18450
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.977218329151888e-06,
+      "loss": 0.2784,
+      "step": 18451
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.976641577991176e-06,
+      "loss": 0.4558,
+      "step": 18452
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.976064826830465e-06,
+      "loss": 0.4384,
+      "step": 18453
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.975488075669753e-06,
+      "loss": 0.387,
+      "step": 18454
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.97491132450904e-06,
+      "loss": 0.4788,
+      "step": 18455
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.97433457334833e-06,
+      "loss": 0.434,
+      "step": 18456
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.973757822187618e-06,
+      "loss": 0.4285,
+      "step": 18457
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.973181071026907e-06,
+      "loss": 0.3961,
+      "step": 18458
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.972604319866195e-06,
+      "loss": 0.4644,
+      "step": 18459
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.972027568705482e-06,
+      "loss": 0.4335,
+      "step": 18460
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.971450817544772e-06,
+      "loss": 0.516,
+      "step": 18461
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.97087406638406e-06,
+      "loss": 0.3363,
+      "step": 18462
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.970297315223347e-06,
+      "loss": 0.3951,
+      "step": 18463
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.969720564062636e-06,
+      "loss": 0.504,
+      "step": 18464
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.969143812901924e-06,
+      "loss": 0.4622,
+      "step": 18465
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.968567061741212e-06,
+      "loss": 0.3139,
+      "step": 18466
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.967990310580501e-06,
+      "loss": 0.5197,
+      "step": 18467
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.967413559419789e-06,
+      "loss": 0.4509,
+      "step": 18468
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.966836808259077e-06,
+      "loss": 0.4923,
+      "step": 18469
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.966260057098366e-06,
+      "loss": 0.3832,
+      "step": 18470
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.965683305937655e-06,
+      "loss": 0.434,
+      "step": 18471
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.965106554776941e-06,
+      "loss": 0.4775,
+      "step": 18472
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.96452980361623e-06,
+      "loss": 0.5165,
+      "step": 18473
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.963953052455518e-06,
+      "loss": 0.4711,
+      "step": 18474
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.963376301294808e-06,
+      "loss": 0.4599,
+      "step": 18475
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.962799550134095e-06,
+      "loss": 0.4552,
+      "step": 18476
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.962222798973383e-06,
+      "loss": 0.4284,
+      "step": 18477
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.961646047812673e-06,
+      "loss": 0.4742,
+      "step": 18478
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.96106929665196e-06,
+      "loss": 0.4823,
+      "step": 18479
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.960492545491248e-06,
+      "loss": 0.4601,
+      "step": 18480
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.959915794330537e-06,
+      "loss": 0.4481,
+      "step": 18481
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.959339043169825e-06,
+      "loss": 0.416,
+      "step": 18482
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958762292009113e-06,
+      "loss": 0.5298,
+      "step": 18483
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.958185540848402e-06,
+      "loss": 0.3863,
+      "step": 18484
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.95760878968769e-06,
+      "loss": 0.4655,
+      "step": 18485
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.957032038526977e-06,
+      "loss": 0.4735,
+      "step": 18486
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.956455287366267e-06,
+      "loss": 0.443,
+      "step": 18487
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.955878536205556e-06,
+      "loss": 0.4526,
+      "step": 18488
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.955301785044844e-06,
+      "loss": 0.3423,
+      "step": 18489
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.954725033884132e-06,
+      "loss": 0.5601,
+      "step": 18490
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.95414828272342e-06,
+      "loss": 0.5176,
+      "step": 18491
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.953571531562709e-06,
+      "loss": 0.3818,
+      "step": 18492
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.952994780401996e-06,
+      "loss": 0.5451,
+      "step": 18493
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.952418029241284e-06,
+      "loss": 0.4194,
+      "step": 18494
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.951841278080573e-06,
+      "loss": 0.5649,
+      "step": 18495
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.951264526919861e-06,
+      "loss": 0.452,
+      "step": 18496
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.950687775759149e-06,
+      "loss": 0.4238,
+      "step": 18497
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.950111024598438e-06,
+      "loss": 0.3026,
+      "step": 18498
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.949534273437726e-06,
+      "loss": 0.3031,
+      "step": 18499
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 9.948957522277013e-06,
+      "loss": 0.5971,
+      "step": 18500
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.948380771116303e-06,
+      "loss": 0.3476,
+      "step": 18501
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.947804019955592e-06,
+      "loss": 0.3857,
+      "step": 18502
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.947227268794878e-06,
+      "loss": 0.4007,
+      "step": 18503
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.946650517634168e-06,
+      "loss": 0.4638,
+      "step": 18504
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.946073766473457e-06,
+      "loss": 0.4803,
+      "step": 18505
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.945497015312745e-06,
+      "loss": 0.3787,
+      "step": 18506
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.944920264152032e-06,
+      "loss": 0.394,
+      "step": 18507
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.94434351299132e-06,
+      "loss": 0.3744,
+      "step": 18508
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.94376676183061e-06,
+      "loss": 0.4972,
+      "step": 18509
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.943190010669897e-06,
+      "loss": 0.545,
+      "step": 18510
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.942613259509185e-06,
+      "loss": 0.3898,
+      "step": 18511
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.942036508348474e-06,
+      "loss": 0.4984,
+      "step": 18512
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.941459757187762e-06,
+      "loss": 0.5857,
+      "step": 18513
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.94088300602705e-06,
+      "loss": 0.3909,
+      "step": 18514
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.940306254866339e-06,
+      "loss": 0.3996,
+      "step": 18515
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.939729503705627e-06,
+      "loss": 0.4263,
+      "step": 18516
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.939152752544914e-06,
+      "loss": 0.4165,
+      "step": 18517
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.938576001384204e-06,
+      "loss": 0.4963,
+      "step": 18518
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.937999250223493e-06,
+      "loss": 0.4827,
+      "step": 18519
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.93742249906278e-06,
+      "loss": 0.3993,
+      "step": 18520
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.936845747902068e-06,
+      "loss": 0.4512,
+      "step": 18521
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.936268996741358e-06,
+      "loss": 0.5088,
+      "step": 18522
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.935692245580645e-06,
+      "loss": 0.5265,
+      "step": 18523
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.935115494419933e-06,
+      "loss": 0.4893,
+      "step": 18524
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.93453874325922e-06,
+      "loss": 0.4386,
+      "step": 18525
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.93396199209851e-06,
+      "loss": 0.4584,
+      "step": 18526
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.933385240937798e-06,
+      "loss": 0.4758,
+      "step": 18527
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.932808489777086e-06,
+      "loss": 0.3737,
+      "step": 18528
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.932231738616375e-06,
+      "loss": 0.501,
+      "step": 18529
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.931654987455663e-06,
+      "loss": 0.5264,
+      "step": 18530
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.93107823629495e-06,
+      "loss": 0.3336,
+      "step": 18531
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.93050148513424e-06,
+      "loss": 0.3658,
+      "step": 18532
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.929924733973529e-06,
+      "loss": 0.5598,
+      "step": 18533
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.929347982812815e-06,
+      "loss": 0.4348,
+      "step": 18534
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.928771231652104e-06,
+      "loss": 0.6017,
+      "step": 18535
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.928194480491394e-06,
+      "loss": 0.5184,
+      "step": 18536
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.927617729330682e-06,
+      "loss": 0.544,
+      "step": 18537
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.92704097816997e-06,
+      "loss": 0.3833,
+      "step": 18538
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.926464227009259e-06,
+      "loss": 0.4873,
+      "step": 18539
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.925887475848546e-06,
+      "loss": 0.5069,
+      "step": 18540
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.925310724687834e-06,
+      "loss": 0.4555,
+      "step": 18541
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.924733973527123e-06,
+      "loss": 0.4994,
+      "step": 18542
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.924157222366411e-06,
+      "loss": 0.4314,
+      "step": 18543
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.923580471205699e-06,
+      "loss": 0.5288,
+      "step": 18544
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.923003720044986e-06,
+      "loss": 0.4607,
+      "step": 18545
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.922426968884276e-06,
+      "loss": 0.5857,
+      "step": 18546
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.921850217723563e-06,
+      "loss": 0.4179,
+      "step": 18547
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.921273466562851e-06,
+      "loss": 0.4676,
+      "step": 18548
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.92069671540214e-06,
+      "loss": 0.4717,
+      "step": 18549
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.92011996424143e-06,
+      "loss": 0.3362,
+      "step": 18550
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.919543213080718e-06,
+      "loss": 0.4273,
+      "step": 18551
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.918966461920005e-06,
+      "loss": 0.4534,
+      "step": 18552
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.918389710759295e-06,
+      "loss": 0.493,
+      "step": 18553
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.917812959598582e-06,
+      "loss": 0.3528,
+      "step": 18554
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91723620843787e-06,
+      "loss": 0.4739,
+      "step": 18555
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91665945727716e-06,
+      "loss": 0.4359,
+      "step": 18556
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.916082706116447e-06,
+      "loss": 0.4436,
+      "step": 18557
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.915505954955735e-06,
+      "loss": 0.4737,
+      "step": 18558
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.914929203795024e-06,
+      "loss": 0.4198,
+      "step": 18559
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.914352452634312e-06,
+      "loss": 0.4031,
+      "step": 18560
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.9137757014736e-06,
+      "loss": 0.3427,
+      "step": 18561
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.913198950312887e-06,
+      "loss": 0.515,
+      "step": 18562
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.912622199152177e-06,
+      "loss": 0.423,
+      "step": 18563
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.912045447991466e-06,
+      "loss": 0.5118,
+      "step": 18564
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.911468696830752e-06,
+      "loss": 0.5378,
+      "step": 18565
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.910891945670041e-06,
+      "loss": 0.4673,
+      "step": 18566
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.91031519450933e-06,
+      "loss": 0.5755,
+      "step": 18567
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.909738443348618e-06,
+      "loss": 0.4168,
+      "step": 18568
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.909161692187906e-06,
+      "loss": 0.4576,
+      "step": 18569
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.908584941027195e-06,
+      "loss": 0.4106,
+      "step": 18570
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.908008189866483e-06,
+      "loss": 0.4019,
+      "step": 18571
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.90743143870577e-06,
+      "loss": 0.5376,
+      "step": 18572
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.90685468754506e-06,
+      "loss": 0.4476,
+      "step": 18573
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.906277936384348e-06,
+      "loss": 0.6515,
+      "step": 18574
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.905701185223636e-06,
+      "loss": 0.3055,
+      "step": 18575
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.905124434062925e-06,
+      "loss": 0.561,
+      "step": 18576
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.904547682902213e-06,
+      "loss": 0.3754,
+      "step": 18577
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.9039709317415e-06,
+      "loss": 0.4874,
+      "step": 18578
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.903394180580788e-06,
+      "loss": 0.4333,
+      "step": 18579
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.902817429420077e-06,
+      "loss": 0.3884,
+      "step": 18580
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.902240678259367e-06,
+      "loss": 0.4569,
+      "step": 18581
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.901663927098654e-06,
+      "loss": 0.5969,
+      "step": 18582
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.901087175937942e-06,
+      "loss": 0.3466,
+      "step": 18583
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.900510424777232e-06,
+      "loss": 0.4711,
+      "step": 18584
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.89993367361652e-06,
+      "loss": 0.4313,
+      "step": 18585
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.899356922455807e-06,
+      "loss": 0.4267,
+      "step": 18586
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.898780171295096e-06,
+      "loss": 0.4876,
+      "step": 18587
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.898203420134384e-06,
+      "loss": 0.454,
+      "step": 18588
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897626668973672e-06,
+      "loss": 0.5605,
+      "step": 18589
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.897049917812961e-06,
+      "loss": 0.5056,
+      "step": 18590
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.896473166652249e-06,
+      "loss": 0.4588,
+      "step": 18591
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.895896415491536e-06,
+      "loss": 0.4469,
+      "step": 18592
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.895319664330826e-06,
+      "loss": 0.5683,
+      "step": 18593
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.894742913170113e-06,
+      "loss": 0.3996,
+      "step": 18594
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.894166162009403e-06,
+      "loss": 0.3961,
+      "step": 18595
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.893589410848689e-06,
+      "loss": 0.3483,
+      "step": 18596
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.893012659687978e-06,
+      "loss": 0.3871,
+      "step": 18597
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.892435908527268e-06,
+      "loss": 0.5465,
+      "step": 18598
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.891859157366555e-06,
+      "loss": 0.4681,
+      "step": 18599
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.891282406205843e-06,
+      "loss": 0.4359,
+      "step": 18600
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.890705655045132e-06,
+      "loss": 0.5081,
+      "step": 18601
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.89012890388442e-06,
+      "loss": 0.3908,
+      "step": 18602
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.889552152723708e-06,
+      "loss": 0.4501,
+      "step": 18603
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.888975401562997e-06,
+      "loss": 0.4058,
+      "step": 18604
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.888398650402285e-06,
+      "loss": 0.5571,
+      "step": 18605
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.887821899241572e-06,
+      "loss": 0.4485,
+      "step": 18606
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.887245148080862e-06,
+      "loss": 0.5447,
+      "step": 18607
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.88666839692015e-06,
+      "loss": 0.5593,
+      "step": 18608
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.886091645759437e-06,
+      "loss": 0.3342,
+      "step": 18609
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.885514894598727e-06,
+      "loss": 0.4268,
+      "step": 18610
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.884938143438014e-06,
+      "loss": 0.5025,
+      "step": 18611
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.884361392277304e-06,
+      "loss": 0.391,
+      "step": 18612
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.883784641116591e-06,
+      "loss": 0.4451,
+      "step": 18613
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.883207889955879e-06,
+      "loss": 0.3814,
+      "step": 18614
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.882631138795168e-06,
+      "loss": 0.4852,
+      "step": 18615
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.882054387634456e-06,
+      "loss": 0.3715,
+      "step": 18616
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.881477636473744e-06,
+      "loss": 0.4242,
+      "step": 18617
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.880900885313033e-06,
+      "loss": 0.5082,
+      "step": 18618
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.88032413415232e-06,
+      "loss": 0.4863,
+      "step": 18619
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.879747382991609e-06,
+      "loss": 0.3149,
+      "step": 18620
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.879170631830898e-06,
+      "loss": 0.3445,
+      "step": 18621
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.878593880670186e-06,
+      "loss": 0.6019,
+      "step": 18622
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.878017129509473e-06,
+      "loss": 0.3658,
+      "step": 18623
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.877440378348763e-06,
+      "loss": 0.4155,
+      "step": 18624
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.87686362718805e-06,
+      "loss": 0.3822,
+      "step": 18625
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.87628687602734e-06,
+      "loss": 0.4855,
+      "step": 18626
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.875710124866627e-06,
+      "loss": 0.5809,
+      "step": 18627
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.875133373705915e-06,
+      "loss": 0.3457,
+      "step": 18628
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.874556622545204e-06,
+      "loss": 0.4454,
+      "step": 18629
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.873979871384492e-06,
+      "loss": 0.6165,
+      "step": 18630
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.87340312022378e-06,
+      "loss": 0.4937,
+      "step": 18631
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.87282636906307e-06,
+      "loss": 0.4416,
+      "step": 18632
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.872249617902357e-06,
+      "loss": 0.5523,
+      "step": 18633
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.871672866741645e-06,
+      "loss": 0.4832,
+      "step": 18634
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.871096115580934e-06,
+      "loss": 0.3636,
+      "step": 18635
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.870519364420222e-06,
+      "loss": 0.3896,
+      "step": 18636
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86994261325951e-06,
+      "loss": 0.4583,
+      "step": 18637
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.869365862098799e-06,
+      "loss": 0.4548,
+      "step": 18638
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.868789110938086e-06,
+      "loss": 0.381,
+      "step": 18639
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.868212359777374e-06,
+      "loss": 0.5206,
+      "step": 18640
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.867635608616663e-06,
+      "loss": 0.4085,
+      "step": 18641
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.867058857455951e-06,
+      "loss": 0.4844,
+      "step": 18642
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86648210629524e-06,
+      "loss": 0.4138,
+      "step": 18643
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.865905355134528e-06,
+      "loss": 0.468,
+      "step": 18644
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.865328603973816e-06,
+      "loss": 0.5346,
+      "step": 18645
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.864751852813105e-06,
+      "loss": 0.4034,
+      "step": 18646
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.864175101652393e-06,
+      "loss": 0.3455,
+      "step": 18647
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86359835049168e-06,
+      "loss": 0.4149,
+      "step": 18648
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86302159933097e-06,
+      "loss": 0.3918,
+      "step": 18649
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.862444848170258e-06,
+      "loss": 0.4673,
+      "step": 18650
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.861868097009545e-06,
+      "loss": 0.3177,
+      "step": 18651
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.861291345848835e-06,
+      "loss": 0.5269,
+      "step": 18652
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.860714594688122e-06,
+      "loss": 0.4365,
+      "step": 18653
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.86013784352741e-06,
+      "loss": 0.4268,
+      "step": 18654
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8595610923667e-06,
+      "loss": 0.4083,
+      "step": 18655
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.858984341205987e-06,
+      "loss": 0.5336,
+      "step": 18656
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.858407590045277e-06,
+      "loss": 0.6764,
+      "step": 18657
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.857830838884564e-06,
+      "loss": 0.4003,
+      "step": 18658
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.857254087723852e-06,
+      "loss": 0.4151,
+      "step": 18659
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.856677336563141e-06,
+      "loss": 0.3997,
+      "step": 18660
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.856100585402429e-06,
+      "loss": 0.4649,
+      "step": 18661
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.855523834241717e-06,
+      "loss": 0.4946,
+      "step": 18662
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.854947083081006e-06,
+      "loss": 0.4448,
+      "step": 18663
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.854370331920294e-06,
+      "loss": 0.4099,
+      "step": 18664
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.853793580759581e-06,
+      "loss": 0.385,
+      "step": 18665
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.85321682959887e-06,
+      "loss": 0.4558,
+      "step": 18666
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.852640078438159e-06,
+      "loss": 0.4452,
+      "step": 18667
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.852063327277446e-06,
+      "loss": 0.4142,
+      "step": 18668
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.851486576116736e-06,
+      "loss": 0.468,
+      "step": 18669
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.850909824956023e-06,
+      "loss": 0.3918,
+      "step": 18670
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.850333073795311e-06,
+      "loss": 0.3933,
+      "step": 18671
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.8497563226346e-06,
+      "loss": 0.4342,
+      "step": 18672
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.849179571473888e-06,
+      "loss": 0.4078,
+      "step": 18673
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.848602820313177e-06,
+      "loss": 0.5337,
+      "step": 18674
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.848026069152465e-06,
+      "loss": 0.4438,
+      "step": 18675
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.847449317991753e-06,
+      "loss": 0.4334,
+      "step": 18676
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.846872566831042e-06,
+      "loss": 0.4102,
+      "step": 18677
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.84629581567033e-06,
+      "loss": 0.4725,
+      "step": 18678
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 9.845719064509618e-06,
+      "loss": 0.4419,
+      "step": 18679
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.845142313348907e-06,
+      "loss": 0.498,
+      "step": 18680
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.844565562188195e-06,
+      "loss": 0.5452,
+      "step": 18681
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.843988811027482e-06,
+      "loss": 0.3827,
+      "step": 18682
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.843412059866772e-06,
+      "loss": 0.4991,
+      "step": 18683
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.84283530870606e-06,
+      "loss": 0.4466,
+      "step": 18684
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.842258557545347e-06,
+      "loss": 0.5142,
+      "step": 18685
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.841681806384636e-06,
+      "loss": 0.486,
+      "step": 18686
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.841105055223924e-06,
+      "loss": 0.3946,
+      "step": 18687
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.840528304063213e-06,
+      "loss": 0.4815,
+      "step": 18688
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.839951552902501e-06,
+      "loss": 0.4769,
+      "step": 18689
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.839374801741789e-06,
+      "loss": 0.4444,
+      "step": 18690
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.838798050581078e-06,
+      "loss": 0.4001,
+      "step": 18691
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.838221299420366e-06,
+      "loss": 0.4242,
+      "step": 18692
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.837644548259654e-06,
+      "loss": 0.4523,
+      "step": 18693
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.837067797098943e-06,
+      "loss": 0.2689,
+      "step": 18694
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.83649104593823e-06,
+      "loss": 0.3871,
+      "step": 18695
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835914294777518e-06,
+      "loss": 0.4564,
+      "step": 18696
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.835337543616808e-06,
+      "loss": 0.4398,
+      "step": 18697
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.834760792456095e-06,
+      "loss": 0.4589,
+      "step": 18698
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.834184041295383e-06,
+      "loss": 0.448,
+      "step": 18699
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.833607290134672e-06,
+      "loss": 0.3952,
+      "step": 18700
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.83303053897396e-06,
+      "loss": 0.4223,
+      "step": 18701
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.832453787813248e-06,
+      "loss": 0.4573,
+      "step": 18702
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.831877036652537e-06,
+      "loss": 0.4283,
+      "step": 18703
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.831300285491825e-06,
+      "loss": 0.3395,
+      "step": 18704
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.830723534331114e-06,
+      "loss": 0.3544,
+      "step": 18705
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.830146783170402e-06,
+      "loss": 0.4664,
+      "step": 18706
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.82957003200969e-06,
+      "loss": 0.4196,
+      "step": 18707
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.828993280848979e-06,
+      "loss": 0.4529,
+      "step": 18708
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.828416529688267e-06,
+      "loss": 0.3057,
+      "step": 18709
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.827839778527554e-06,
+      "loss": 0.4992,
+      "step": 18710
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.827263027366844e-06,
+      "loss": 0.3588,
+      "step": 18711
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.826686276206131e-06,
+      "loss": 0.4449,
+      "step": 18712
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.826109525045419e-06,
+      "loss": 0.3986,
+      "step": 18713
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.825532773884709e-06,
+      "loss": 0.4627,
+      "step": 18714
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.824956022723998e-06,
+      "loss": 0.4269,
+      "step": 18715
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.824379271563284e-06,
+      "loss": 0.5605,
+      "step": 18716
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.823802520402573e-06,
+      "loss": 0.5394,
+      "step": 18717
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.823225769241861e-06,
+      "loss": 0.399,
+      "step": 18718
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.82264901808115e-06,
+      "loss": 0.4593,
+      "step": 18719
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.822072266920438e-06,
+      "loss": 0.4385,
+      "step": 18720
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.821495515759726e-06,
+      "loss": 0.3048,
+      "step": 18721
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.820918764599015e-06,
+      "loss": 0.4157,
+      "step": 18722
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.820342013438303e-06,
+      "loss": 0.386,
+      "step": 18723
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.81976526227759e-06,
+      "loss": 0.5487,
+      "step": 18724
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.81918851111688e-06,
+      "loss": 0.3643,
+      "step": 18725
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.818611759956168e-06,
+      "loss": 0.3073,
+      "step": 18726
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.818035008795455e-06,
+      "loss": 0.4494,
+      "step": 18727
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.817458257634745e-06,
+      "loss": 0.4961,
+      "step": 18728
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.816881506474032e-06,
+      "loss": 0.5669,
+      "step": 18729
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.81630475531332e-06,
+      "loss": 0.436,
+      "step": 18730
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.81572800415261e-06,
+      "loss": 0.5719,
+      "step": 18731
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.815151252991899e-06,
+      "loss": 0.4696,
+      "step": 18732
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.814574501831185e-06,
+      "loss": 0.5585,
+      "step": 18733
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.813997750670474e-06,
+      "loss": 0.4836,
+      "step": 18734
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.813420999509762e-06,
+      "loss": 0.5457,
+      "step": 18735
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.812844248349051e-06,
+      "loss": 0.4125,
+      "step": 18736
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.812267497188339e-06,
+      "loss": 0.4033,
+      "step": 18737
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.811690746027627e-06,
+      "loss": 0.3339,
+      "step": 18738
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.811113994866916e-06,
+      "loss": 0.4815,
+      "step": 18739
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.810537243706204e-06,
+      "loss": 0.456,
+      "step": 18740
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.809960492545491e-06,
+      "loss": 0.4284,
+      "step": 18741
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.80938374138478e-06,
+      "loss": 0.4055,
+      "step": 18742
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.808806990224068e-06,
+      "loss": 0.4687,
+      "step": 18743
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.808230239063356e-06,
+      "loss": 0.5277,
+      "step": 18744
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.807653487902645e-06,
+      "loss": 0.4118,
+      "step": 18745
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.807076736741935e-06,
+      "loss": 0.4046,
+      "step": 18746
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.80649998558122e-06,
+      "loss": 0.5619,
+      "step": 18747
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.80592323442051e-06,
+      "loss": 0.5332,
+      "step": 18748
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.8053464832598e-06,
+      "loss": 0.4017,
+      "step": 18749
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.804769732099087e-06,
+      "loss": 0.3446,
+      "step": 18750
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.804192980938375e-06,
+      "loss": 0.4992,
+      "step": 18751
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.803616229777663e-06,
+      "loss": 0.4265,
+      "step": 18752
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.803039478616952e-06,
+      "loss": 0.4466,
+      "step": 18753
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.80246272745624e-06,
+      "loss": 0.4496,
+      "step": 18754
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.801885976295527e-06,
+      "loss": 0.485,
+      "step": 18755
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.801309225134817e-06,
+      "loss": 0.5154,
+      "step": 18756
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.800732473974104e-06,
+      "loss": 0.4708,
+      "step": 18757
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.800155722813392e-06,
+      "loss": 0.5496,
+      "step": 18758
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.799578971652681e-06,
+      "loss": 0.4494,
+      "step": 18759
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.799002220491969e-06,
+      "loss": 0.4507,
+      "step": 18760
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.798425469331257e-06,
+      "loss": 0.3549,
+      "step": 18761
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.797848718170546e-06,
+      "loss": 0.3973,
+      "step": 18762
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.797271967009836e-06,
+      "loss": 0.4455,
+      "step": 18763
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.796695215849122e-06,
+      "loss": 0.4545,
+      "step": 18764
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.796118464688411e-06,
+      "loss": 0.4973,
+      "step": 18765
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.7955417135277e-06,
+      "loss": 0.3818,
+      "step": 18766
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794964962366988e-06,
+      "loss": 0.5265,
+      "step": 18767
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.794388211206276e-06,
+      "loss": 0.3924,
+      "step": 18768
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.793811460045563e-06,
+      "loss": 0.417,
+      "step": 18769
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.793234708884853e-06,
+      "loss": 0.4808,
+      "step": 18770
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.79265795772414e-06,
+      "loss": 0.4311,
+      "step": 18771
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.792081206563428e-06,
+      "loss": 0.4672,
+      "step": 18772
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.791504455402718e-06,
+      "loss": 0.4652,
+      "step": 18773
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.790927704242005e-06,
+      "loss": 0.4907,
+      "step": 18774
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.790350953081293e-06,
+      "loss": 0.4323,
+      "step": 18775
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.789774201920582e-06,
+      "loss": 0.4306,
+      "step": 18776
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.789197450759872e-06,
+      "loss": 0.4894,
+      "step": 18777
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.788620699599158e-06,
+      "loss": 0.4193,
+      "step": 18778
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.788043948438447e-06,
+      "loss": 0.4133,
+      "step": 18779
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.787467197277736e-06,
+      "loss": 0.5153,
+      "step": 18780
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.786890446117024e-06,
+      "loss": 0.5494,
+      "step": 18781
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.786313694956312e-06,
+      "loss": 0.367,
+      "step": 18782
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.785736943795601e-06,
+      "loss": 0.4449,
+      "step": 18783
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.785160192634889e-06,
+      "loss": 0.4408,
+      "step": 18784
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.784583441474177e-06,
+      "loss": 0.5272,
+      "step": 18785
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.784006690313464e-06,
+      "loss": 0.4993,
+      "step": 18786
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.783429939152754e-06,
+      "loss": 0.4985,
+      "step": 18787
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.782853187992041e-06,
+      "loss": 0.4899,
+      "step": 18788
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.782276436831329e-06,
+      "loss": 0.6076,
+      "step": 18789
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.781699685670618e-06,
+      "loss": 0.3431,
+      "step": 18790
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.781122934509906e-06,
+      "loss": 0.4475,
+      "step": 18791
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.780546183349194e-06,
+      "loss": 0.4154,
+      "step": 18792
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.779969432188483e-06,
+      "loss": 0.4409,
+      "step": 18793
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.779392681027772e-06,
+      "loss": 0.5379,
+      "step": 18794
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77881592986706e-06,
+      "loss": 0.4739,
+      "step": 18795
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.778239178706348e-06,
+      "loss": 0.3721,
+      "step": 18796
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.777662427545637e-06,
+      "loss": 0.4395,
+      "step": 18797
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.777085676384925e-06,
+      "loss": 0.4066,
+      "step": 18798
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.776508925224213e-06,
+      "loss": 0.5303,
+      "step": 18799
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.775932174063502e-06,
+      "loss": 0.5107,
+      "step": 18800
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77535542290279e-06,
+      "loss": 0.4897,
+      "step": 18801
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.774778671742077e-06,
+      "loss": 0.3778,
+      "step": 18802
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.774201920581367e-06,
+      "loss": 0.5734,
+      "step": 18803
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.773625169420654e-06,
+      "loss": 0.4442,
+      "step": 18804
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.773048418259942e-06,
+      "loss": 0.4911,
+      "step": 18805
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.77247166709923e-06,
+      "loss": 0.3887,
+      "step": 18806
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.771894915938519e-06,
+      "loss": 0.4972,
+      "step": 18807
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.771318164777809e-06,
+      "loss": 0.377,
+      "step": 18808
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.770741413617095e-06,
+      "loss": 0.4109,
+      "step": 18809
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.770164662456384e-06,
+      "loss": 0.4638,
+      "step": 18810
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.769587911295673e-06,
+      "loss": 0.4993,
+      "step": 18811
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.769011160134961e-06,
+      "loss": 0.3269,
+      "step": 18812
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.768434408974249e-06,
+      "loss": 0.3534,
+      "step": 18813
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.767857657813538e-06,
+      "loss": 0.4381,
+      "step": 18814
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.767280906652826e-06,
+      "loss": 0.4323,
+      "step": 18815
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.766704155492113e-06,
+      "loss": 0.5238,
+      "step": 18816
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.766127404331403e-06,
+      "loss": 0.4749,
+      "step": 18817
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.76555065317069e-06,
+      "loss": 0.4326,
+      "step": 18818
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.764973902009978e-06,
+      "loss": 0.456,
+      "step": 18819
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.764397150849268e-06,
+      "loss": 0.4244,
+      "step": 18820
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.763820399688555e-06,
+      "loss": 0.4538,
+      "step": 18821
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.763243648527843e-06,
+      "loss": 0.4307,
+      "step": 18822
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.76266689736713e-06,
+      "loss": 0.3898,
+      "step": 18823
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.76209014620642e-06,
+      "loss": 0.4496,
+      "step": 18824
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.76151339504571e-06,
+      "loss": 0.3344,
+      "step": 18825
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.760936643884997e-06,
+      "loss": 0.4241,
+      "step": 18826
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.760359892724285e-06,
+      "loss": 0.5967,
+      "step": 18827
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.759783141563574e-06,
+      "loss": 0.3991,
+      "step": 18828
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.759206390402862e-06,
+      "loss": 0.4312,
+      "step": 18829
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.75862963924215e-06,
+      "loss": 0.5476,
+      "step": 18830
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.758052888081439e-06,
+      "loss": 0.5638,
+      "step": 18831
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.757476136920727e-06,
+      "loss": 0.6939,
+      "step": 18832
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.756899385760014e-06,
+      "loss": 0.4588,
+      "step": 18833
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.756322634599304e-06,
+      "loss": 0.469,
+      "step": 18834
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.755745883438591e-06,
+      "loss": 0.3465,
+      "step": 18835
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.755169132277879e-06,
+      "loss": 0.4767,
+      "step": 18836
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.754592381117168e-06,
+      "loss": 0.439,
+      "step": 18837
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.754015629956456e-06,
+      "loss": 0.4612,
+      "step": 18838
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.753438878795745e-06,
+      "loss": 0.4811,
+      "step": 18839
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.752862127635031e-06,
+      "loss": 0.4351,
+      "step": 18840
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.75228537647432e-06,
+      "loss": 0.4239,
+      "step": 18841
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.75170862531361e-06,
+      "loss": 0.4449,
+      "step": 18842
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.751131874152898e-06,
+      "loss": 0.3336,
+      "step": 18843
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.750555122992186e-06,
+      "loss": 0.4487,
+      "step": 18844
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.749978371831475e-06,
+      "loss": 0.4173,
+      "step": 18845
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.749401620670763e-06,
+      "loss": 0.4577,
+      "step": 18846
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.74882486951005e-06,
+      "loss": 0.6976,
+      "step": 18847
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.74824811834934e-06,
+      "loss": 0.2402,
+      "step": 18848
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.747671367188627e-06,
+      "loss": 0.4911,
+      "step": 18849
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.747094616027915e-06,
+      "loss": 0.4474,
+      "step": 18850
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.746517864867204e-06,
+      "loss": 0.4017,
+      "step": 18851
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.745941113706492e-06,
+      "loss": 0.3115,
+      "step": 18852
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.74536436254578e-06,
+      "loss": 0.3676,
+      "step": 18853
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.744787611385069e-06,
+      "loss": 0.5285,
+      "step": 18854
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.744210860224357e-06,
+      "loss": 0.5517,
+      "step": 18855
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.743634109063646e-06,
+      "loss": 0.4403,
+      "step": 18856
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.743057357902934e-06,
+      "loss": 0.4925,
+      "step": 18857
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 9.742480606742222e-06,
+      "loss": 0.3718,
+      "step": 18858
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.741903855581511e-06,
+      "loss": 0.4805,
+      "step": 18859
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.741327104420799e-06,
+      "loss": 0.447,
+      "step": 18860
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.740750353260086e-06,
+      "loss": 0.3859,
+      "step": 18861
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.740173602099376e-06,
+      "loss": 0.382,
+      "step": 18862
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.739596850938663e-06,
+      "loss": 0.3659,
+      "step": 18863
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.739020099777951e-06,
+      "loss": 0.5745,
+      "step": 18864
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.73844334861724e-06,
+      "loss": 0.4122,
+      "step": 18865
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.737866597456528e-06,
+      "loss": 0.3879,
+      "step": 18866
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.737289846295816e-06,
+      "loss": 0.5605,
+      "step": 18867
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.736713095135105e-06,
+      "loss": 0.5627,
+      "step": 18868
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.736136343974393e-06,
+      "loss": 0.4816,
+      "step": 18869
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.735559592813682e-06,
+      "loss": 0.5332,
+      "step": 18870
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.73498284165297e-06,
+      "loss": 0.6084,
+      "step": 18871
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.734406090492258e-06,
+      "loss": 0.3702,
+      "step": 18872
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.733829339331547e-06,
+      "loss": 0.4136,
+      "step": 18873
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.733252588170835e-06,
+      "loss": 0.431,
+      "step": 18874
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732675837010122e-06,
+      "loss": 0.4633,
+      "step": 18875
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.732099085849412e-06,
+      "loss": 0.4921,
+      "step": 18876
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.7315223346887e-06,
+      "loss": 0.3752,
+      "step": 18877
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.730945583527987e-06,
+      "loss": 0.4436,
+      "step": 18878
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.730368832367277e-06,
+      "loss": 0.3693,
+      "step": 18879
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.729792081206564e-06,
+      "loss": 0.503,
+      "step": 18880
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.729215330045852e-06,
+      "loss": 0.4269,
+      "step": 18881
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.728638578885141e-06,
+      "loss": 0.4653,
+      "step": 18882
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.728061827724429e-06,
+      "loss": 0.5006,
+      "step": 18883
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.727485076563717e-06,
+      "loss": 0.4889,
+      "step": 18884
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.726908325403006e-06,
+      "loss": 0.4329,
+      "step": 18885
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.726331574242294e-06,
+      "loss": 0.3509,
+      "step": 18886
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.725754823081583e-06,
+      "loss": 0.3418,
+      "step": 18887
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.72517807192087e-06,
+      "loss": 0.3696,
+      "step": 18888
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.724601320760158e-06,
+      "loss": 0.4795,
+      "step": 18889
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.724024569599448e-06,
+      "loss": 0.3551,
+      "step": 18890
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.723447818438736e-06,
+      "loss": 0.4721,
+      "step": 18891
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.722871067278023e-06,
+      "loss": 0.4923,
+      "step": 18892
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.722294316117313e-06,
+      "loss": 0.3898,
+      "step": 18893
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.7217175649566e-06,
+      "loss": 0.4902,
+      "step": 18894
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.721140813795888e-06,
+      "loss": 0.4488,
+      "step": 18895
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.720564062635177e-06,
+      "loss": 0.4071,
+      "step": 18896
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.719987311474465e-06,
+      "loss": 0.4732,
+      "step": 18897
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.719410560313753e-06,
+      "loss": 0.5539,
+      "step": 18898
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.718833809153042e-06,
+      "loss": 0.4335,
+      "step": 18899
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.71825705799233e-06,
+      "loss": 0.5874,
+      "step": 18900
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.717680306831619e-06,
+      "loss": 0.4243,
+      "step": 18901
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.717103555670907e-06,
+      "loss": 0.3878,
+      "step": 18902
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.716526804510195e-06,
+      "loss": 0.347,
+      "step": 18903
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.715950053349484e-06,
+      "loss": 0.5538,
+      "step": 18904
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.715373302188772e-06,
+      "loss": 0.5411,
+      "step": 18905
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.71479655102806e-06,
+      "loss": 0.321,
+      "step": 18906
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.714219799867349e-06,
+      "loss": 0.4802,
+      "step": 18907
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.713643048706636e-06,
+      "loss": 0.4292,
+      "step": 18908
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.713066297545924e-06,
+      "loss": 0.4663,
+      "step": 18909
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.712489546385213e-06,
+      "loss": 0.5043,
+      "step": 18910
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711912795224501e-06,
+      "loss": 0.3686,
+      "step": 18911
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.711336044063789e-06,
+      "loss": 0.3983,
+      "step": 18912
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.710759292903078e-06,
+      "loss": 0.3777,
+      "step": 18913
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.710182541742366e-06,
+      "loss": 0.3521,
+      "step": 18914
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.709605790581654e-06,
+      "loss": 0.4941,
+      "step": 18915
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.709029039420943e-06,
+      "loss": 0.3591,
+      "step": 18916
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.70845228826023e-06,
+      "loss": 0.4116,
+      "step": 18917
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.70787553709952e-06,
+      "loss": 0.3852,
+      "step": 18918
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.707298785938808e-06,
+      "loss": 0.4156,
+      "step": 18919
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.706722034778095e-06,
+      "loss": 0.4947,
+      "step": 18920
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.706145283617385e-06,
+      "loss": 0.5112,
+      "step": 18921
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.705568532456672e-06,
+      "loss": 0.474,
+      "step": 18922
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.70499178129596e-06,
+      "loss": 0.4546,
+      "step": 18923
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.70441503013525e-06,
+      "loss": 0.548,
+      "step": 18924
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.703838278974537e-06,
+      "loss": 0.3687,
+      "step": 18925
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.703261527813825e-06,
+      "loss": 0.4827,
+      "step": 18926
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.702684776653114e-06,
+      "loss": 0.5354,
+      "step": 18927
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.702108025492402e-06,
+      "loss": 0.2746,
+      "step": 18928
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.70153127433169e-06,
+      "loss": 0.51,
+      "step": 18929
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.700954523170979e-06,
+      "loss": 0.5465,
+      "step": 18930
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.700377772010267e-06,
+      "loss": 0.4127,
+      "step": 18931
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.699801020849556e-06,
+      "loss": 0.3398,
+      "step": 18932
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.699224269688844e-06,
+      "loss": 0.4483,
+      "step": 18933
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.698647518528131e-06,
+      "loss": 0.4827,
+      "step": 18934
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.69807076736742e-06,
+      "loss": 0.415,
+      "step": 18935
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.697494016206708e-06,
+      "loss": 0.4159,
+      "step": 18936
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.696917265045996e-06,
+      "loss": 0.4523,
+      "step": 18937
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.696340513885286e-06,
+      "loss": 0.4091,
+      "step": 18938
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.695763762724573e-06,
+      "loss": 0.4664,
+      "step": 18939
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.695187011563861e-06,
+      "loss": 0.4904,
+      "step": 18940
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.69461026040315e-06,
+      "loss": 0.4008,
+      "step": 18941
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.694033509242438e-06,
+      "loss": 0.5345,
+      "step": 18942
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.693456758081726e-06,
+      "loss": 0.3781,
+      "step": 18943
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.692880006921015e-06,
+      "loss": 0.5049,
+      "step": 18944
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.692303255760303e-06,
+      "loss": 0.3309,
+      "step": 18945
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.69172650459959e-06,
+      "loss": 0.4424,
+      "step": 18946
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.69114975343888e-06,
+      "loss": 0.3347,
+      "step": 18947
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.690573002278167e-06,
+      "loss": 0.3768,
+      "step": 18948
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.689996251117457e-06,
+      "loss": 0.4244,
+      "step": 18949
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.689419499956745e-06,
+      "loss": 0.614,
+      "step": 18950
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.688842748796032e-06,
+      "loss": 0.4812,
+      "step": 18951
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.688265997635322e-06,
+      "loss": 0.4638,
+      "step": 18952
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.68768924647461e-06,
+      "loss": 0.5305,
+      "step": 18953
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.687112495313897e-06,
+      "loss": 0.4765,
+      "step": 18954
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.686535744153186e-06,
+      "loss": 0.4195,
+      "step": 18955
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.685958992992474e-06,
+      "loss": 0.4233,
+      "step": 18956
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.685382241831762e-06,
+      "loss": 0.4987,
+      "step": 18957
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.684805490671051e-06,
+      "loss": 0.4191,
+      "step": 18958
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.684228739510339e-06,
+      "loss": 0.4641,
+      "step": 18959
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.683651988349626e-06,
+      "loss": 0.4653,
+      "step": 18960
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.683075237188916e-06,
+      "loss": 0.4775,
+      "step": 18961
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.682498486028204e-06,
+      "loss": 0.4782,
+      "step": 18962
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.681921734867493e-06,
+      "loss": 0.4098,
+      "step": 18963
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.68134498370678e-06,
+      "loss": 0.4246,
+      "step": 18964
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.680768232546068e-06,
+      "loss": 0.3969,
+      "step": 18965
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.680191481385358e-06,
+      "loss": 0.5199,
+      "step": 18966
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.679614730224645e-06,
+      "loss": 0.4561,
+      "step": 18967
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.679037979063933e-06,
+      "loss": 0.5383,
+      "step": 18968
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.678461227903222e-06,
+      "loss": 0.5516,
+      "step": 18969
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.67788447674251e-06,
+      "loss": 0.465,
+      "step": 18970
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.677307725581798e-06,
+      "loss": 0.3656,
+      "step": 18971
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.676730974421087e-06,
+      "loss": 0.4469,
+      "step": 18972
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.676154223260375e-06,
+      "loss": 0.3495,
+      "step": 18973
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.675577472099663e-06,
+      "loss": 0.3484,
+      "step": 18974
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.675000720938952e-06,
+      "loss": 0.359,
+      "step": 18975
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.674423969778241e-06,
+      "loss": 0.554,
+      "step": 18976
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.673847218617527e-06,
+      "loss": 0.3983,
+      "step": 18977
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.673270467456817e-06,
+      "loss": 0.3426,
+      "step": 18978
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.672693716296104e-06,
+      "loss": 0.5061,
+      "step": 18979
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.672116965135394e-06,
+      "loss": 0.4113,
+      "step": 18980
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.671540213974681e-06,
+      "loss": 0.4109,
+      "step": 18981
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670963462813969e-06,
+      "loss": 0.4345,
+      "step": 18982
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.670386711653258e-06,
+      "loss": 0.5126,
+      "step": 18983
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.669809960492546e-06,
+      "loss": 0.368,
+      "step": 18984
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.669233209331834e-06,
+      "loss": 0.4165,
+      "step": 18985
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.668656458171123e-06,
+      "loss": 0.4531,
+      "step": 18986
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.668079707010411e-06,
+      "loss": 0.3368,
+      "step": 18987
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.667502955849699e-06,
+      "loss": 0.3974,
+      "step": 18988
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.666926204688988e-06,
+      "loss": 0.3519,
+      "step": 18989
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.666349453528276e-06,
+      "loss": 0.4879,
+      "step": 18990
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.665772702367563e-06,
+      "loss": 0.4631,
+      "step": 18991
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.665195951206853e-06,
+      "loss": 0.4251,
+      "step": 18992
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.664619200046142e-06,
+      "loss": 0.4047,
+      "step": 18993
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.66404244888543e-06,
+      "loss": 0.5075,
+      "step": 18994
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.663465697724717e-06,
+      "loss": 0.5203,
+      "step": 18995
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.662888946564005e-06,
+      "loss": 0.3088,
+      "step": 18996
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.662312195403295e-06,
+      "loss": 0.3304,
+      "step": 18997
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.661735444242582e-06,
+      "loss": 0.4137,
+      "step": 18998
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.66115869308187e-06,
+      "loss": 0.369,
+      "step": 18999
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.66058194192116e-06,
+      "loss": 0.4741,
+      "step": 19000
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.660005190760447e-06,
+      "loss": 0.465,
+      "step": 19001
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.659428439599735e-06,
+      "loss": 0.3718,
+      "step": 19002
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.658851688439024e-06,
+      "loss": 0.3749,
+      "step": 19003
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.658274937278312e-06,
+      "loss": 0.3575,
+      "step": 19004
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.6576981861176e-06,
+      "loss": 0.4719,
+      "step": 19005
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.657121434956889e-06,
+      "loss": 0.4005,
+      "step": 19006
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.656544683796178e-06,
+      "loss": 0.36,
+      "step": 19007
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.655967932635464e-06,
+      "loss": 0.3943,
+      "step": 19008
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.655391181474754e-06,
+      "loss": 0.5419,
+      "step": 19009
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.654814430314043e-06,
+      "loss": 0.4541,
+      "step": 19010
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.65423767915333e-06,
+      "loss": 0.5251,
+      "step": 19011
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.653660927992618e-06,
+      "loss": 0.3593,
+      "step": 19012
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.653084176831906e-06,
+      "loss": 0.4901,
+      "step": 19013
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.652507425671195e-06,
+      "loss": 0.512,
+      "step": 19014
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.651930674510483e-06,
+      "loss": 0.4173,
+      "step": 19015
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.65135392334977e-06,
+      "loss": 0.5545,
+      "step": 19016
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.65077717218906e-06,
+      "loss": 0.3749,
+      "step": 19017
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.650200421028348e-06,
+      "loss": 0.4314,
+      "step": 19018
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.649623669867635e-06,
+      "loss": 0.4708,
+      "step": 19019
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.649046918706925e-06,
+      "loss": 0.4272,
+      "step": 19020
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.648470167546214e-06,
+      "loss": 0.4394,
+      "step": 19021
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.6478934163855e-06,
+      "loss": 0.377,
+      "step": 19022
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.64731666522479e-06,
+      "loss": 0.451,
+      "step": 19023
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.646739914064079e-06,
+      "loss": 0.4485,
+      "step": 19024
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.646163162903367e-06,
+      "loss": 0.3251,
+      "step": 19025
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.645586411742654e-06,
+      "loss": 0.43,
+      "step": 19026
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.645009660581944e-06,
+      "loss": 0.4178,
+      "step": 19027
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.644432909421231e-06,
+      "loss": 0.3675,
+      "step": 19028
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.643856158260519e-06,
+      "loss": 0.4896,
+      "step": 19029
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.643279407099807e-06,
+      "loss": 0.4527,
+      "step": 19030
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.642702655939096e-06,
+      "loss": 0.3954,
+      "step": 19031
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.642125904778384e-06,
+      "loss": 0.4242,
+      "step": 19032
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.641549153617671e-06,
+      "loss": 0.3825,
+      "step": 19033
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.640972402456961e-06,
+      "loss": 0.3657,
+      "step": 19034
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.640395651296249e-06,
+      "loss": 0.4685,
+      "step": 19035
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 9.639818900135536e-06,
+      "loss": 0.4706,
+      "step": 19036
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.639242148974826e-06,
+      "loss": 0.4077,
+      "step": 19037
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.638665397814115e-06,
+      "loss": 0.5107,
+      "step": 19038
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.638088646653401e-06,
+      "loss": 0.3984,
+      "step": 19039
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.63751189549269e-06,
+      "loss": 0.5667,
+      "step": 19040
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.63693514433198e-06,
+      "loss": 0.6,
+      "step": 19041
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.636358393171267e-06,
+      "loss": 0.4593,
+      "step": 19042
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.635781642010555e-06,
+      "loss": 0.4521,
+      "step": 19043
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.635204890849845e-06,
+      "loss": 0.4162,
+      "step": 19044
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.634628139689132e-06,
+      "loss": 0.4423,
+      "step": 19045
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.63405138852842e-06,
+      "loss": 0.409,
+      "step": 19046
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.633474637367708e-06,
+      "loss": 0.3331,
+      "step": 19047
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.632897886206997e-06,
+      "loss": 0.4266,
+      "step": 19048
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.632321135046285e-06,
+      "loss": 0.3489,
+      "step": 19049
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.631744383885572e-06,
+      "loss": 0.4621,
+      "step": 19050
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.631167632724862e-06,
+      "loss": 0.5064,
+      "step": 19051
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.630590881564151e-06,
+      "loss": 0.3702,
+      "step": 19052
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.630014130403437e-06,
+      "loss": 0.4077,
+      "step": 19053
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.629437379242726e-06,
+      "loss": 0.4709,
+      "step": 19054
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.628860628082016e-06,
+      "loss": 0.3014,
+      "step": 19055
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.628283876921304e-06,
+      "loss": 0.5316,
+      "step": 19056
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.627707125760591e-06,
+      "loss": 0.3694,
+      "step": 19057
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62713037459988e-06,
+      "loss": 0.4287,
+      "step": 19058
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.626553623439168e-06,
+      "loss": 0.4744,
+      "step": 19059
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.625976872278456e-06,
+      "loss": 0.3356,
+      "step": 19060
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.625400121117745e-06,
+      "loss": 0.4263,
+      "step": 19061
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.624823369957033e-06,
+      "loss": 0.393,
+      "step": 19062
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62424661879632e-06,
+      "loss": 0.439,
+      "step": 19063
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.62366986763561e-06,
+      "loss": 0.4575,
+      "step": 19064
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.623093116474898e-06,
+      "loss": 0.427,
+      "step": 19065
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.622516365314185e-06,
+      "loss": 0.4861,
+      "step": 19066
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.621939614153473e-06,
+      "loss": 0.3958,
+      "step": 19067
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.621362862992762e-06,
+      "loss": 0.4412,
+      "step": 19068
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.620786111832052e-06,
+      "loss": 0.5483,
+      "step": 19069
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.620209360671338e-06,
+      "loss": 0.5809,
+      "step": 19070
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.619632609510627e-06,
+      "loss": 0.483,
+      "step": 19071
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.619055858349917e-06,
+      "loss": 0.4163,
+      "step": 19072
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.618479107189204e-06,
+      "loss": 0.5332,
+      "step": 19073
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.617902356028492e-06,
+      "loss": 0.3959,
+      "step": 19074
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.617325604867781e-06,
+      "loss": 0.4415,
+      "step": 19075
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.616748853707069e-06,
+      "loss": 0.3493,
+      "step": 19076
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.616172102546357e-06,
+      "loss": 0.6091,
+      "step": 19077
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.615595351385646e-06,
+      "loss": 0.4526,
+      "step": 19078
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.615018600224934e-06,
+      "loss": 0.4952,
+      "step": 19079
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.614441849064221e-06,
+      "loss": 0.3758,
+      "step": 19080
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.613865097903511e-06,
+      "loss": 0.4395,
+      "step": 19081
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.613288346742799e-06,
+      "loss": 0.36,
+      "step": 19082
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.612711595582088e-06,
+      "loss": 0.4493,
+      "step": 19083
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.612134844421374e-06,
+      "loss": 0.4465,
+      "step": 19084
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.611558093260663e-06,
+      "loss": 0.455,
+      "step": 19085
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.610981342099953e-06,
+      "loss": 0.4371,
+      "step": 19086
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.61040459093924e-06,
+      "loss": 0.3009,
+      "step": 19087
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609827839778528e-06,
+      "loss": 0.58,
+      "step": 19088
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.609251088617817e-06,
+      "loss": 0.3951,
+      "step": 19089
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.608674337457105e-06,
+      "loss": 0.5869,
+      "step": 19090
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.608097586296393e-06,
+      "loss": 0.5778,
+      "step": 19091
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.607520835135682e-06,
+      "loss": 0.6078,
+      "step": 19092
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.60694408397497e-06,
+      "loss": 0.38,
+      "step": 19093
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.606367332814258e-06,
+      "loss": 0.435,
+      "step": 19094
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.605790581653547e-06,
+      "loss": 0.4292,
+      "step": 19095
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.605213830492835e-06,
+      "loss": 0.387,
+      "step": 19096
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.604637079332122e-06,
+      "loss": 0.4274,
+      "step": 19097
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.604060328171412e-06,
+      "loss": 0.5705,
+      "step": 19098
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.6034835770107e-06,
+      "loss": 0.3191,
+      "step": 19099
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.602906825849989e-06,
+      "loss": 0.3936,
+      "step": 19100
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.602330074689275e-06,
+      "loss": 0.4379,
+      "step": 19101
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.601753323528564e-06,
+      "loss": 0.4411,
+      "step": 19102
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.601176572367854e-06,
+      "loss": 0.4867,
+      "step": 19103
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.600599821207141e-06,
+      "loss": 0.4255,
+      "step": 19104
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.600023070046429e-06,
+      "loss": 0.3837,
+      "step": 19105
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.599446318885718e-06,
+      "loss": 0.3396,
+      "step": 19106
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.598869567725006e-06,
+      "loss": 0.5523,
+      "step": 19107
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.598292816564294e-06,
+      "loss": 0.4132,
+      "step": 19108
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.597716065403583e-06,
+      "loss": 0.4779,
+      "step": 19109
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.59713931424287e-06,
+      "loss": 0.5171,
+      "step": 19110
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.596562563082158e-06,
+      "loss": 0.4254,
+      "step": 19111
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.595985811921448e-06,
+      "loss": 0.565,
+      "step": 19112
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.595409060760735e-06,
+      "loss": 0.409,
+      "step": 19113
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.594832309600025e-06,
+      "loss": 0.5756,
+      "step": 19114
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.594255558439312e-06,
+      "loss": 0.2883,
+      "step": 19115
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.5936788072786e-06,
+      "loss": 0.479,
+      "step": 19116
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.59310205611789e-06,
+      "loss": 0.6024,
+      "step": 19117
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.592525304957177e-06,
+      "loss": 0.521,
+      "step": 19118
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.591948553796465e-06,
+      "loss": 0.4696,
+      "step": 19119
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.591371802635754e-06,
+      "loss": 0.5794,
+      "step": 19120
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.590795051475042e-06,
+      "loss": 0.502,
+      "step": 19121
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.59021830031433e-06,
+      "loss": 0.4552,
+      "step": 19122
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.589641549153619e-06,
+      "loss": 0.4734,
+      "step": 19123
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.589064797992907e-06,
+      "loss": 0.4607,
+      "step": 19124
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.588488046832194e-06,
+      "loss": 0.4629,
+      "step": 19125
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.587911295671484e-06,
+      "loss": 0.4778,
+      "step": 19126
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.587334544510771e-06,
+      "loss": 0.3936,
+      "step": 19127
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.58675779335006e-06,
+      "loss": 0.5509,
+      "step": 19128
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.586181042189349e-06,
+      "loss": 0.4745,
+      "step": 19129
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.585604291028636e-06,
+      "loss": 0.3671,
+      "step": 19130
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.585027539867926e-06,
+      "loss": 0.4714,
+      "step": 19131
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.584450788707213e-06,
+      "loss": 0.5045,
+      "step": 19132
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.583874037546501e-06,
+      "loss": 0.3084,
+      "step": 19133
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.58329728638579e-06,
+      "loss": 0.5432,
+      "step": 19134
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.582720535225078e-06,
+      "loss": 0.4066,
+      "step": 19135
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.582143784064366e-06,
+      "loss": 0.4743,
+      "step": 19136
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.581567032903655e-06,
+      "loss": 0.556,
+      "step": 19137
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.580990281742943e-06,
+      "loss": 0.4599,
+      "step": 19138
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.58041353058223e-06,
+      "loss": 0.4569,
+      "step": 19139
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.57983677942152e-06,
+      "loss": 0.2991,
+      "step": 19140
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.579260028260808e-06,
+      "loss": 0.4967,
+      "step": 19141
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.578683277100095e-06,
+      "loss": 0.4376,
+      "step": 19142
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.578106525939385e-06,
+      "loss": 0.4356,
+      "step": 19143
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.577529774778672e-06,
+      "loss": 0.4314,
+      "step": 19144
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.576953023617962e-06,
+      "loss": 0.4113,
+      "step": 19145
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.57637627245725e-06,
+      "loss": 0.3656,
+      "step": 19146
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.575799521296537e-06,
+      "loss": 0.4665,
+      "step": 19147
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.575222770135826e-06,
+      "loss": 0.4949,
+      "step": 19148
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.574646018975114e-06,
+      "loss": 0.5013,
+      "step": 19149
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.574069267814402e-06,
+      "loss": 0.4674,
+      "step": 19150
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.573492516653691e-06,
+      "loss": 0.3597,
+      "step": 19151
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.572915765492979e-06,
+      "loss": 0.5359,
+      "step": 19152
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.572339014332267e-06,
+      "loss": 0.4013,
+      "step": 19153
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.571762263171556e-06,
+      "loss": 0.4468,
+      "step": 19154
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.571185512010844e-06,
+      "loss": 0.4191,
+      "step": 19155
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.570608760850131e-06,
+      "loss": 0.4053,
+      "step": 19156
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.57003200968942e-06,
+      "loss": 0.4741,
+      "step": 19157
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.569455258528708e-06,
+      "loss": 0.453,
+      "step": 19158
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.568878507367996e-06,
+      "loss": 0.3392,
+      "step": 19159
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.568301756207285e-06,
+      "loss": 0.4444,
+      "step": 19160
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567725005046573e-06,
+      "loss": 0.4721,
+      "step": 19161
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.567148253885862e-06,
+      "loss": 0.5226,
+      "step": 19162
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.56657150272515e-06,
+      "loss": 0.4091,
+      "step": 19163
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.565994751564438e-06,
+      "loss": 0.4859,
+      "step": 19164
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.565418000403727e-06,
+      "loss": 0.4707,
+      "step": 19165
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.564841249243015e-06,
+      "loss": 0.4181,
+      "step": 19166
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.564264498082303e-06,
+      "loss": 0.3615,
+      "step": 19167
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.563687746921592e-06,
+      "loss": 0.3981,
+      "step": 19168
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.56311099576088e-06,
+      "loss": 0.4318,
+      "step": 19169
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.562534244600167e-06,
+      "loss": 0.3285,
+      "step": 19170
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.561957493439457e-06,
+      "loss": 0.3997,
+      "step": 19171
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.561380742278744e-06,
+      "loss": 0.4045,
+      "step": 19172
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.560803991118032e-06,
+      "loss": 0.382,
+      "step": 19173
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.560227239957321e-06,
+      "loss": 0.4817,
+      "step": 19174
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.55965048879661e-06,
+      "loss": 0.5232,
+      "step": 19175
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.559073737635899e-06,
+      "loss": 0.4388,
+      "step": 19176
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.558496986475186e-06,
+      "loss": 0.424,
+      "step": 19177
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.557920235314474e-06,
+      "loss": 0.386,
+      "step": 19178
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.557343484153763e-06,
+      "loss": 0.4287,
+      "step": 19179
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.556766732993051e-06,
+      "loss": 0.462,
+      "step": 19180
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.556189981832339e-06,
+      "loss": 0.5362,
+      "step": 19181
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.555613230671628e-06,
+      "loss": 0.4279,
+      "step": 19182
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.555036479510916e-06,
+      "loss": 0.5174,
+      "step": 19183
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.554459728350203e-06,
+      "loss": 0.4785,
+      "step": 19184
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.553882977189493e-06,
+      "loss": 0.41,
+      "step": 19185
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.55330622602878e-06,
+      "loss": 0.3849,
+      "step": 19186
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.552729474868068e-06,
+      "loss": 0.3795,
+      "step": 19187
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.552152723707358e-06,
+      "loss": 0.4626,
+      "step": 19188
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.551575972546645e-06,
+      "loss": 0.5443,
+      "step": 19189
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.550999221385933e-06,
+      "loss": 0.5377,
+      "step": 19190
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.550422470225222e-06,
+      "loss": 0.5216,
+      "step": 19191
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.54984571906451e-06,
+      "loss": 0.4856,
+      "step": 19192
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.5492689679038e-06,
+      "loss": 0.4054,
+      "step": 19193
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.548692216743087e-06,
+      "loss": 0.3199,
+      "step": 19194
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.548115465582375e-06,
+      "loss": 0.4735,
+      "step": 19195
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.547538714421664e-06,
+      "loss": 0.4942,
+      "step": 19196
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.546961963260952e-06,
+      "loss": 0.3776,
+      "step": 19197
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.54638521210024e-06,
+      "loss": 0.521,
+      "step": 19198
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.545808460939529e-06,
+      "loss": 0.4417,
+      "step": 19199
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.545231709778817e-06,
+      "loss": 0.4681,
+      "step": 19200
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.544654958618104e-06,
+      "loss": 0.4203,
+      "step": 19201
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.544078207457394e-06,
+      "loss": 0.4754,
+      "step": 19202
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.543501456296681e-06,
+      "loss": 0.4281,
+      "step": 19203
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.542924705135969e-06,
+      "loss": 0.4648,
+      "step": 19204
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.542347953975258e-06,
+      "loss": 0.439,
+      "step": 19205
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.541771202814546e-06,
+      "loss": 0.4204,
+      "step": 19206
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.541194451653835e-06,
+      "loss": 0.5179,
+      "step": 19207
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.540617700493123e-06,
+      "loss": 0.5376,
+      "step": 19208
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.54004094933241e-06,
+      "loss": 0.5422,
+      "step": 19209
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.5394641981717e-06,
+      "loss": 0.4966,
+      "step": 19210
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.538887447010988e-06,
+      "loss": 0.4015,
+      "step": 19211
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.538310695850276e-06,
+      "loss": 0.4463,
+      "step": 19212
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.537733944689565e-06,
+      "loss": 0.4556,
+      "step": 19213
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.537157193528853e-06,
+      "loss": 0.4175,
+      "step": 19214
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 9.53658044236814e-06,
+      "loss": 0.3782,
+      "step": 19215
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.53600369120743e-06,
+      "loss": 0.4878,
+      "step": 19216
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.535426940046717e-06,
+      "loss": 0.3397,
+      "step": 19217
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.534850188886005e-06,
+      "loss": 0.4466,
+      "step": 19218
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.534273437725294e-06,
+      "loss": 0.4674,
+      "step": 19219
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.533696686564584e-06,
+      "loss": 0.5704,
+      "step": 19220
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.53311993540387e-06,
+      "loss": 0.4442,
+      "step": 19221
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.53254318424316e-06,
+      "loss": 0.505,
+      "step": 19222
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.531966433082447e-06,
+      "loss": 0.4559,
+      "step": 19223
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.531389681921736e-06,
+      "loss": 0.4543,
+      "step": 19224
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.530812930761024e-06,
+      "loss": 0.4164,
+      "step": 19225
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.530236179600312e-06,
+      "loss": 0.4113,
+      "step": 19226
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.529659428439601e-06,
+      "loss": 0.4095,
+      "step": 19227
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.529082677278889e-06,
+      "loss": 0.4429,
+      "step": 19228
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.528505926118176e-06,
+      "loss": 0.6145,
+      "step": 19229
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.527929174957466e-06,
+      "loss": 0.4266,
+      "step": 19230
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.527352423796753e-06,
+      "loss": 0.3981,
+      "step": 19231
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.526775672636041e-06,
+      "loss": 0.5054,
+      "step": 19232
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.52619892147533e-06,
+      "loss": 0.4673,
+      "step": 19233
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.525622170314618e-06,
+      "loss": 0.5545,
+      "step": 19234
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.525045419153906e-06,
+      "loss": 0.3618,
+      "step": 19235
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.524468667993195e-06,
+      "loss": 0.4064,
+      "step": 19236
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.523891916832485e-06,
+      "loss": 0.4771,
+      "step": 19237
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.523315165671772e-06,
+      "loss": 0.5323,
+      "step": 19238
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.52273841451106e-06,
+      "loss": 0.543,
+      "step": 19239
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.522161663350348e-06,
+      "loss": 0.4236,
+      "step": 19240
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.521584912189637e-06,
+      "loss": 0.3485,
+      "step": 19241
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.521008161028925e-06,
+      "loss": 0.472,
+      "step": 19242
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.520431409868212e-06,
+      "loss": 0.4,
+      "step": 19243
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.519854658707502e-06,
+      "loss": 0.4415,
+      "step": 19244
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.51927790754679e-06,
+      "loss": 0.3695,
+      "step": 19245
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.518701156386077e-06,
+      "loss": 0.4775,
+      "step": 19246
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.518124405225367e-06,
+      "loss": 0.5356,
+      "step": 19247
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.517547654064654e-06,
+      "loss": 0.4677,
+      "step": 19248
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.516970902903942e-06,
+      "loss": 0.39,
+      "step": 19249
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.516394151743231e-06,
+      "loss": 0.3531,
+      "step": 19250
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.51581740058252e-06,
+      "loss": 0.4642,
+      "step": 19251
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.515240649421807e-06,
+      "loss": 0.4359,
+      "step": 19252
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.514663898261096e-06,
+      "loss": 0.3607,
+      "step": 19253
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.514087147100385e-06,
+      "loss": 0.5181,
+      "step": 19254
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.513510395939673e-06,
+      "loss": 0.374,
+      "step": 19255
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.51293364477896e-06,
+      "loss": 0.4512,
+      "step": 19256
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.512356893618248e-06,
+      "loss": 0.5243,
+      "step": 19257
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.511780142457538e-06,
+      "loss": 0.4712,
+      "step": 19258
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.511203391296826e-06,
+      "loss": 0.3827,
+      "step": 19259
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.510626640136113e-06,
+      "loss": 0.458,
+      "step": 19260
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.510049888975403e-06,
+      "loss": 0.4233,
+      "step": 19261
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.50947313781469e-06,
+      "loss": 0.4823,
+      "step": 19262
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.508896386653978e-06,
+      "loss": 0.4032,
+      "step": 19263
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.508319635493267e-06,
+      "loss": 0.376,
+      "step": 19264
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.507742884332555e-06,
+      "loss": 0.3758,
+      "step": 19265
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.507166133171843e-06,
+      "loss": 0.5641,
+      "step": 19266
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506589382011132e-06,
+      "loss": 0.312,
+      "step": 19267
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.506012630850421e-06,
+      "loss": 0.5807,
+      "step": 19268
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.50543587968971e-06,
+      "loss": 0.4382,
+      "step": 19269
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.504859128528997e-06,
+      "loss": 0.3553,
+      "step": 19270
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.504282377368286e-06,
+      "loss": 0.3825,
+      "step": 19271
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.503705626207574e-06,
+      "loss": 0.4651,
+      "step": 19272
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.503128875046862e-06,
+      "loss": 0.4138,
+      "step": 19273
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.50255212388615e-06,
+      "loss": 0.4845,
+      "step": 19274
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.501975372725439e-06,
+      "loss": 0.4517,
+      "step": 19275
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.501398621564726e-06,
+      "loss": 0.4675,
+      "step": 19276
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.500821870404014e-06,
+      "loss": 0.4855,
+      "step": 19277
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.500245119243303e-06,
+      "loss": 0.5414,
+      "step": 19278
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.499668368082591e-06,
+      "loss": 0.4402,
+      "step": 19279
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.499091616921879e-06,
+      "loss": 0.3281,
+      "step": 19280
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.498514865761168e-06,
+      "loss": 0.4632,
+      "step": 19281
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.497938114600458e-06,
+      "loss": 0.4516,
+      "step": 19282
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.497361363439744e-06,
+      "loss": 0.2819,
+      "step": 19283
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.496784612279033e-06,
+      "loss": 0.3501,
+      "step": 19284
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.496207861118322e-06,
+      "loss": 0.4914,
+      "step": 19285
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.49563110995761e-06,
+      "loss": 0.4047,
+      "step": 19286
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.495054358796898e-06,
+      "loss": 0.406,
+      "step": 19287
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.494477607636187e-06,
+      "loss": 0.4698,
+      "step": 19288
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.493900856475475e-06,
+      "loss": 0.4562,
+      "step": 19289
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.493324105314762e-06,
+      "loss": 0.3791,
+      "step": 19290
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.49274735415405e-06,
+      "loss": 0.5131,
+      "step": 19291
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.49217060299334e-06,
+      "loss": 0.4234,
+      "step": 19292
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.491593851832627e-06,
+      "loss": 0.3569,
+      "step": 19293
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.491017100671915e-06,
+      "loss": 0.5251,
+      "step": 19294
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.490440349511204e-06,
+      "loss": 0.4384,
+      "step": 19295
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.489863598350492e-06,
+      "loss": 0.3292,
+      "step": 19296
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48928684718978e-06,
+      "loss": 0.3958,
+      "step": 19297
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.488710096029069e-06,
+      "loss": 0.4778,
+      "step": 19298
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.488133344868358e-06,
+      "loss": 0.4701,
+      "step": 19299
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.487556593707646e-06,
+      "loss": 0.3816,
+      "step": 19300
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.486979842546934e-06,
+      "loss": 0.3922,
+      "step": 19301
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.486403091386223e-06,
+      "loss": 0.4087,
+      "step": 19302
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48582634022551e-06,
+      "loss": 0.5345,
+      "step": 19303
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.485249589064798e-06,
+      "loss": 0.3758,
+      "step": 19304
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.484672837904088e-06,
+      "loss": 0.2705,
+      "step": 19305
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.484096086743376e-06,
+      "loss": 0.4635,
+      "step": 19306
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.483519335582663e-06,
+      "loss": 0.3277,
+      "step": 19307
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.482942584421953e-06,
+      "loss": 0.503,
+      "step": 19308
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.48236583326124e-06,
+      "loss": 0.5031,
+      "step": 19309
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.481789082100528e-06,
+      "loss": 0.39,
+      "step": 19310
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.481212330939816e-06,
+      "loss": 0.4752,
+      "step": 19311
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.480635579779105e-06,
+      "loss": 0.4896,
+      "step": 19312
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.480058828618394e-06,
+      "loss": 0.5526,
+      "step": 19313
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.47948207745768e-06,
+      "loss": 0.4169,
+      "step": 19314
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.47890532629697e-06,
+      "loss": 0.5197,
+      "step": 19315
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.47832857513626e-06,
+      "loss": 0.3416,
+      "step": 19316
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.477751823975547e-06,
+      "loss": 0.4573,
+      "step": 19317
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.477175072814835e-06,
+      "loss": 0.4438,
+      "step": 19318
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.476598321654124e-06,
+      "loss": 0.5174,
+      "step": 19319
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.476021570493412e-06,
+      "loss": 0.4439,
+      "step": 19320
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.4754448193327e-06,
+      "loss": 0.372,
+      "step": 19321
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.474868068171989e-06,
+      "loss": 0.3957,
+      "step": 19322
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.474291317011276e-06,
+      "loss": 0.5634,
+      "step": 19323
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.473714565850564e-06,
+      "loss": 0.3287,
+      "step": 19324
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.473137814689853e-06,
+      "loss": 0.4258,
+      "step": 19325
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.472561063529141e-06,
+      "loss": 0.4876,
+      "step": 19326
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.471984312368429e-06,
+      "loss": 0.362,
+      "step": 19327
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.471407561207716e-06,
+      "loss": 0.5304,
+      "step": 19328
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.470830810047006e-06,
+      "loss": 0.4814,
+      "step": 19329
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.470254058886295e-06,
+      "loss": 0.4962,
+      "step": 19330
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.469677307725583e-06,
+      "loss": 0.4987,
+      "step": 19331
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.46910055656487e-06,
+      "loss": 0.5135,
+      "step": 19332
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.46852380540416e-06,
+      "loss": 0.4761,
+      "step": 19333
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.467947054243448e-06,
+      "loss": 0.4972,
+      "step": 19334
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.467370303082735e-06,
+      "loss": 0.527,
+      "step": 19335
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.466793551922025e-06,
+      "loss": 0.3974,
+      "step": 19336
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.466216800761312e-06,
+      "loss": 0.3725,
+      "step": 19337
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.4656400496006e-06,
+      "loss": 0.4784,
+      "step": 19338
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.46506329843989e-06,
+      "loss": 0.4717,
+      "step": 19339
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.464486547279177e-06,
+      "loss": 0.3869,
+      "step": 19340
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.463909796118465e-06,
+      "loss": 0.4033,
+      "step": 19341
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.463333044957754e-06,
+      "loss": 0.4499,
+      "step": 19342
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.462756293797042e-06,
+      "loss": 0.4722,
+      "step": 19343
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.462179542636331e-06,
+      "loss": 0.5096,
+      "step": 19344
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.461602791475617e-06,
+      "loss": 0.4176,
+      "step": 19345
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.461026040314907e-06,
+      "loss": 0.4563,
+      "step": 19346
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.460449289154196e-06,
+      "loss": 0.4384,
+      "step": 19347
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.459872537993484e-06,
+      "loss": 0.5319,
+      "step": 19348
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.459295786832771e-06,
+      "loss": 0.4768,
+      "step": 19349
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.45871903567206e-06,
+      "loss": 0.3454,
+      "step": 19350
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.458142284511348e-06,
+      "loss": 0.4863,
+      "step": 19351
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.457565533350636e-06,
+      "loss": 0.5329,
+      "step": 19352
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.456988782189926e-06,
+      "loss": 0.3224,
+      "step": 19353
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.456412031029213e-06,
+      "loss": 0.5383,
+      "step": 19354
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.455835279868501e-06,
+      "loss": 0.5476,
+      "step": 19355
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.45525852870779e-06,
+      "loss": 0.3857,
+      "step": 19356
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.454681777547078e-06,
+      "loss": 0.445,
+      "step": 19357
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.454105026386366e-06,
+      "loss": 0.4791,
+      "step": 19358
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.453528275225655e-06,
+      "loss": 0.575,
+      "step": 19359
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.452951524064943e-06,
+      "loss": 0.5899,
+      "step": 19360
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.452374772904232e-06,
+      "loss": 0.4323,
+      "step": 19361
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.45179802174352e-06,
+      "loss": 0.4589,
+      "step": 19362
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.451221270582807e-06,
+      "loss": 0.4314,
+      "step": 19363
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.450644519422097e-06,
+      "loss": 0.5164,
+      "step": 19364
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.450067768261385e-06,
+      "loss": 0.4553,
+      "step": 19365
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.449491017100672e-06,
+      "loss": 0.3951,
+      "step": 19366
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.448914265939962e-06,
+      "loss": 0.3973,
+      "step": 19367
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.44833751477925e-06,
+      "loss": 0.3889,
+      "step": 19368
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.447760763618537e-06,
+      "loss": 0.4012,
+      "step": 19369
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.447184012457826e-06,
+      "loss": 0.3842,
+      "step": 19370
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.446607261297114e-06,
+      "loss": 0.4329,
+      "step": 19371
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.446030510136402e-06,
+      "loss": 0.4737,
+      "step": 19372
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.445453758975691e-06,
+      "loss": 0.5046,
+      "step": 19373
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444877007814979e-06,
+      "loss": 0.4302,
+      "step": 19374
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.444300256654268e-06,
+      "loss": 0.4768,
+      "step": 19375
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.443723505493556e-06,
+      "loss": 0.3585,
+      "step": 19376
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.443146754332844e-06,
+      "loss": 0.4828,
+      "step": 19377
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.442570003172133e-06,
+      "loss": 0.4951,
+      "step": 19378
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.44199325201142e-06,
+      "loss": 0.4663,
+      "step": 19379
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.441416500850708e-06,
+      "loss": 0.4335,
+      "step": 19380
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.440839749689998e-06,
+      "loss": 0.4091,
+      "step": 19381
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.440262998529285e-06,
+      "loss": 0.4295,
+      "step": 19382
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.439686247368573e-06,
+      "loss": 0.4582,
+      "step": 19383
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.439109496207862e-06,
+      "loss": 0.4011,
+      "step": 19384
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.43853274504715e-06,
+      "loss": 0.2847,
+      "step": 19385
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.437955993886438e-06,
+      "loss": 0.4913,
+      "step": 19386
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.437379242725727e-06,
+      "loss": 0.4488,
+      "step": 19387
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.436802491565015e-06,
+      "loss": 0.3946,
+      "step": 19388
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.436225740404304e-06,
+      "loss": 0.4304,
+      "step": 19389
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.435648989243592e-06,
+      "loss": 0.4793,
+      "step": 19390
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.43507223808288e-06,
+      "loss": 0.4611,
+      "step": 19391
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.434495486922169e-06,
+      "loss": 0.5197,
+      "step": 19392
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.433918735761457e-06,
+      "loss": 0.3882,
+      "step": 19393
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 9.433341984600744e-06,
+      "loss": 0.3513,
+      "step": 19394
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.432765233440034e-06,
+      "loss": 0.3708,
+      "step": 19395
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.432188482279321e-06,
+      "loss": 0.3867,
+      "step": 19396
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.431611731118609e-06,
+      "loss": 0.6042,
+      "step": 19397
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.431034979957898e-06,
+      "loss": 0.4363,
+      "step": 19398
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.430458228797186e-06,
+      "loss": 0.4301,
+      "step": 19399
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.429881477636474e-06,
+      "loss": 0.4,
+      "step": 19400
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.429304726475763e-06,
+      "loss": 0.5843,
+      "step": 19401
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.428727975315051e-06,
+      "loss": 0.497,
+      "step": 19402
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.428151224154339e-06,
+      "loss": 0.4054,
+      "step": 19403
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.427574472993628e-06,
+      "loss": 0.4911,
+      "step": 19404
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.426997721832916e-06,
+      "loss": 0.3923,
+      "step": 19405
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.426420970672205e-06,
+      "loss": 0.467,
+      "step": 19406
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.425844219511493e-06,
+      "loss": 0.5117,
+      "step": 19407
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.42526746835078e-06,
+      "loss": 0.3598,
+      "step": 19408
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.42469071719007e-06,
+      "loss": 0.5742,
+      "step": 19409
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.424113966029357e-06,
+      "loss": 0.5522,
+      "step": 19410
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.423537214868645e-06,
+      "loss": 0.4997,
+      "step": 19411
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.422960463707935e-06,
+      "loss": 0.4805,
+      "step": 19412
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.422383712547222e-06,
+      "loss": 0.484,
+      "step": 19413
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.42180696138651e-06,
+      "loss": 0.4763,
+      "step": 19414
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.4212302102258e-06,
+      "loss": 0.4825,
+      "step": 19415
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.420653459065087e-06,
+      "loss": 0.5748,
+      "step": 19416
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.420076707904375e-06,
+      "loss": 0.3722,
+      "step": 19417
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.419499956743664e-06,
+      "loss": 0.3693,
+      "step": 19418
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.418923205582952e-06,
+      "loss": 0.3858,
+      "step": 19419
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.418346454422241e-06,
+      "loss": 0.4838,
+      "step": 19420
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.417769703261529e-06,
+      "loss": 0.4172,
+      "step": 19421
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.417192952100816e-06,
+      "loss": 0.4252,
+      "step": 19422
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.416616200940106e-06,
+      "loss": 0.4233,
+      "step": 19423
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.416039449779394e-06,
+      "loss": 0.5065,
+      "step": 19424
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.415462698618681e-06,
+      "loss": 0.4168,
+      "step": 19425
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.41488594745797e-06,
+      "loss": 0.419,
+      "step": 19426
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.414309196297258e-06,
+      "loss": 0.4775,
+      "step": 19427
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.413732445136546e-06,
+      "loss": 0.4374,
+      "step": 19428
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.413155693975835e-06,
+      "loss": 0.4871,
+      "step": 19429
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.412578942815123e-06,
+      "loss": 0.3513,
+      "step": 19430
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.41200219165441e-06,
+      "loss": 0.4153,
+      "step": 19431
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.4114254404937e-06,
+      "loss": 0.4458,
+      "step": 19432
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.410848689332988e-06,
+      "loss": 0.4206,
+      "step": 19433
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.410271938172275e-06,
+      "loss": 0.4758,
+      "step": 19434
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.409695187011565e-06,
+      "loss": 0.3385,
+      "step": 19435
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.409118435850853e-06,
+      "loss": 0.4825,
+      "step": 19436
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.408541684690142e-06,
+      "loss": 0.3933,
+      "step": 19437
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.40796493352943e-06,
+      "loss": 0.3505,
+      "step": 19438
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.407388182368717e-06,
+      "loss": 0.5458,
+      "step": 19439
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.406811431208007e-06,
+      "loss": 0.3759,
+      "step": 19440
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.406234680047294e-06,
+      "loss": 0.4979,
+      "step": 19441
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.405657928886582e-06,
+      "loss": 0.4537,
+      "step": 19442
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.405081177725871e-06,
+      "loss": 0.5328,
+      "step": 19443
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.404504426565159e-06,
+      "loss": 0.5189,
+      "step": 19444
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403927675404447e-06,
+      "loss": 0.4167,
+      "step": 19445
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.403350924243736e-06,
+      "loss": 0.4041,
+      "step": 19446
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.402774173083024e-06,
+      "loss": 0.3875,
+      "step": 19447
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.402197421922312e-06,
+      "loss": 0.5099,
+      "step": 19448
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.401620670761601e-06,
+      "loss": 0.5087,
+      "step": 19449
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.401043919600889e-06,
+      "loss": 0.3312,
+      "step": 19450
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.400467168440178e-06,
+      "loss": 0.4413,
+      "step": 19451
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.399890417279466e-06,
+      "loss": 0.4508,
+      "step": 19452
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.399313666118753e-06,
+      "loss": 0.395,
+      "step": 19453
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.398736914958043e-06,
+      "loss": 0.3788,
+      "step": 19454
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.39816016379733e-06,
+      "loss": 0.5401,
+      "step": 19455
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.397583412636618e-06,
+      "loss": 0.4138,
+      "step": 19456
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.397006661475907e-06,
+      "loss": 0.5129,
+      "step": 19457
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.396429910315195e-06,
+      "loss": 0.36,
+      "step": 19458
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.395853159154483e-06,
+      "loss": 0.3743,
+      "step": 19459
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.395276407993772e-06,
+      "loss": 0.393,
+      "step": 19460
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.39469965683306e-06,
+      "loss": 0.5025,
+      "step": 19461
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.394122905672348e-06,
+      "loss": 0.5358,
+      "step": 19462
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.393546154511637e-06,
+      "loss": 0.4092,
+      "step": 19463
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.392969403350926e-06,
+      "loss": 0.3705,
+      "step": 19464
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.392392652190212e-06,
+      "loss": 0.5015,
+      "step": 19465
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.391815901029502e-06,
+      "loss": 0.4021,
+      "step": 19466
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.39123914986879e-06,
+      "loss": 0.3323,
+      "step": 19467
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.390662398708079e-06,
+      "loss": 0.3487,
+      "step": 19468
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.390085647547366e-06,
+      "loss": 0.4737,
+      "step": 19469
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.389508896386654e-06,
+      "loss": 0.4507,
+      "step": 19470
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.388932145225944e-06,
+      "loss": 0.5674,
+      "step": 19471
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.388355394065231e-06,
+      "loss": 0.3959,
+      "step": 19472
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.387778642904519e-06,
+      "loss": 0.6163,
+      "step": 19473
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.387201891743808e-06,
+      "loss": 0.4373,
+      "step": 19474
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.386625140583096e-06,
+      "loss": 0.4525,
+      "step": 19475
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.386048389422384e-06,
+      "loss": 0.3226,
+      "step": 19476
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.385471638261673e-06,
+      "loss": 0.3931,
+      "step": 19477
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.38489488710096e-06,
+      "loss": 0.3538,
+      "step": 19478
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.384318135940248e-06,
+      "loss": 0.5689,
+      "step": 19479
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.383741384779538e-06,
+      "loss": 0.4422,
+      "step": 19480
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.383164633618827e-06,
+      "loss": 0.4557,
+      "step": 19481
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382587882458115e-06,
+      "loss": 0.3454,
+      "step": 19482
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.382011131297403e-06,
+      "loss": 0.4654,
+      "step": 19483
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.38143438013669e-06,
+      "loss": 0.4154,
+      "step": 19484
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.38085762897598e-06,
+      "loss": 0.4356,
+      "step": 19485
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.380280877815267e-06,
+      "loss": 0.4066,
+      "step": 19486
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.379704126654555e-06,
+      "loss": 0.4491,
+      "step": 19487
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.379127375493844e-06,
+      "loss": 0.3779,
+      "step": 19488
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.378550624333132e-06,
+      "loss": 0.4872,
+      "step": 19489
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.37797387317242e-06,
+      "loss": 0.4132,
+      "step": 19490
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.377397122011709e-06,
+      "loss": 0.5425,
+      "step": 19491
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.376820370850997e-06,
+      "loss": 0.5483,
+      "step": 19492
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.376243619690284e-06,
+      "loss": 0.4529,
+      "step": 19493
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.375666868529574e-06,
+      "loss": 0.4041,
+      "step": 19494
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.375090117368863e-06,
+      "loss": 0.4127,
+      "step": 19495
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.37451336620815e-06,
+      "loss": 0.4504,
+      "step": 19496
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.373936615047439e-06,
+      "loss": 0.4142,
+      "step": 19497
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.373359863886728e-06,
+      "loss": 0.5451,
+      "step": 19498
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.372783112726016e-06,
+      "loss": 0.4519,
+      "step": 19499
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.372206361565303e-06,
+      "loss": 0.4641,
+      "step": 19500
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.371629610404591e-06,
+      "loss": 0.4416,
+      "step": 19501
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.37105285924388e-06,
+      "loss": 0.3359,
+      "step": 19502
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.370476108083168e-06,
+      "loss": 0.4687,
+      "step": 19503
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.369899356922456e-06,
+      "loss": 0.4924,
+      "step": 19504
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.369322605761745e-06,
+      "loss": 0.4112,
+      "step": 19505
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.368745854601033e-06,
+      "loss": 0.4309,
+      "step": 19506
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.36816910344032e-06,
+      "loss": 0.4805,
+      "step": 19507
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.36759235227961e-06,
+      "loss": 0.4873,
+      "step": 19508
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.367015601118898e-06,
+      "loss": 0.4481,
+      "step": 19509
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.366438849958185e-06,
+      "loss": 0.5338,
+      "step": 19510
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.365862098797475e-06,
+      "loss": 0.4199,
+      "step": 19511
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.365285347636764e-06,
+      "loss": 0.5267,
+      "step": 19512
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.364708596476052e-06,
+      "loss": 0.3696,
+      "step": 19513
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.36413184531534e-06,
+      "loss": 0.5377,
+      "step": 19514
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.363555094154629e-06,
+      "loss": 0.4145,
+      "step": 19515
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362978342993916e-06,
+      "loss": 0.5028,
+      "step": 19516
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.362401591833204e-06,
+      "loss": 0.4569,
+      "step": 19517
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.361824840672492e-06,
+      "loss": 0.3603,
+      "step": 19518
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.361248089511781e-06,
+      "loss": 0.4311,
+      "step": 19519
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.360671338351069e-06,
+      "loss": 0.4363,
+      "step": 19520
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.360094587190357e-06,
+      "loss": 0.4678,
+      "step": 19521
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.359517836029646e-06,
+      "loss": 0.437,
+      "step": 19522
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.358941084868934e-06,
+      "loss": 0.5664,
+      "step": 19523
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.358364333708221e-06,
+      "loss": 0.4628,
+      "step": 19524
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.35778758254751e-06,
+      "loss": 0.6008,
+      "step": 19525
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.3572108313868e-06,
+      "loss": 0.4732,
+      "step": 19526
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.356634080226086e-06,
+      "loss": 0.4517,
+      "step": 19527
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.356057329065375e-06,
+      "loss": 0.4115,
+      "step": 19528
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.355480577904665e-06,
+      "loss": 0.4898,
+      "step": 19529
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.354903826743953e-06,
+      "loss": 0.3603,
+      "step": 19530
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.35432707558324e-06,
+      "loss": 0.444,
+      "step": 19531
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.35375032442253e-06,
+      "loss": 0.517,
+      "step": 19532
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.353173573261817e-06,
+      "loss": 0.5261,
+      "step": 19533
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.352596822101105e-06,
+      "loss": 0.4398,
+      "step": 19534
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.352020070940393e-06,
+      "loss": 0.4157,
+      "step": 19535
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.351443319779682e-06,
+      "loss": 0.5536,
+      "step": 19536
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.35086656861897e-06,
+      "loss": 0.5367,
+      "step": 19537
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.350289817458257e-06,
+      "loss": 0.4473,
+      "step": 19538
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.349713066297547e-06,
+      "loss": 0.5118,
+      "step": 19539
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.349136315136834e-06,
+      "loss": 0.4584,
+      "step": 19540
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.348559563976122e-06,
+      "loss": 0.42,
+      "step": 19541
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.347982812815412e-06,
+      "loss": 0.5607,
+      "step": 19542
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.347406061654701e-06,
+      "loss": 0.5176,
+      "step": 19543
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.346829310493989e-06,
+      "loss": 0.4461,
+      "step": 19544
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.346252559333276e-06,
+      "loss": 0.475,
+      "step": 19545
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.345675808172566e-06,
+      "loss": 0.4373,
+      "step": 19546
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.345099057011853e-06,
+      "loss": 0.4326,
+      "step": 19547
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.344522305851141e-06,
+      "loss": 0.4328,
+      "step": 19548
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.34394555469043e-06,
+      "loss": 0.4749,
+      "step": 19549
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.343368803529718e-06,
+      "loss": 0.4532,
+      "step": 19550
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.342792052369006e-06,
+      "loss": 0.4166,
+      "step": 19551
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.342215301208293e-06,
+      "loss": 0.4221,
+      "step": 19552
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.341638550047583e-06,
+      "loss": 0.4129,
+      "step": 19553
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.34106179888687e-06,
+      "loss": 0.4079,
+      "step": 19554
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.340485047726158e-06,
+      "loss": 0.5809,
+      "step": 19555
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.339908296565448e-06,
+      "loss": 0.4374,
+      "step": 19556
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.339331545404737e-06,
+      "loss": 0.4624,
+      "step": 19557
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.338754794244023e-06,
+      "loss": 0.4289,
+      "step": 19558
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.338178043083312e-06,
+      "loss": 0.3679,
+      "step": 19559
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.337601291922602e-06,
+      "loss": 0.4857,
+      "step": 19560
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.33702454076189e-06,
+      "loss": 0.4743,
+      "step": 19561
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.336447789601177e-06,
+      "loss": 0.5471,
+      "step": 19562
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.335871038440466e-06,
+      "loss": 0.3501,
+      "step": 19563
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.335294287279754e-06,
+      "loss": 0.4549,
+      "step": 19564
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.334717536119042e-06,
+      "loss": 0.5354,
+      "step": 19565
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.334140784958331e-06,
+      "loss": 0.3577,
+      "step": 19566
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.333564033797619e-06,
+      "loss": 0.5548,
+      "step": 19567
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.332987282636907e-06,
+      "loss": 0.424,
+      "step": 19568
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.332410531476196e-06,
+      "loss": 0.5056,
+      "step": 19569
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.331833780315484e-06,
+      "loss": 0.4331,
+      "step": 19570
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.331257029154771e-06,
+      "loss": 0.5878,
+      "step": 19571
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.330680277994059e-06,
+      "loss": 0.4575,
+      "step": 19572
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 9.330103526833348e-06,
+      "loss": 0.3121,
+      "step": 19573
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.329526775672638e-06,
+      "loss": 0.4305,
+      "step": 19574
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.328950024511925e-06,
+      "loss": 0.4958,
+      "step": 19575
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.328373273351213e-06,
+      "loss": 0.4329,
+      "step": 19576
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.327796522190503e-06,
+      "loss": 0.3897,
+      "step": 19577
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.32721977102979e-06,
+      "loss": 0.4352,
+      "step": 19578
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.326643019869078e-06,
+      "loss": 0.4349,
+      "step": 19579
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.326066268708367e-06,
+      "loss": 0.3968,
+      "step": 19580
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.325489517547655e-06,
+      "loss": 0.3929,
+      "step": 19581
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.324912766386943e-06,
+      "loss": 0.5108,
+      "step": 19582
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.324336015226232e-06,
+      "loss": 0.3222,
+      "step": 19583
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.32375926406552e-06,
+      "loss": 0.4152,
+      "step": 19584
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.323182512904807e-06,
+      "loss": 0.3081,
+      "step": 19585
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.322605761744097e-06,
+      "loss": 0.4197,
+      "step": 19586
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.322029010583384e-06,
+      "loss": 0.4284,
+      "step": 19587
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.321452259422674e-06,
+      "loss": 0.4792,
+      "step": 19588
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.32087550826196e-06,
+      "loss": 0.4443,
+      "step": 19589
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.32029875710125e-06,
+      "loss": 0.3727,
+      "step": 19590
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.319722005940539e-06,
+      "loss": 0.318,
+      "step": 19591
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.319145254779826e-06,
+      "loss": 0.4652,
+      "step": 19592
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.318568503619114e-06,
+      "loss": 0.4143,
+      "step": 19593
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.317991752458403e-06,
+      "loss": 0.5429,
+      "step": 19594
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.317415001297691e-06,
+      "loss": 0.5051,
+      "step": 19595
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.316838250136979e-06,
+      "loss": 0.5281,
+      "step": 19596
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.316261498976268e-06,
+      "loss": 0.599,
+      "step": 19597
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.315684747815556e-06,
+      "loss": 0.4007,
+      "step": 19598
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.315107996654843e-06,
+      "loss": 0.4056,
+      "step": 19599
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.314531245494133e-06,
+      "loss": 0.3812,
+      "step": 19600
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.31395449433342e-06,
+      "loss": 0.4012,
+      "step": 19601
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.313377743172708e-06,
+      "loss": 0.3478,
+      "step": 19602
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.312800992011998e-06,
+      "loss": 0.4693,
+      "step": 19603
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.312224240851285e-06,
+      "loss": 0.3641,
+      "step": 19604
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.311647489690575e-06,
+      "loss": 0.4252,
+      "step": 19605
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.311070738529862e-06,
+      "loss": 0.3736,
+      "step": 19606
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.31049398736915e-06,
+      "loss": 0.4789,
+      "step": 19607
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.30991723620844e-06,
+      "loss": 0.5341,
+      "step": 19608
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.309340485047727e-06,
+      "loss": 0.4066,
+      "step": 19609
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.308763733887015e-06,
+      "loss": 0.4374,
+      "step": 19610
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.308186982726304e-06,
+      "loss": 0.5857,
+      "step": 19611
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.307610231565592e-06,
+      "loss": 0.4595,
+      "step": 19612
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.30703348040488e-06,
+      "loss": 0.4025,
+      "step": 19613
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.306456729244169e-06,
+      "loss": 0.4061,
+      "step": 19614
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.305879978083457e-06,
+      "loss": 0.4126,
+      "step": 19615
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.305303226922744e-06,
+      "loss": 0.5466,
+      "step": 19616
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.304726475762034e-06,
+      "loss": 0.5742,
+      "step": 19617
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.304149724601321e-06,
+      "loss": 0.4398,
+      "step": 19618
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.30357297344061e-06,
+      "loss": 0.3731,
+      "step": 19619
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.302996222279898e-06,
+      "loss": 0.4456,
+      "step": 19620
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.302419471119186e-06,
+      "loss": 0.3766,
+      "step": 19621
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.301842719958475e-06,
+      "loss": 0.4851,
+      "step": 19622
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.301265968797763e-06,
+      "loss": 0.3472,
+      "step": 19623
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.30068921763705e-06,
+      "loss": 0.4091,
+      "step": 19624
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.30011246647634e-06,
+      "loss": 0.4058,
+      "step": 19625
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.299535715315628e-06,
+      "loss": 0.4468,
+      "step": 19626
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.298958964154916e-06,
+      "loss": 0.4229,
+      "step": 19627
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.298382212994205e-06,
+      "loss": 0.4332,
+      "step": 19628
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.297805461833493e-06,
+      "loss": 0.4138,
+      "step": 19629
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.29722871067278e-06,
+      "loss": 0.3723,
+      "step": 19630
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.29665195951207e-06,
+      "loss": 0.4019,
+      "step": 19631
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.296075208351357e-06,
+      "loss": 0.5485,
+      "step": 19632
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.295498457190645e-06,
+      "loss": 0.4105,
+      "step": 19633
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.294921706029934e-06,
+      "loss": 0.4544,
+      "step": 19634
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.294344954869222e-06,
+      "loss": 0.5012,
+      "step": 19635
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.293768203708512e-06,
+      "loss": 0.4731,
+      "step": 19636
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.2931914525478e-06,
+      "loss": 0.573,
+      "step": 19637
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.292614701387087e-06,
+      "loss": 0.4021,
+      "step": 19638
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.292037950226376e-06,
+      "loss": 0.4048,
+      "step": 19639
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.291461199065664e-06,
+      "loss": 0.3707,
+      "step": 19640
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.290884447904952e-06,
+      "loss": 0.477,
+      "step": 19641
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.290307696744241e-06,
+      "loss": 0.4128,
+      "step": 19642
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.289730945583529e-06,
+      "loss": 0.5537,
+      "step": 19643
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.289154194422816e-06,
+      "loss": 0.4602,
+      "step": 19644
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.288577443262106e-06,
+      "loss": 0.4528,
+      "step": 19645
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.288000692101393e-06,
+      "loss": 0.4894,
+      "step": 19646
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.287423940940681e-06,
+      "loss": 0.4855,
+      "step": 19647
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.28684718977997e-06,
+      "loss": 0.4497,
+      "step": 19648
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.286270438619258e-06,
+      "loss": 0.5017,
+      "step": 19649
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.285693687458548e-06,
+      "loss": 0.4591,
+      "step": 19650
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.285116936297835e-06,
+      "loss": 0.3713,
+      "step": 19651
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.284540185137123e-06,
+      "loss": 0.5232,
+      "step": 19652
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.283963433976412e-06,
+      "loss": 0.4649,
+      "step": 19653
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.2833866828157e-06,
+      "loss": 0.4389,
+      "step": 19654
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.282809931654988e-06,
+      "loss": 0.5409,
+      "step": 19655
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.282233180494277e-06,
+      "loss": 0.6013,
+      "step": 19656
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.281656429333565e-06,
+      "loss": 0.4066,
+      "step": 19657
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.281079678172852e-06,
+      "loss": 0.4611,
+      "step": 19658
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.280502927012142e-06,
+      "loss": 0.4473,
+      "step": 19659
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.27992617585143e-06,
+      "loss": 0.4625,
+      "step": 19660
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.279349424690717e-06,
+      "loss": 0.3874,
+      "step": 19661
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.278772673530007e-06,
+      "loss": 0.3992,
+      "step": 19662
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.278195922369294e-06,
+      "loss": 0.4619,
+      "step": 19663
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.277619171208582e-06,
+      "loss": 0.4108,
+      "step": 19664
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.277042420047871e-06,
+      "loss": 0.5194,
+      "step": 19665
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.276465668887159e-06,
+      "loss": 0.4259,
+      "step": 19666
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.275888917726448e-06,
+      "loss": 0.4028,
+      "step": 19667
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.275312166565736e-06,
+      "loss": 0.4073,
+      "step": 19668
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.274735415405024e-06,
+      "loss": 0.4119,
+      "step": 19669
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.274158664244313e-06,
+      "loss": 0.4852,
+      "step": 19670
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.2735819130836e-06,
+      "loss": 0.427,
+      "step": 19671
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.273005161922889e-06,
+      "loss": 0.3593,
+      "step": 19672
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.272428410762178e-06,
+      "loss": 0.4098,
+      "step": 19673
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.271851659601466e-06,
+      "loss": 0.4051,
+      "step": 19674
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.271274908440753e-06,
+      "loss": 0.3833,
+      "step": 19675
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.270698157280043e-06,
+      "loss": 0.3326,
+      "step": 19676
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.27012140611933e-06,
+      "loss": 0.3882,
+      "step": 19677
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.269544654958618e-06,
+      "loss": 0.3262,
+      "step": 19678
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.268967903797907e-06,
+      "loss": 0.4338,
+      "step": 19679
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.268391152637195e-06,
+      "loss": 0.4368,
+      "step": 19680
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.267814401476484e-06,
+      "loss": 0.3981,
+      "step": 19681
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.267237650315772e-06,
+      "loss": 0.4156,
+      "step": 19682
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.26666089915506e-06,
+      "loss": 0.3442,
+      "step": 19683
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.26608414799435e-06,
+      "loss": 0.5411,
+      "step": 19684
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.265507396833637e-06,
+      "loss": 0.4507,
+      "step": 19685
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.264930645672925e-06,
+      "loss": 0.3865,
+      "step": 19686
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.264353894512214e-06,
+      "loss": 0.4388,
+      "step": 19687
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.263777143351502e-06,
+      "loss": 0.534,
+      "step": 19688
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.26320039219079e-06,
+      "loss": 0.4062,
+      "step": 19689
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.262623641030079e-06,
+      "loss": 0.5027,
+      "step": 19690
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.262046889869366e-06,
+      "loss": 0.6339,
+      "step": 19691
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.261470138708654e-06,
+      "loss": 0.2727,
+      "step": 19692
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.260893387547943e-06,
+      "loss": 0.3981,
+      "step": 19693
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.260316636387231e-06,
+      "loss": 0.4341,
+      "step": 19694
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.259739885226519e-06,
+      "loss": 0.2557,
+      "step": 19695
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.259163134065808e-06,
+      "loss": 0.44,
+      "step": 19696
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.258586382905096e-06,
+      "loss": 0.5404,
+      "step": 19697
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.258009631744385e-06,
+      "loss": 0.5092,
+      "step": 19698
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.257432880583673e-06,
+      "loss": 0.4426,
+      "step": 19699
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25685612942296e-06,
+      "loss": 0.4787,
+      "step": 19700
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25627937826225e-06,
+      "loss": 0.4919,
+      "step": 19701
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.255702627101538e-06,
+      "loss": 0.4262,
+      "step": 19702
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.255125875940825e-06,
+      "loss": 0.3797,
+      "step": 19703
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.254549124780115e-06,
+      "loss": 0.4774,
+      "step": 19704
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.253972373619402e-06,
+      "loss": 0.5111,
+      "step": 19705
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25339562245869e-06,
+      "loss": 0.3929,
+      "step": 19706
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.25281887129798e-06,
+      "loss": 0.3291,
+      "step": 19707
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.252242120137269e-06,
+      "loss": 0.4329,
+      "step": 19708
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.251665368976555e-06,
+      "loss": 0.4166,
+      "step": 19709
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.251088617815844e-06,
+      "loss": 0.4524,
+      "step": 19710
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.250511866655132e-06,
+      "loss": 0.4135,
+      "step": 19711
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.249935115494421e-06,
+      "loss": 0.3487,
+      "step": 19712
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.249358364333709e-06,
+      "loss": 0.3947,
+      "step": 19713
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.248781613172997e-06,
+      "loss": 0.426,
+      "step": 19714
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.248204862012286e-06,
+      "loss": 0.5131,
+      "step": 19715
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.247628110851574e-06,
+      "loss": 0.3663,
+      "step": 19716
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.247051359690861e-06,
+      "loss": 0.5968,
+      "step": 19717
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.24647460853015e-06,
+      "loss": 0.4512,
+      "step": 19718
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.245897857369439e-06,
+      "loss": 0.4537,
+      "step": 19719
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.245321106208726e-06,
+      "loss": 0.4602,
+      "step": 19720
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.244744355048016e-06,
+      "loss": 0.3705,
+      "step": 19721
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.244167603887303e-06,
+      "loss": 0.4348,
+      "step": 19722
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.243590852726591e-06,
+      "loss": 0.3505,
+      "step": 19723
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.24301410156588e-06,
+      "loss": 0.4096,
+      "step": 19724
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.24243735040517e-06,
+      "loss": 0.5697,
+      "step": 19725
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.241860599244456e-06,
+      "loss": 0.3853,
+      "step": 19726
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.241283848083745e-06,
+      "loss": 0.3667,
+      "step": 19727
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.240707096923033e-06,
+      "loss": 0.3454,
+      "step": 19728
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.240130345762322e-06,
+      "loss": 0.528,
+      "step": 19729
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.23955359460161e-06,
+      "loss": 0.5492,
+      "step": 19730
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238976843440898e-06,
+      "loss": 0.5038,
+      "step": 19731
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.238400092280187e-06,
+      "loss": 0.4765,
+      "step": 19732
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.237823341119475e-06,
+      "loss": 0.5031,
+      "step": 19733
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.237246589958762e-06,
+      "loss": 0.4933,
+      "step": 19734
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.236669838798052e-06,
+      "loss": 0.4872,
+      "step": 19735
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.23609308763734e-06,
+      "loss": 0.4689,
+      "step": 19736
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.235516336476627e-06,
+      "loss": 0.4452,
+      "step": 19737
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.234939585315916e-06,
+      "loss": 0.4732,
+      "step": 19738
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.234362834155206e-06,
+      "loss": 0.3419,
+      "step": 19739
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.233786082994492e-06,
+      "loss": 0.4607,
+      "step": 19740
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.233209331833781e-06,
+      "loss": 0.4403,
+      "step": 19741
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.23263258067307e-06,
+      "loss": 0.5233,
+      "step": 19742
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.232055829512358e-06,
+      "loss": 0.4828,
+      "step": 19743
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.231479078351646e-06,
+      "loss": 0.48,
+      "step": 19744
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.230902327190934e-06,
+      "loss": 0.5419,
+      "step": 19745
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.230325576030223e-06,
+      "loss": 0.4661,
+      "step": 19746
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.22974882486951e-06,
+      "loss": 0.2851,
+      "step": 19747
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.229172073708798e-06,
+      "loss": 0.4134,
+      "step": 19748
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.228595322548088e-06,
+      "loss": 0.4758,
+      "step": 19749
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.228018571387375e-06,
+      "loss": 0.4329,
+      "step": 19750
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 9.227441820226663e-06,
+      "loss": 0.4803,
+      "step": 19751
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.226865069065952e-06,
+      "loss": 0.5211,
+      "step": 19752
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.22628831790524e-06,
+      "loss": 0.4084,
+      "step": 19753
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.225711566744528e-06,
+      "loss": 0.4901,
+      "step": 19754
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.225134815583817e-06,
+      "loss": 0.4951,
+      "step": 19755
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.224558064423107e-06,
+      "loss": 0.4568,
+      "step": 19756
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.223981313262394e-06,
+      "loss": 0.437,
+      "step": 19757
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.223404562101682e-06,
+      "loss": 0.2916,
+      "step": 19758
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.222827810940971e-06,
+      "loss": 0.4303,
+      "step": 19759
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.222251059780259e-06,
+      "loss": 0.4881,
+      "step": 19760
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.221674308619547e-06,
+      "loss": 0.4637,
+      "step": 19761
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.221097557458834e-06,
+      "loss": 0.4339,
+      "step": 19762
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.220520806298124e-06,
+      "loss": 0.5524,
+      "step": 19763
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.219944055137411e-06,
+      "loss": 0.3542,
+      "step": 19764
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.219367303976699e-06,
+      "loss": 0.2913,
+      "step": 19765
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218790552815989e-06,
+      "loss": 0.4818,
+      "step": 19766
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.218213801655276e-06,
+      "loss": 0.5108,
+      "step": 19767
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.217637050494564e-06,
+      "loss": 0.4221,
+      "step": 19768
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.217060299333853e-06,
+      "loss": 0.4648,
+      "step": 19769
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.216483548173143e-06,
+      "loss": 0.4343,
+      "step": 19770
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.215906797012429e-06,
+      "loss": 0.4899,
+      "step": 19771
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.215330045851718e-06,
+      "loss": 0.6411,
+      "step": 19772
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.214753294691007e-06,
+      "loss": 0.4268,
+      "step": 19773
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.214176543530295e-06,
+      "loss": 0.427,
+      "step": 19774
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.213599792369583e-06,
+      "loss": 0.5291,
+      "step": 19775
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.213023041208872e-06,
+      "loss": 0.4867,
+      "step": 19776
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.21244629004816e-06,
+      "loss": 0.4372,
+      "step": 19777
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.211869538887448e-06,
+      "loss": 0.4586,
+      "step": 19778
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.211292787726735e-06,
+      "loss": 0.4345,
+      "step": 19779
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.210716036566025e-06,
+      "loss": 0.4952,
+      "step": 19780
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.210139285405312e-06,
+      "loss": 0.4919,
+      "step": 19781
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.2095625342446e-06,
+      "loss": 0.479,
+      "step": 19782
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.20898578308389e-06,
+      "loss": 0.4971,
+      "step": 19783
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.208409031923177e-06,
+      "loss": 0.4798,
+      "step": 19784
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.207832280762465e-06,
+      "loss": 0.3271,
+      "step": 19785
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.207255529601754e-06,
+      "loss": 0.3086,
+      "step": 19786
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.206678778441043e-06,
+      "loss": 0.3903,
+      "step": 19787
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.206102027280331e-06,
+      "loss": 0.4191,
+      "step": 19788
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.205525276119619e-06,
+      "loss": 0.4196,
+      "step": 19789
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.204948524958908e-06,
+      "loss": 0.4175,
+      "step": 19790
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.204371773798196e-06,
+      "loss": 0.4257,
+      "step": 19791
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.203795022637484e-06,
+      "loss": 0.416,
+      "step": 19792
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.203218271476773e-06,
+      "loss": 0.246,
+      "step": 19793
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.20264152031606e-06,
+      "loss": 0.488,
+      "step": 19794
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.202064769155348e-06,
+      "loss": 0.5518,
+      "step": 19795
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.201488017994636e-06,
+      "loss": 0.5296,
+      "step": 19796
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.200911266833925e-06,
+      "loss": 0.4111,
+      "step": 19797
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.200334515673213e-06,
+      "loss": 0.4665,
+      "step": 19798
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.1997577645125e-06,
+      "loss": 0.4514,
+      "step": 19799
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.19918101335179e-06,
+      "loss": 0.5046,
+      "step": 19800
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.19860426219108e-06,
+      "loss": 0.4531,
+      "step": 19801
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.198027511030366e-06,
+      "loss": 0.4804,
+      "step": 19802
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.197450759869655e-06,
+      "loss": 0.4854,
+      "step": 19803
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.196874008708944e-06,
+      "loss": 0.5217,
+      "step": 19804
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.196297257548232e-06,
+      "loss": 0.4326,
+      "step": 19805
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.19572050638752e-06,
+      "loss": 0.4063,
+      "step": 19806
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.195143755226809e-06,
+      "loss": 0.4331,
+      "step": 19807
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.194567004066097e-06,
+      "loss": 0.5445,
+      "step": 19808
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.193990252905384e-06,
+      "loss": 0.3986,
+      "step": 19809
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.193413501744674e-06,
+      "loss": 0.454,
+      "step": 19810
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.192836750583961e-06,
+      "loss": 0.5011,
+      "step": 19811
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.192259999423249e-06,
+      "loss": 0.4514,
+      "step": 19812
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.191683248262537e-06,
+      "loss": 0.3898,
+      "step": 19813
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.191106497101826e-06,
+      "loss": 0.3852,
+      "step": 19814
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.190529745941114e-06,
+      "loss": 0.5736,
+      "step": 19815
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.189952994780402e-06,
+      "loss": 0.5034,
+      "step": 19816
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.189376243619691e-06,
+      "loss": 0.3845,
+      "step": 19817
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.18879949245898e-06,
+      "loss": 0.4971,
+      "step": 19818
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.188222741298268e-06,
+      "loss": 0.5083,
+      "step": 19819
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.187645990137556e-06,
+      "loss": 0.3806,
+      "step": 19820
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.187069238976845e-06,
+      "loss": 0.3446,
+      "step": 19821
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.186492487816133e-06,
+      "loss": 0.384,
+      "step": 19822
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.18591573665542e-06,
+      "loss": 0.4711,
+      "step": 19823
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.18533898549471e-06,
+      "loss": 0.3437,
+      "step": 19824
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.184762234333998e-06,
+      "loss": 0.4588,
+      "step": 19825
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.184185483173285e-06,
+      "loss": 0.4305,
+      "step": 19826
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.183608732012575e-06,
+      "loss": 0.5016,
+      "step": 19827
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.183031980851862e-06,
+      "loss": 0.5125,
+      "step": 19828
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.18245522969115e-06,
+      "loss": 0.2437,
+      "step": 19829
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.18187847853044e-06,
+      "loss": 0.307,
+      "step": 19830
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.181301727369727e-06,
+      "loss": 0.5043,
+      "step": 19831
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.180724976209016e-06,
+      "loss": 0.425,
+      "step": 19832
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.180148225048302e-06,
+      "loss": 0.3796,
+      "step": 19833
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.179571473887592e-06,
+      "loss": 0.3869,
+      "step": 19834
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.178994722726881e-06,
+      "loss": 0.5118,
+      "step": 19835
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.178417971566169e-06,
+      "loss": 0.4166,
+      "step": 19836
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.177841220405457e-06,
+      "loss": 0.4437,
+      "step": 19837
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.177264469244746e-06,
+      "loss": 0.4495,
+      "step": 19838
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176687718084034e-06,
+      "loss": 0.4774,
+      "step": 19839
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.176110966923321e-06,
+      "loss": 0.4011,
+      "step": 19840
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.17553421576261e-06,
+      "loss": 0.4643,
+      "step": 19841
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.174957464601898e-06,
+      "loss": 0.376,
+      "step": 19842
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.174380713441186e-06,
+      "loss": 0.4616,
+      "step": 19843
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.173803962280475e-06,
+      "loss": 0.3875,
+      "step": 19844
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.173227211119763e-06,
+      "loss": 0.3847,
+      "step": 19845
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.17265045995905e-06,
+      "loss": 0.451,
+      "step": 19846
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.17207370879834e-06,
+      "loss": 0.3726,
+      "step": 19847
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.171496957637628e-06,
+      "loss": 0.4202,
+      "step": 19848
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.170920206476917e-06,
+      "loss": 0.5045,
+      "step": 19849
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.170343455316205e-06,
+      "loss": 0.3522,
+      "step": 19850
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.169766704155493e-06,
+      "loss": 0.4995,
+      "step": 19851
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.169189952994782e-06,
+      "loss": 0.5066,
+      "step": 19852
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.16861320183407e-06,
+      "loss": 0.4137,
+      "step": 19853
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.168036450673357e-06,
+      "loss": 0.483,
+      "step": 19854
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.167459699512647e-06,
+      "loss": 0.5117,
+      "step": 19855
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.166882948351934e-06,
+      "loss": 0.4505,
+      "step": 19856
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.166306197191222e-06,
+      "loss": 0.4004,
+      "step": 19857
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.165729446030511e-06,
+      "loss": 0.5135,
+      "step": 19858
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.165152694869799e-06,
+      "loss": 0.4225,
+      "step": 19859
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.164575943709087e-06,
+      "loss": 0.4162,
+      "step": 19860
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.163999192548376e-06,
+      "loss": 0.409,
+      "step": 19861
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.163422441387664e-06,
+      "loss": 0.436,
+      "step": 19862
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.162845690226953e-06,
+      "loss": 0.351,
+      "step": 19863
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.162268939066241e-06,
+      "loss": 0.4121,
+      "step": 19864
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.161692187905529e-06,
+      "loss": 0.395,
+      "step": 19865
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.161115436744818e-06,
+      "loss": 0.4319,
+      "step": 19866
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.160538685584106e-06,
+      "loss": 0.5183,
+      "step": 19867
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.159961934423393e-06,
+      "loss": 0.4757,
+      "step": 19868
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.159385183262683e-06,
+      "loss": 0.391,
+      "step": 19869
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.15880843210197e-06,
+      "loss": 0.36,
+      "step": 19870
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.158231680941258e-06,
+      "loss": 0.3421,
+      "step": 19871
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.157654929780548e-06,
+      "loss": 0.5801,
+      "step": 19872
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.157078178619835e-06,
+      "loss": 0.5445,
+      "step": 19873
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.156501427459123e-06,
+      "loss": 0.4508,
+      "step": 19874
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.155924676298412e-06,
+      "loss": 0.4782,
+      "step": 19875
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.1553479251377e-06,
+      "loss": 0.5363,
+      "step": 19876
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.154771173976988e-06,
+      "loss": 0.3942,
+      "step": 19877
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.154194422816277e-06,
+      "loss": 0.4582,
+      "step": 19878
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.153617671655565e-06,
+      "loss": 0.4227,
+      "step": 19879
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.153040920494854e-06,
+      "loss": 0.55,
+      "step": 19880
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.152464169334142e-06,
+      "loss": 0.5017,
+      "step": 19881
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.15188741817343e-06,
+      "loss": 0.4872,
+      "step": 19882
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.151310667012719e-06,
+      "loss": 0.471,
+      "step": 19883
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.150733915852007e-06,
+      "loss": 0.5504,
+      "step": 19884
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.150157164691294e-06,
+      "loss": 0.3193,
+      "step": 19885
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.149580413530584e-06,
+      "loss": 0.5351,
+      "step": 19886
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.149003662369871e-06,
+      "loss": 0.3517,
+      "step": 19887
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.148426911209159e-06,
+      "loss": 0.3512,
+      "step": 19888
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.147850160048448e-06,
+      "loss": 0.356,
+      "step": 19889
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.147273408887736e-06,
+      "loss": 0.5231,
+      "step": 19890
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.146696657727024e-06,
+      "loss": 0.5269,
+      "step": 19891
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.146119906566313e-06,
+      "loss": 0.4728,
+      "step": 19892
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.1455431554056e-06,
+      "loss": 0.4061,
+      "step": 19893
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.14496640424489e-06,
+      "loss": 0.4654,
+      "step": 19894
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.144389653084178e-06,
+      "loss": 0.4253,
+      "step": 19895
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.143812901923466e-06,
+      "loss": 0.4541,
+      "step": 19896
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.143236150762755e-06,
+      "loss": 0.4643,
+      "step": 19897
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.142659399602043e-06,
+      "loss": 0.4248,
+      "step": 19898
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.14208264844133e-06,
+      "loss": 0.4421,
+      "step": 19899
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.14150589728062e-06,
+      "loss": 0.4998,
+      "step": 19900
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.140929146119907e-06,
+      "loss": 0.4735,
+      "step": 19901
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.140352394959195e-06,
+      "loss": 0.5805,
+      "step": 19902
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.139775643798484e-06,
+      "loss": 0.4215,
+      "step": 19903
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.139198892637772e-06,
+      "loss": 0.4053,
+      "step": 19904
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.13862214147706e-06,
+      "loss": 0.5306,
+      "step": 19905
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.138045390316349e-06,
+      "loss": 0.4696,
+      "step": 19906
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.137468639155637e-06,
+      "loss": 0.5258,
+      "step": 19907
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.136891887994925e-06,
+      "loss": 0.4186,
+      "step": 19908
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.136315136834214e-06,
+      "loss": 0.4952,
+      "step": 19909
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135738385673502e-06,
+      "loss": 0.4863,
+      "step": 19910
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.135161634512791e-06,
+      "loss": 0.4056,
+      "step": 19911
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.134584883352079e-06,
+      "loss": 0.447,
+      "step": 19912
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.134008132191366e-06,
+      "loss": 0.4349,
+      "step": 19913
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.133431381030656e-06,
+      "loss": 0.4261,
+      "step": 19914
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.132854629869943e-06,
+      "loss": 0.3178,
+      "step": 19915
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.132277878709231e-06,
+      "loss": 0.5692,
+      "step": 19916
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.13170112754852e-06,
+      "loss": 0.4335,
+      "step": 19917
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.131124376387808e-06,
+      "loss": 0.4113,
+      "step": 19918
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.130547625227096e-06,
+      "loss": 0.3504,
+      "step": 19919
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.129970874066385e-06,
+      "loss": 0.3914,
+      "step": 19920
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.129394122905673e-06,
+      "loss": 0.3634,
+      "step": 19921
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.12881737174496e-06,
+      "loss": 0.3369,
+      "step": 19922
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.12824062058425e-06,
+      "loss": 0.3612,
+      "step": 19923
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.127663869423538e-06,
+      "loss": 0.3454,
+      "step": 19924
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.127087118262827e-06,
+      "loss": 0.4594,
+      "step": 19925
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.126510367102115e-06,
+      "loss": 0.5529,
+      "step": 19926
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.125933615941402e-06,
+      "loss": 0.5069,
+      "step": 19927
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.125356864780692e-06,
+      "loss": 0.4351,
+      "step": 19928
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.12478011361998e-06,
+      "loss": 0.4545,
+      "step": 19929
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 9.124203362459267e-06,
+      "loss": 0.3781,
+      "step": 19930
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.123626611298557e-06,
+      "loss": 0.4514,
+      "step": 19931
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.123049860137844e-06,
+      "loss": 0.4973,
+      "step": 19932
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.122473108977132e-06,
+      "loss": 0.5035,
+      "step": 19933
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.121896357816421e-06,
+      "loss": 0.5345,
+      "step": 19934
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.121319606655709e-06,
+      "loss": 0.4683,
+      "step": 19935
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.120742855494997e-06,
+      "loss": 0.4451,
+      "step": 19936
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.120166104334286e-06,
+      "loss": 0.4134,
+      "step": 19937
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.119589353173574e-06,
+      "loss": 0.4403,
+      "step": 19938
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.119012602012861e-06,
+      "loss": 0.5655,
+      "step": 19939
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11843585085215e-06,
+      "loss": 0.4385,
+      "step": 19940
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.117859099691438e-06,
+      "loss": 0.4733,
+      "step": 19941
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.117282348530728e-06,
+      "loss": 0.3758,
+      "step": 19942
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.116705597370016e-06,
+      "loss": 0.3773,
+      "step": 19943
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.116128846209303e-06,
+      "loss": 0.5321,
+      "step": 19944
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.115552095048593e-06,
+      "loss": 0.3984,
+      "step": 19945
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11497534388788e-06,
+      "loss": 0.5124,
+      "step": 19946
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.114398592727168e-06,
+      "loss": 0.4384,
+      "step": 19947
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.113821841566457e-06,
+      "loss": 0.5271,
+      "step": 19948
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.113245090405745e-06,
+      "loss": 0.4755,
+      "step": 19949
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.112668339245033e-06,
+      "loss": 0.3299,
+      "step": 19950
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.112091588084322e-06,
+      "loss": 0.4496,
+      "step": 19951
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.11151483692361e-06,
+      "loss": 0.5113,
+      "step": 19952
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.110938085762897e-06,
+      "loss": 0.4165,
+      "step": 19953
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.110361334602187e-06,
+      "loss": 0.4,
+      "step": 19954
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.109784583441475e-06,
+      "loss": 0.4281,
+      "step": 19955
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.109207832280764e-06,
+      "loss": 0.4254,
+      "step": 19956
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.108631081120052e-06,
+      "loss": 0.4607,
+      "step": 19957
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.10805432995934e-06,
+      "loss": 0.3374,
+      "step": 19958
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.107477578798629e-06,
+      "loss": 0.4202,
+      "step": 19959
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.106900827637916e-06,
+      "loss": 0.5575,
+      "step": 19960
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.106324076477204e-06,
+      "loss": 0.4821,
+      "step": 19961
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.105747325316493e-06,
+      "loss": 0.4522,
+      "step": 19962
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.105170574155781e-06,
+      "loss": 0.4295,
+      "step": 19963
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.104593822995069e-06,
+      "loss": 0.4639,
+      "step": 19964
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.104017071834358e-06,
+      "loss": 0.4772,
+      "step": 19965
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.103440320673646e-06,
+      "loss": 0.4498,
+      "step": 19966
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.102863569512934e-06,
+      "loss": 0.3791,
+      "step": 19967
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.102286818352223e-06,
+      "loss": 0.5,
+      "step": 19968
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.101710067191512e-06,
+      "loss": 0.4876,
+      "step": 19969
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.101133316030798e-06,
+      "loss": 0.3877,
+      "step": 19970
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.100556564870088e-06,
+      "loss": 0.4717,
+      "step": 19971
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.099979813709375e-06,
+      "loss": 0.4988,
+      "step": 19972
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.099403062548665e-06,
+      "loss": 0.4494,
+      "step": 19973
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.098826311387952e-06,
+      "loss": 0.3483,
+      "step": 19974
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.09824956022724e-06,
+      "loss": 0.4681,
+      "step": 19975
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.09767280906653e-06,
+      "loss": 0.4348,
+      "step": 19976
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.097096057905817e-06,
+      "loss": 0.4163,
+      "step": 19977
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.096519306745105e-06,
+      "loss": 0.4457,
+      "step": 19978
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.095942555584394e-06,
+      "loss": 0.5027,
+      "step": 19979
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.095365804423682e-06,
+      "loss": 0.3962,
+      "step": 19980
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.09478905326297e-06,
+      "loss": 0.5437,
+      "step": 19981
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.094212302102259e-06,
+      "loss": 0.3946,
+      "step": 19982
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.093635550941548e-06,
+      "loss": 0.4012,
+      "step": 19983
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.093058799780834e-06,
+      "loss": 0.5266,
+      "step": 19984
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.092482048620124e-06,
+      "loss": 0.3444,
+      "step": 19985
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.091905297459413e-06,
+      "loss": 0.4277,
+      "step": 19986
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.0913285462987e-06,
+      "loss": 0.3873,
+      "step": 19987
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.090751795137988e-06,
+      "loss": 0.4659,
+      "step": 19988
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.090175043977276e-06,
+      "loss": 0.3907,
+      "step": 19989
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.089598292816566e-06,
+      "loss": 0.4652,
+      "step": 19990
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.089021541655853e-06,
+      "loss": 0.4229,
+      "step": 19991
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.088444790495141e-06,
+      "loss": 0.4618,
+      "step": 19992
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.08786803933443e-06,
+      "loss": 0.4496,
+      "step": 19993
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.087291288173718e-06,
+      "loss": 0.436,
+      "step": 19994
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.086714537013006e-06,
+      "loss": 0.3997,
+      "step": 19995
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.086137785852295e-06,
+      "loss": 0.5917,
+      "step": 19996
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.085561034691583e-06,
+      "loss": 0.5232,
+      "step": 19997
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.08498428353087e-06,
+      "loss": 0.4393,
+      "step": 19998
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.08440753237016e-06,
+      "loss": 0.3434,
+      "step": 19999
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.083830781209449e-06,
+      "loss": 0.4515,
+      "step": 20000
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.083254030048735e-06,
+      "loss": 0.4481,
+      "step": 20001
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.082677278888025e-06,
+      "loss": 0.4533,
+      "step": 20002
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.082100527727314e-06,
+      "loss": 0.4146,
+      "step": 20003
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.081523776566602e-06,
+      "loss": 0.3607,
+      "step": 20004
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.08094702540589e-06,
+      "loss": 0.4043,
+      "step": 20005
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.080370274245177e-06,
+      "loss": 0.4394,
+      "step": 20006
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.079793523084466e-06,
+      "loss": 0.4712,
+      "step": 20007
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.079216771923754e-06,
+      "loss": 0.4158,
+      "step": 20008
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.078640020763042e-06,
+      "loss": 0.407,
+      "step": 20009
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.078063269602331e-06,
+      "loss": 0.4996,
+      "step": 20010
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.077486518441619e-06,
+      "loss": 0.3651,
+      "step": 20011
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.076909767280906e-06,
+      "loss": 0.4516,
+      "step": 20012
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.076333016120196e-06,
+      "loss": 0.3939,
+      "step": 20013
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.075756264959485e-06,
+      "loss": 0.4643,
+      "step": 20014
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.075179513798771e-06,
+      "loss": 0.4708,
+      "step": 20015
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.07460276263806e-06,
+      "loss": 0.3525,
+      "step": 20016
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.07402601147735e-06,
+      "loss": 0.4133,
+      "step": 20017
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.073449260316638e-06,
+      "loss": 0.4527,
+      "step": 20018
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.072872509155925e-06,
+      "loss": 0.3208,
+      "step": 20019
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.072295757995215e-06,
+      "loss": 0.5055,
+      "step": 20020
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.071719006834502e-06,
+      "loss": 0.4155,
+      "step": 20021
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.07114225567379e-06,
+      "loss": 0.4694,
+      "step": 20022
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.070565504513078e-06,
+      "loss": 0.4281,
+      "step": 20023
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.069988753352367e-06,
+      "loss": 0.501,
+      "step": 20024
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.069412002191655e-06,
+      "loss": 0.3938,
+      "step": 20025
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.068835251030943e-06,
+      "loss": 0.4417,
+      "step": 20026
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.068258499870232e-06,
+      "loss": 0.5257,
+      "step": 20027
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.06768174870952e-06,
+      "loss": 0.4272,
+      "step": 20028
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.067104997548807e-06,
+      "loss": 0.4203,
+      "step": 20029
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.066528246388097e-06,
+      "loss": 0.6136,
+      "step": 20030
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.065951495227386e-06,
+      "loss": 0.4383,
+      "step": 20031
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.065374744066672e-06,
+      "loss": 0.4897,
+      "step": 20032
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.064797992905961e-06,
+      "loss": 0.4809,
+      "step": 20033
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.06422124174525e-06,
+      "loss": 0.4279,
+      "step": 20034
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.063644490584538e-06,
+      "loss": 0.501,
+      "step": 20035
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.063067739423826e-06,
+      "loss": 0.6064,
+      "step": 20036
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.062490988263116e-06,
+      "loss": 0.4305,
+      "step": 20037
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.061914237102403e-06,
+      "loss": 0.4312,
+      "step": 20038
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.061337485941691e-06,
+      "loss": 0.4352,
+      "step": 20039
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.060760734780979e-06,
+      "loss": 0.4302,
+      "step": 20040
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.060183983620268e-06,
+      "loss": 0.3691,
+      "step": 20041
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.059607232459556e-06,
+      "loss": 0.485,
+      "step": 20042
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.059030481298843e-06,
+      "loss": 0.353,
+      "step": 20043
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.058453730138133e-06,
+      "loss": 0.3176,
+      "step": 20044
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.057876978977422e-06,
+      "loss": 0.4314,
+      "step": 20045
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.057300227816708e-06,
+      "loss": 0.4157,
+      "step": 20046
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.056723476655997e-06,
+      "loss": 0.4216,
+      "step": 20047
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.056146725495287e-06,
+      "loss": 0.4326,
+      "step": 20048
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.055569974334575e-06,
+      "loss": 0.4021,
+      "step": 20049
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.054993223173862e-06,
+      "loss": 0.551,
+      "step": 20050
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.054416472013152e-06,
+      "loss": 0.4343,
+      "step": 20051
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.05383972085244e-06,
+      "loss": 0.4644,
+      "step": 20052
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.053262969691727e-06,
+      "loss": 0.4145,
+      "step": 20053
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.052686218531016e-06,
+      "loss": 0.4784,
+      "step": 20054
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.052109467370304e-06,
+      "loss": 0.5206,
+      "step": 20055
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.051532716209592e-06,
+      "loss": 0.355,
+      "step": 20056
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.05095596504888e-06,
+      "loss": 0.5225,
+      "step": 20057
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.050379213888169e-06,
+      "loss": 0.4178,
+      "step": 20058
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.049802462727456e-06,
+      "loss": 0.2878,
+      "step": 20059
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.049225711566744e-06,
+      "loss": 0.4457,
+      "step": 20060
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.048648960406034e-06,
+      "loss": 0.5203,
+      "step": 20061
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.048072209245323e-06,
+      "loss": 0.3782,
+      "step": 20062
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.047495458084609e-06,
+      "loss": 0.4551,
+      "step": 20063
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.046918706923898e-06,
+      "loss": 0.5032,
+      "step": 20064
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.046341955763188e-06,
+      "loss": 0.3936,
+      "step": 20065
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.045765204602475e-06,
+      "loss": 0.3939,
+      "step": 20066
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.045188453441763e-06,
+      "loss": 0.4458,
+      "step": 20067
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.044611702281052e-06,
+      "loss": 0.3751,
+      "step": 20068
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.04403495112034e-06,
+      "loss": 0.4855,
+      "step": 20069
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.043458199959628e-06,
+      "loss": 0.482,
+      "step": 20070
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.042881448798917e-06,
+      "loss": 0.5341,
+      "step": 20071
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.042304697638205e-06,
+      "loss": 0.3508,
+      "step": 20072
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.041727946477493e-06,
+      "loss": 0.4962,
+      "step": 20073
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.04115119531678e-06,
+      "loss": 0.5237,
+      "step": 20074
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.04057444415607e-06,
+      "loss": 0.4819,
+      "step": 20075
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.039997692995359e-06,
+      "loss": 0.504,
+      "step": 20076
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.039420941834645e-06,
+      "loss": 0.5426,
+      "step": 20077
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.038844190673934e-06,
+      "loss": 0.4254,
+      "step": 20078
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.038267439513224e-06,
+      "loss": 0.4129,
+      "step": 20079
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.037690688352511e-06,
+      "loss": 0.4999,
+      "step": 20080
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.037113937191799e-06,
+      "loss": 0.3733,
+      "step": 20081
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.036537186031088e-06,
+      "loss": 0.4126,
+      "step": 20082
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.035960434870376e-06,
+      "loss": 0.5434,
+      "step": 20083
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.035383683709664e-06,
+      "loss": 0.3924,
+      "step": 20084
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.034806932548953e-06,
+      "loss": 0.4426,
+      "step": 20085
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.034230181388241e-06,
+      "loss": 0.547,
+      "step": 20086
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.033653430227529e-06,
+      "loss": 0.4951,
+      "step": 20087
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.033076679066818e-06,
+      "loss": 0.4593,
+      "step": 20088
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.032499927906106e-06,
+      "loss": 0.4488,
+      "step": 20089
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.031923176745393e-06,
+      "loss": 0.5227,
+      "step": 20090
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.031346425584683e-06,
+      "loss": 0.4155,
+      "step": 20091
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.03076967442397e-06,
+      "loss": 0.5017,
+      "step": 20092
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.03019292326326e-06,
+      "loss": 0.5182,
+      "step": 20093
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.029616172102546e-06,
+      "loss": 0.3674,
+      "step": 20094
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.029039420941835e-06,
+      "loss": 0.4935,
+      "step": 20095
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.028462669781125e-06,
+      "loss": 0.4193,
+      "step": 20096
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.027885918620412e-06,
+      "loss": 0.4332,
+      "step": 20097
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.0273091674597e-06,
+      "loss": 0.434,
+      "step": 20098
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.02673241629899e-06,
+      "loss": 0.4245,
+      "step": 20099
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.026155665138277e-06,
+      "loss": 0.4502,
+      "step": 20100
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.025578913977565e-06,
+      "loss": 0.4221,
+      "step": 20101
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.025002162816854e-06,
+      "loss": 0.4938,
+      "step": 20102
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.024425411656142e-06,
+      "loss": 0.4764,
+      "step": 20103
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.02384866049543e-06,
+      "loss": 0.4502,
+      "step": 20104
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.023271909334719e-06,
+      "loss": 0.6106,
+      "step": 20105
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.022695158174006e-06,
+      "loss": 0.267,
+      "step": 20106
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.022118407013296e-06,
+      "loss": 0.3927,
+      "step": 20107
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.021541655852584e-06,
+      "loss": 0.4436,
+      "step": 20108
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 9.020964904691871e-06,
+      "loss": 0.505,
+      "step": 20109
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.02038815353116e-06,
+      "loss": 0.3212,
+      "step": 20110
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.019811402370448e-06,
+      "loss": 0.5014,
+      "step": 20111
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.019234651209736e-06,
+      "loss": 0.4169,
+      "step": 20112
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.018657900049025e-06,
+      "loss": 0.547,
+      "step": 20113
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.018081148888313e-06,
+      "loss": 0.3577,
+      "step": 20114
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.0175043977276e-06,
+      "loss": 0.5297,
+      "step": 20115
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.01692764656689e-06,
+      "loss": 0.4175,
+      "step": 20116
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.016350895406178e-06,
+      "loss": 0.4769,
+      "step": 20117
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.015774144245465e-06,
+      "loss": 0.4893,
+      "step": 20118
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.015197393084755e-06,
+      "loss": 0.4621,
+      "step": 20119
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.014620641924043e-06,
+      "loss": 0.5594,
+      "step": 20120
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.01404389076333e-06,
+      "loss": 0.4034,
+      "step": 20121
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.01346713960262e-06,
+      "loss": 0.3867,
+      "step": 20122
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012890388441907e-06,
+      "loss": 0.5038,
+      "step": 20123
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.012313637281197e-06,
+      "loss": 0.4863,
+      "step": 20124
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.011736886120484e-06,
+      "loss": 0.438,
+      "step": 20125
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.011160134959772e-06,
+      "loss": 0.4429,
+      "step": 20126
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.010583383799061e-06,
+      "loss": 0.403,
+      "step": 20127
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.010006632638349e-06,
+      "loss": 0.5253,
+      "step": 20128
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.009429881477637e-06,
+      "loss": 0.3291,
+      "step": 20129
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.008853130316926e-06,
+      "loss": 0.402,
+      "step": 20130
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.008276379156214e-06,
+      "loss": 0.5609,
+      "step": 20131
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.007699627995502e-06,
+      "loss": 0.5314,
+      "step": 20132
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.007122876834791e-06,
+      "loss": 0.3551,
+      "step": 20133
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.006546125674079e-06,
+      "loss": 0.3594,
+      "step": 20134
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.005969374513366e-06,
+      "loss": 0.4221,
+      "step": 20135
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.005392623352656e-06,
+      "loss": 0.4605,
+      "step": 20136
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.004815872191943e-06,
+      "loss": 0.3756,
+      "step": 20137
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.004239121031233e-06,
+      "loss": 0.4123,
+      "step": 20138
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.00366236987052e-06,
+      "loss": 0.4041,
+      "step": 20139
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.003085618709808e-06,
+      "loss": 0.501,
+      "step": 20140
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.002508867549097e-06,
+      "loss": 0.3624,
+      "step": 20141
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.001932116388385e-06,
+      "loss": 0.4215,
+      "step": 20142
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.001355365227673e-06,
+      "loss": 0.4444,
+      "step": 20143
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.000778614066962e-06,
+      "loss": 0.501,
+      "step": 20144
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 9.00020186290625e-06,
+      "loss": 0.399,
+      "step": 20145
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.999625111745538e-06,
+      "loss": 0.5058,
+      "step": 20146
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.999048360584827e-06,
+      "loss": 0.4119,
+      "step": 20147
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.998471609424115e-06,
+      "loss": 0.481,
+      "step": 20148
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.997894858263402e-06,
+      "loss": 0.3891,
+      "step": 20149
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.997318107102692e-06,
+      "loss": 0.454,
+      "step": 20150
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.99674135594198e-06,
+      "loss": 0.4403,
+      "step": 20151
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.996164604781267e-06,
+      "loss": 0.5342,
+      "step": 20152
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.995587853620556e-06,
+      "loss": 0.5038,
+      "step": 20153
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.995011102459844e-06,
+      "loss": 0.371,
+      "step": 20154
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.994434351299134e-06,
+      "loss": 0.3468,
+      "step": 20155
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.993857600138421e-06,
+      "loss": 0.457,
+      "step": 20156
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.993280848977709e-06,
+      "loss": 0.4264,
+      "step": 20157
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.992704097816998e-06,
+      "loss": 0.4825,
+      "step": 20158
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.992127346656286e-06,
+      "loss": 0.4086,
+      "step": 20159
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.991550595495574e-06,
+      "loss": 0.4401,
+      "step": 20160
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.990973844334863e-06,
+      "loss": 0.554,
+      "step": 20161
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.99039709317415e-06,
+      "loss": 0.4485,
+      "step": 20162
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.989820342013438e-06,
+      "loss": 0.3812,
+      "step": 20163
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.989243590852728e-06,
+      "loss": 0.4117,
+      "step": 20164
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.988666839692015e-06,
+      "loss": 0.4494,
+      "step": 20165
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.988090088531303e-06,
+      "loss": 0.4914,
+      "step": 20166
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.987513337370593e-06,
+      "loss": 0.349,
+      "step": 20167
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.98693658620988e-06,
+      "loss": 0.3553,
+      "step": 20168
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.98635983504917e-06,
+      "loss": 0.401,
+      "step": 20169
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.985783083888457e-06,
+      "loss": 0.4486,
+      "step": 20170
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.985206332727745e-06,
+      "loss": 0.4802,
+      "step": 20171
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.984629581567034e-06,
+      "loss": 0.486,
+      "step": 20172
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.984052830406322e-06,
+      "loss": 0.468,
+      "step": 20173
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.98347607924561e-06,
+      "loss": 0.4831,
+      "step": 20174
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.982899328084899e-06,
+      "loss": 0.4404,
+      "step": 20175
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.982322576924187e-06,
+      "loss": 0.3287,
+      "step": 20176
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.981745825763474e-06,
+      "loss": 0.3592,
+      "step": 20177
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.981169074602764e-06,
+      "loss": 0.519,
+      "step": 20178
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.980592323442051e-06,
+      "loss": 0.4361,
+      "step": 20179
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.98001557228134e-06,
+      "loss": 0.4859,
+      "step": 20180
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.979438821120629e-06,
+      "loss": 0.3588,
+      "step": 20181
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.978862069959916e-06,
+      "loss": 0.4408,
+      "step": 20182
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.978285318799204e-06,
+      "loss": 0.4759,
+      "step": 20183
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.977708567638493e-06,
+      "loss": 0.5403,
+      "step": 20184
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.977131816477781e-06,
+      "loss": 0.3013,
+      "step": 20185
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97655506531707e-06,
+      "loss": 0.4687,
+      "step": 20186
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.975978314156358e-06,
+      "loss": 0.4055,
+      "step": 20187
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.975401562995646e-06,
+      "loss": 0.4781,
+      "step": 20188
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.974824811834935e-06,
+      "loss": 0.3704,
+      "step": 20189
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.974248060674223e-06,
+      "loss": 0.3931,
+      "step": 20190
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97367130951351e-06,
+      "loss": 0.4585,
+      "step": 20191
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.9730945583528e-06,
+      "loss": 0.5256,
+      "step": 20192
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.972517807192088e-06,
+      "loss": 0.4319,
+      "step": 20193
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.971941056031375e-06,
+      "loss": 0.4235,
+      "step": 20194
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.971364304870665e-06,
+      "loss": 0.3245,
+      "step": 20195
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.970787553709952e-06,
+      "loss": 0.4819,
+      "step": 20196
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.97021080254924e-06,
+      "loss": 0.5467,
+      "step": 20197
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.96963405138853e-06,
+      "loss": 0.6236,
+      "step": 20198
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.969057300227817e-06,
+      "loss": 0.5508,
+      "step": 20199
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.968480549067106e-06,
+      "loss": 0.4329,
+      "step": 20200
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.967903797906394e-06,
+      "loss": 0.4205,
+      "step": 20201
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.967327046745682e-06,
+      "loss": 0.4996,
+      "step": 20202
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.966750295584971e-06,
+      "loss": 0.4772,
+      "step": 20203
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.966173544424259e-06,
+      "loss": 0.4516,
+      "step": 20204
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.965596793263547e-06,
+      "loss": 0.4682,
+      "step": 20205
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.965020042102836e-06,
+      "loss": 0.3805,
+      "step": 20206
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.964443290942124e-06,
+      "loss": 0.4895,
+      "step": 20207
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.963866539781411e-06,
+      "loss": 0.5066,
+      "step": 20208
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.9632897886207e-06,
+      "loss": 0.4067,
+      "step": 20209
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.962713037459988e-06,
+      "loss": 0.3943,
+      "step": 20210
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.962136286299276e-06,
+      "loss": 0.408,
+      "step": 20211
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.961559535138565e-06,
+      "loss": 0.3951,
+      "step": 20212
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.960982783977855e-06,
+      "loss": 0.4426,
+      "step": 20213
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.96040603281714e-06,
+      "loss": 0.3952,
+      "step": 20214
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.95982928165643e-06,
+      "loss": 0.468,
+      "step": 20215
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.959252530495718e-06,
+      "loss": 0.4674,
+      "step": 20216
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.958675779335007e-06,
+      "loss": 0.4044,
+      "step": 20217
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.958099028174295e-06,
+      "loss": 0.3876,
+      "step": 20218
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.957522277013583e-06,
+      "loss": 0.4811,
+      "step": 20219
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.956945525852872e-06,
+      "loss": 0.6046,
+      "step": 20220
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.95636877469216e-06,
+      "loss": 0.3711,
+      "step": 20221
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.955792023531447e-06,
+      "loss": 0.4845,
+      "step": 20222
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.955215272370737e-06,
+      "loss": 0.4119,
+      "step": 20223
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.954638521210024e-06,
+      "loss": 0.4306,
+      "step": 20224
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.954061770049312e-06,
+      "loss": 0.5473,
+      "step": 20225
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.953485018888601e-06,
+      "loss": 0.34,
+      "step": 20226
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.95290826772789e-06,
+      "loss": 0.4366,
+      "step": 20227
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.952331516567177e-06,
+      "loss": 0.415,
+      "step": 20228
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.951754765406466e-06,
+      "loss": 0.3801,
+      "step": 20229
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.951178014245756e-06,
+      "loss": 0.4563,
+      "step": 20230
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950601263085043e-06,
+      "loss": 0.4616,
+      "step": 20231
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.950024511924331e-06,
+      "loss": 0.4576,
+      "step": 20232
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.949447760763619e-06,
+      "loss": 0.357,
+      "step": 20233
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.948871009602908e-06,
+      "loss": 0.5088,
+      "step": 20234
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.948294258442196e-06,
+      "loss": 0.5162,
+      "step": 20235
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.947717507281483e-06,
+      "loss": 0.4626,
+      "step": 20236
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.947140756120773e-06,
+      "loss": 0.4545,
+      "step": 20237
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.94656400496006e-06,
+      "loss": 0.5545,
+      "step": 20238
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.945987253799348e-06,
+      "loss": 0.4556,
+      "step": 20239
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.945410502638638e-06,
+      "loss": 0.4503,
+      "step": 20240
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.944833751477925e-06,
+      "loss": 0.4033,
+      "step": 20241
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.944257000317213e-06,
+      "loss": 0.5585,
+      "step": 20242
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.943680249156502e-06,
+      "loss": 0.512,
+      "step": 20243
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.943103497995792e-06,
+      "loss": 0.476,
+      "step": 20244
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.942526746835078e-06,
+      "loss": 0.3381,
+      "step": 20245
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.941949995674367e-06,
+      "loss": 0.4674,
+      "step": 20246
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.941373244513656e-06,
+      "loss": 0.5248,
+      "step": 20247
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.940796493352944e-06,
+      "loss": 0.3722,
+      "step": 20248
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.940219742192232e-06,
+      "loss": 0.5584,
+      "step": 20249
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.93964299103152e-06,
+      "loss": 0.3849,
+      "step": 20250
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.939066239870809e-06,
+      "loss": 0.2544,
+      "step": 20251
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.938489488710097e-06,
+      "loss": 0.3569,
+      "step": 20252
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.937912737549384e-06,
+      "loss": 0.4257,
+      "step": 20253
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.937335986388674e-06,
+      "loss": 0.4936,
+      "step": 20254
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.936759235227961e-06,
+      "loss": 0.4921,
+      "step": 20255
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.936182484067249e-06,
+      "loss": 0.5145,
+      "step": 20256
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.935605732906538e-06,
+      "loss": 0.5052,
+      "step": 20257
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.935028981745826e-06,
+      "loss": 0.4103,
+      "step": 20258
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.934452230585114e-06,
+      "loss": 0.4837,
+      "step": 20259
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.933875479424403e-06,
+      "loss": 0.4521,
+      "step": 20260
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.933298728263692e-06,
+      "loss": 0.4236,
+      "step": 20261
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.93272197710298e-06,
+      "loss": 0.4168,
+      "step": 20262
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.932145225942268e-06,
+      "loss": 0.35,
+      "step": 20263
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.931568474781557e-06,
+      "loss": 0.4013,
+      "step": 20264
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930991723620845e-06,
+      "loss": 0.4421,
+      "step": 20265
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.930414972460133e-06,
+      "loss": 0.3451,
+      "step": 20266
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.92983822129942e-06,
+      "loss": 0.4516,
+      "step": 20267
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.92926147013871e-06,
+      "loss": 0.4326,
+      "step": 20268
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.928684718977997e-06,
+      "loss": 0.3871,
+      "step": 20269
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.928107967817285e-06,
+      "loss": 0.4836,
+      "step": 20270
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.927531216656574e-06,
+      "loss": 0.3504,
+      "step": 20271
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.926954465495862e-06,
+      "loss": 0.4717,
+      "step": 20272
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.92637771433515e-06,
+      "loss": 0.3877,
+      "step": 20273
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.92580096317444e-06,
+      "loss": 0.543,
+      "step": 20274
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.925224212013729e-06,
+      "loss": 0.415,
+      "step": 20275
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.924647460853015e-06,
+      "loss": 0.4015,
+      "step": 20276
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.924070709692304e-06,
+      "loss": 0.3994,
+      "step": 20277
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.923493958531593e-06,
+      "loss": 0.3982,
+      "step": 20278
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.922917207370881e-06,
+      "loss": 0.4112,
+      "step": 20279
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.922340456210169e-06,
+      "loss": 0.3756,
+      "step": 20280
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.921763705049458e-06,
+      "loss": 0.4402,
+      "step": 20281
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.921186953888746e-06,
+      "loss": 0.4068,
+      "step": 20282
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.920610202728033e-06,
+      "loss": 0.5658,
+      "step": 20283
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.920033451567321e-06,
+      "loss": 0.4492,
+      "step": 20284
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.91945670040661e-06,
+      "loss": 0.4354,
+      "step": 20285
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.918879949245898e-06,
+      "loss": 0.4901,
+      "step": 20286
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.918303198085186e-06,
+      "loss": 0.4355,
+      "step": 20287
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 8.917726446924475e-06,
+      "loss": 0.3926,
+      "step": 20288
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.917149695763763e-06,
+      "loss": 0.4315,
+      "step": 20289
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.91657294460305e-06,
+      "loss": 0.5541,
+      "step": 20290
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.91599619344234e-06,
+      "loss": 0.4872,
+      "step": 20291
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.91541944228163e-06,
+      "loss": 0.429,
+      "step": 20292
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.914842691120917e-06,
+      "loss": 0.6031,
+      "step": 20293
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.914265939960205e-06,
+      "loss": 0.38,
+      "step": 20294
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.913689188799494e-06,
+      "loss": 0.3892,
+      "step": 20295
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.913112437638782e-06,
+      "loss": 0.3767,
+      "step": 20296
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.91253568647807e-06,
+      "loss": 0.4269,
+      "step": 20297
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.911958935317359e-06,
+      "loss": 0.4749,
+      "step": 20298
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.911382184156647e-06,
+      "loss": 0.471,
+      "step": 20299
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.910805432995934e-06,
+      "loss": 0.4931,
+      "step": 20300
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.910228681835222e-06,
+      "loss": 0.4456,
+      "step": 20301
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909651930674511e-06,
+      "loss": 0.477,
+      "step": 20302
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.909075179513799e-06,
+      "loss": 0.4705,
+      "step": 20303
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.908498428353087e-06,
+      "loss": 0.491,
+      "step": 20304
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.907921677192376e-06,
+      "loss": 0.4126,
+      "step": 20305
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.907344926031665e-06,
+      "loss": 0.3927,
+      "step": 20306
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.906768174870951e-06,
+      "loss": 0.402,
+      "step": 20307
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.90619142371024e-06,
+      "loss": 0.4067,
+      "step": 20308
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.90561467254953e-06,
+      "loss": 0.5575,
+      "step": 20309
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.905037921388818e-06,
+      "loss": 0.4127,
+      "step": 20310
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.904461170228106e-06,
+      "loss": 0.4425,
+      "step": 20311
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.903884419067395e-06,
+      "loss": 0.4974,
+      "step": 20312
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.903307667906683e-06,
+      "loss": 0.4484,
+      "step": 20313
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.90273091674597e-06,
+      "loss": 0.4469,
+      "step": 20314
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.90215416558526e-06,
+      "loss": 0.4009,
+      "step": 20315
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.901577414424547e-06,
+      "loss": 0.4667,
+      "step": 20316
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.901000663263835e-06,
+      "loss": 0.4627,
+      "step": 20317
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.900423912103123e-06,
+      "loss": 0.4528,
+      "step": 20318
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.899847160942412e-06,
+      "loss": 0.4995,
+      "step": 20319
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.8992704097817e-06,
+      "loss": 0.3858,
+      "step": 20320
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.898693658620987e-06,
+      "loss": 0.4247,
+      "step": 20321
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.898116907460277e-06,
+      "loss": 0.4516,
+      "step": 20322
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.897540156299566e-06,
+      "loss": 0.4257,
+      "step": 20323
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.896963405138854e-06,
+      "loss": 0.4101,
+      "step": 20324
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.896386653978142e-06,
+      "loss": 0.3635,
+      "step": 20325
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.895809902817431e-06,
+      "loss": 0.3857,
+      "step": 20326
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.895233151656719e-06,
+      "loss": 0.6122,
+      "step": 20327
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.894656400496006e-06,
+      "loss": 0.4329,
+      "step": 20328
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.894079649335296e-06,
+      "loss": 0.3759,
+      "step": 20329
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.893502898174583e-06,
+      "loss": 0.5398,
+      "step": 20330
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.892926147013871e-06,
+      "loss": 0.3219,
+      "step": 20331
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.89234939585316e-06,
+      "loss": 0.3362,
+      "step": 20332
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.891772644692448e-06,
+      "loss": 0.4354,
+      "step": 20333
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.891195893531736e-06,
+      "loss": 0.4022,
+      "step": 20334
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.890619142371025e-06,
+      "loss": 0.4952,
+      "step": 20335
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.890042391210313e-06,
+      "loss": 0.3699,
+      "step": 20336
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.889465640049602e-06,
+      "loss": 0.5386,
+      "step": 20337
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888888888888888e-06,
+      "loss": 0.4835,
+      "step": 20338
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.888312137728178e-06,
+      "loss": 0.4095,
+      "step": 20339
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.887735386567467e-06,
+      "loss": 0.5565,
+      "step": 20340
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.887158635406755e-06,
+      "loss": 0.3881,
+      "step": 20341
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.886581884246042e-06,
+      "loss": 0.4288,
+      "step": 20342
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.886005133085332e-06,
+      "loss": 0.4554,
+      "step": 20343
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.88542838192462e-06,
+      "loss": 0.5299,
+      "step": 20344
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.884851630763907e-06,
+      "loss": 0.4433,
+      "step": 20345
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.884274879603197e-06,
+      "loss": 0.548,
+      "step": 20346
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.883698128442484e-06,
+      "loss": 0.5714,
+      "step": 20347
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.883121377281772e-06,
+      "loss": 0.5606,
+      "step": 20348
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.882544626121061e-06,
+      "loss": 0.3931,
+      "step": 20349
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.881967874960349e-06,
+      "loss": 0.4581,
+      "step": 20350
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.881391123799638e-06,
+      "loss": 0.377,
+      "step": 20351
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.880814372638926e-06,
+      "loss": 0.4045,
+      "step": 20352
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.880237621478214e-06,
+      "loss": 0.3347,
+      "step": 20353
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.879660870317503e-06,
+      "loss": 0.4235,
+      "step": 20354
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.87908411915679e-06,
+      "loss": 0.4478,
+      "step": 20355
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.878507367996078e-06,
+      "loss": 0.4999,
+      "step": 20356
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.877930616835368e-06,
+      "loss": 0.4763,
+      "step": 20357
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.877353865674656e-06,
+      "loss": 0.5403,
+      "step": 20358
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.876777114513943e-06,
+      "loss": 0.406,
+      "step": 20359
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.876200363353233e-06,
+      "loss": 0.4722,
+      "step": 20360
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.87562361219252e-06,
+      "loss": 0.5479,
+      "step": 20361
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.875046861031808e-06,
+      "loss": 0.4077,
+      "step": 20362
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.874470109871097e-06,
+      "loss": 0.4556,
+      "step": 20363
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.873893358710385e-06,
+      "loss": 0.4242,
+      "step": 20364
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.873316607549673e-06,
+      "loss": 0.3509,
+      "step": 20365
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.872739856388962e-06,
+      "loss": 0.4163,
+      "step": 20366
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.87216310522825e-06,
+      "loss": 0.3688,
+      "step": 20367
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.87158635406754e-06,
+      "loss": 0.5113,
+      "step": 20368
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.871009602906827e-06,
+      "loss": 0.3953,
+      "step": 20369
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.870432851746115e-06,
+      "loss": 0.4183,
+      "step": 20370
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.869856100585404e-06,
+      "loss": 0.3887,
+      "step": 20371
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.869279349424692e-06,
+      "loss": 0.4835,
+      "step": 20372
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.86870259826398e-06,
+      "loss": 0.4353,
+      "step": 20373
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.868125847103269e-06,
+      "loss": 0.3979,
+      "step": 20374
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.867549095942556e-06,
+      "loss": 0.3308,
+      "step": 20375
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.866972344781844e-06,
+      "loss": 0.4441,
+      "step": 20376
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.866395593621133e-06,
+      "loss": 0.392,
+      "step": 20377
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.865818842460421e-06,
+      "loss": 0.3345,
+      "step": 20378
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.865242091299709e-06,
+      "loss": 0.5134,
+      "step": 20379
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.864665340138998e-06,
+      "loss": 0.4935,
+      "step": 20380
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.864088588978286e-06,
+      "loss": 0.4265,
+      "step": 20381
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.863511837817575e-06,
+      "loss": 0.4338,
+      "step": 20382
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.862935086656863e-06,
+      "loss": 0.4055,
+      "step": 20383
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.86235833549615e-06,
+      "loss": 0.4042,
+      "step": 20384
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.86178158433544e-06,
+      "loss": 0.4282,
+      "step": 20385
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.861204833174728e-06,
+      "loss": 0.4972,
+      "step": 20386
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.860628082014015e-06,
+      "loss": 0.3385,
+      "step": 20387
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.860051330853305e-06,
+      "loss": 0.4008,
+      "step": 20388
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.859474579692592e-06,
+      "loss": 0.3137,
+      "step": 20389
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.85889782853188e-06,
+      "loss": 0.4039,
+      "step": 20390
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.85832107737117e-06,
+      "loss": 0.4384,
+      "step": 20391
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.857744326210457e-06,
+      "loss": 0.4474,
+      "step": 20392
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.857167575049745e-06,
+      "loss": 0.31,
+      "step": 20393
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.856590823889034e-06,
+      "loss": 0.399,
+      "step": 20394
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.856014072728322e-06,
+      "loss": 0.4586,
+      "step": 20395
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.85543732156761e-06,
+      "loss": 0.3926,
+      "step": 20396
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.854860570406899e-06,
+      "loss": 0.5442,
+      "step": 20397
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.854283819246187e-06,
+      "loss": 0.495,
+      "step": 20398
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.853707068085476e-06,
+      "loss": 0.3476,
+      "step": 20399
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.853130316924764e-06,
+      "loss": 0.4011,
+      "step": 20400
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.852553565764051e-06,
+      "loss": 0.5282,
+      "step": 20401
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.85197681460334e-06,
+      "loss": 0.5227,
+      "step": 20402
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.851400063442628e-06,
+      "loss": 0.4178,
+      "step": 20403
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.850823312281916e-06,
+      "loss": 0.3938,
+      "step": 20404
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.850246561121206e-06,
+      "loss": 0.3139,
+      "step": 20405
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.849669809960493e-06,
+      "loss": 0.4239,
+      "step": 20406
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.849093058799781e-06,
+      "loss": 0.4468,
+      "step": 20407
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.84851630763907e-06,
+      "loss": 0.4605,
+      "step": 20408
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847939556478358e-06,
+      "loss": 0.5477,
+      "step": 20409
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.847362805317646e-06,
+      "loss": 0.5029,
+      "step": 20410
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.846786054156935e-06,
+      "loss": 0.4368,
+      "step": 20411
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.846209302996223e-06,
+      "loss": 0.5118,
+      "step": 20412
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.845632551835512e-06,
+      "loss": 0.4178,
+      "step": 20413
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.8450558006748e-06,
+      "loss": 0.3328,
+      "step": 20414
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.844479049514087e-06,
+      "loss": 0.4018,
+      "step": 20415
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.843902298353377e-06,
+      "loss": 0.4955,
+      "step": 20416
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.843325547192665e-06,
+      "loss": 0.6025,
+      "step": 20417
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.842748796031952e-06,
+      "loss": 0.3458,
+      "step": 20418
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.842172044871242e-06,
+      "loss": 0.5115,
+      "step": 20419
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.84159529371053e-06,
+      "loss": 0.406,
+      "step": 20420
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.841018542549817e-06,
+      "loss": 0.5197,
+      "step": 20421
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.840441791389106e-06,
+      "loss": 0.4528,
+      "step": 20422
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.839865040228394e-06,
+      "loss": 0.5466,
+      "step": 20423
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.839288289067682e-06,
+      "loss": 0.5041,
+      "step": 20424
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.838711537906971e-06,
+      "loss": 0.3631,
+      "step": 20425
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.838134786746259e-06,
+      "loss": 0.5048,
+      "step": 20426
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.837558035585546e-06,
+      "loss": 0.3822,
+      "step": 20427
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.836981284424836e-06,
+      "loss": 0.5336,
+      "step": 20428
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.836404533264124e-06,
+      "loss": 0.3694,
+      "step": 20429
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.835827782103413e-06,
+      "loss": 0.4463,
+      "step": 20430
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.8352510309427e-06,
+      "loss": 0.4191,
+      "step": 20431
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.834674279781988e-06,
+      "loss": 0.5061,
+      "step": 20432
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.834097528621278e-06,
+      "loss": 0.4212,
+      "step": 20433
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.833520777460565e-06,
+      "loss": 0.5213,
+      "step": 20434
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.832944026299853e-06,
+      "loss": 0.5409,
+      "step": 20435
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.832367275139142e-06,
+      "loss": 0.5869,
+      "step": 20436
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.83179052397843e-06,
+      "loss": 0.4967,
+      "step": 20437
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.831213772817718e-06,
+      "loss": 0.4084,
+      "step": 20438
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.830637021657007e-06,
+      "loss": 0.4203,
+      "step": 20439
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.830060270496295e-06,
+      "loss": 0.4892,
+      "step": 20440
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.829483519335583e-06,
+      "loss": 0.4057,
+      "step": 20441
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.828906768174872e-06,
+      "loss": 0.5005,
+      "step": 20442
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.82833001701416e-06,
+      "loss": 0.5418,
+      "step": 20443
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827753265853449e-06,
+      "loss": 0.4316,
+      "step": 20444
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.827176514692737e-06,
+      "loss": 0.3777,
+      "step": 20445
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.826599763532024e-06,
+      "loss": 0.5076,
+      "step": 20446
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.826023012371314e-06,
+      "loss": 0.4251,
+      "step": 20447
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.825446261210601e-06,
+      "loss": 0.4686,
+      "step": 20448
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.824869510049889e-06,
+      "loss": 0.4904,
+      "step": 20449
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.824292758889178e-06,
+      "loss": 0.4239,
+      "step": 20450
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.823716007728466e-06,
+      "loss": 0.4683,
+      "step": 20451
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.823139256567754e-06,
+      "loss": 0.3586,
+      "step": 20452
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.822562505407043e-06,
+      "loss": 0.424,
+      "step": 20453
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.821985754246331e-06,
+      "loss": 0.5603,
+      "step": 20454
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.821409003085619e-06,
+      "loss": 0.4114,
+      "step": 20455
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.820832251924908e-06,
+      "loss": 0.3387,
+      "step": 20456
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.820255500764196e-06,
+      "loss": 0.3709,
+      "step": 20457
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.819678749603483e-06,
+      "loss": 0.4006,
+      "step": 20458
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.819101998442773e-06,
+      "loss": 0.4775,
+      "step": 20459
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.81852524728206e-06,
+      "loss": 0.3484,
+      "step": 20460
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.81794849612135e-06,
+      "loss": 0.5464,
+      "step": 20461
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.817371744960637e-06,
+      "loss": 0.5462,
+      "step": 20462
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.816794993799925e-06,
+      "loss": 0.517,
+      "step": 20463
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.816218242639215e-06,
+      "loss": 0.4429,
+      "step": 20464
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.815641491478502e-06,
+      "loss": 0.3337,
+      "step": 20465
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 8.81506474031779e-06,
+      "loss": 0.4632,
+      "step": 20466
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.81448798915708e-06,
+      "loss": 0.4145,
+      "step": 20467
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.813911237996367e-06,
+      "loss": 0.4598,
+      "step": 20468
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.813334486835655e-06,
+      "loss": 0.5243,
+      "step": 20469
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.812757735674944e-06,
+      "loss": 0.4036,
+      "step": 20470
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.812180984514232e-06,
+      "loss": 0.4078,
+      "step": 20471
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.81160423335352e-06,
+      "loss": 0.5491,
+      "step": 20472
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.811027482192809e-06,
+      "loss": 0.5868,
+      "step": 20473
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.810450731032098e-06,
+      "loss": 0.488,
+      "step": 20474
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.809873979871386e-06,
+      "loss": 0.4141,
+      "step": 20475
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.809297228710674e-06,
+      "loss": 0.4286,
+      "step": 20476
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.808720477549961e-06,
+      "loss": 0.4349,
+      "step": 20477
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.80814372638925e-06,
+      "loss": 0.5335,
+      "step": 20478
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.807566975228538e-06,
+      "loss": 0.4487,
+      "step": 20479
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806990224067826e-06,
+      "loss": 0.3051,
+      "step": 20480
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.806413472907115e-06,
+      "loss": 0.6065,
+      "step": 20481
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.805836721746403e-06,
+      "loss": 0.4894,
+      "step": 20482
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.80525997058569e-06,
+      "loss": 0.462,
+      "step": 20483
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.80468321942498e-06,
+      "loss": 0.4496,
+      "step": 20484
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.804106468264268e-06,
+      "loss": 0.4845,
+      "step": 20485
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.803529717103555e-06,
+      "loss": 0.4005,
+      "step": 20486
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.802952965942845e-06,
+      "loss": 0.4997,
+      "step": 20487
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.802376214782134e-06,
+      "loss": 0.3612,
+      "step": 20488
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.80179946362142e-06,
+      "loss": 0.3023,
+      "step": 20489
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.80122271246071e-06,
+      "loss": 0.4262,
+      "step": 20490
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.800645961299999e-06,
+      "loss": 0.5023,
+      "step": 20491
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.800069210139287e-06,
+      "loss": 0.4159,
+      "step": 20492
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.799492458978574e-06,
+      "loss": 0.3985,
+      "step": 20493
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.798915707817862e-06,
+      "loss": 0.4236,
+      "step": 20494
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.798338956657151e-06,
+      "loss": 0.5074,
+      "step": 20495
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.797762205496439e-06,
+      "loss": 0.4712,
+      "step": 20496
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.797185454335727e-06,
+      "loss": 0.3311,
+      "step": 20497
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.796608703175016e-06,
+      "loss": 0.3927,
+      "step": 20498
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.796031952014304e-06,
+      "loss": 0.4969,
+      "step": 20499
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.795455200853592e-06,
+      "loss": 0.6002,
+      "step": 20500
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.794878449692881e-06,
+      "loss": 0.4128,
+      "step": 20501
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.794301698532169e-06,
+      "loss": 0.3383,
+      "step": 20502
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.793724947371456e-06,
+      "loss": 0.4335,
+      "step": 20503
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.793148196210746e-06,
+      "loss": 0.4914,
+      "step": 20504
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.792571445050035e-06,
+      "loss": 0.4569,
+      "step": 20505
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.791994693889323e-06,
+      "loss": 0.3927,
+      "step": 20506
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.79141794272861e-06,
+      "loss": 0.439,
+      "step": 20507
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.7908411915679e-06,
+      "loss": 0.5042,
+      "step": 20508
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.790264440407187e-06,
+      "loss": 0.4901,
+      "step": 20509
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.789687689246475e-06,
+      "loss": 0.4363,
+      "step": 20510
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.789110938085763e-06,
+      "loss": 0.407,
+      "step": 20511
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.788534186925052e-06,
+      "loss": 0.4702,
+      "step": 20512
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.78795743576434e-06,
+      "loss": 0.4414,
+      "step": 20513
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.787380684603628e-06,
+      "loss": 0.4818,
+      "step": 20514
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786803933442917e-06,
+      "loss": 0.4124,
+      "step": 20515
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.786227182282205e-06,
+      "loss": 0.5831,
+      "step": 20516
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.785650431121492e-06,
+      "loss": 0.4359,
+      "step": 20517
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.785073679960782e-06,
+      "loss": 0.4838,
+      "step": 20518
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.784496928800071e-06,
+      "loss": 0.4692,
+      "step": 20519
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.783920177639357e-06,
+      "loss": 0.408,
+      "step": 20520
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.783343426478646e-06,
+      "loss": 0.5271,
+      "step": 20521
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.782766675317936e-06,
+      "loss": 0.4405,
+      "step": 20522
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.782189924157224e-06,
+      "loss": 0.4605,
+      "step": 20523
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.781613172996511e-06,
+      "loss": 0.5181,
+      "step": 20524
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.7810364218358e-06,
+      "loss": 0.5256,
+      "step": 20525
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.780459670675088e-06,
+      "loss": 0.4751,
+      "step": 20526
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.779882919514376e-06,
+      "loss": 0.4283,
+      "step": 20527
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.779306168353664e-06,
+      "loss": 0.3579,
+      "step": 20528
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.778729417192953e-06,
+      "loss": 0.4248,
+      "step": 20529
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.77815266603224e-06,
+      "loss": 0.6532,
+      "step": 20530
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.777575914871528e-06,
+      "loss": 0.4732,
+      "step": 20531
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.776999163710818e-06,
+      "loss": 0.4377,
+      "step": 20532
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.776422412550105e-06,
+      "loss": 0.4851,
+      "step": 20533
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.775845661389393e-06,
+      "loss": 0.4223,
+      "step": 20534
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.775268910228683e-06,
+      "loss": 0.5088,
+      "step": 20535
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.774692159067972e-06,
+      "loss": 0.3979,
+      "step": 20536
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.77411540790726e-06,
+      "loss": 0.3931,
+      "step": 20537
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.773538656746547e-06,
+      "loss": 0.3502,
+      "step": 20538
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.772961905585837e-06,
+      "loss": 0.3329,
+      "step": 20539
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.772385154425124e-06,
+      "loss": 0.4058,
+      "step": 20540
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.771808403264412e-06,
+      "loss": 0.4095,
+      "step": 20541
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.771231652103701e-06,
+      "loss": 0.3738,
+      "step": 20542
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.770654900942989e-06,
+      "loss": 0.411,
+      "step": 20543
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.770078149782277e-06,
+      "loss": 0.3789,
+      "step": 20544
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.769501398621564e-06,
+      "loss": 0.3944,
+      "step": 20545
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.768924647460854e-06,
+      "loss": 0.4095,
+      "step": 20546
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.768347896300142e-06,
+      "loss": 0.5006,
+      "step": 20547
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.76777114513943e-06,
+      "loss": 0.4066,
+      "step": 20548
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.767194393978719e-06,
+      "loss": 0.4723,
+      "step": 20549
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.766617642818008e-06,
+      "loss": 0.3615,
+      "step": 20550
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.766040891657294e-06,
+      "loss": 0.4434,
+      "step": 20551
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.765464140496583e-06,
+      "loss": 0.4345,
+      "step": 20552
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.764887389335873e-06,
+      "loss": 0.3946,
+      "step": 20553
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.76431063817516e-06,
+      "loss": 0.4367,
+      "step": 20554
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.763733887014448e-06,
+      "loss": 0.4664,
+      "step": 20555
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.763157135853737e-06,
+      "loss": 0.3872,
+      "step": 20556
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.762580384693025e-06,
+      "loss": 0.437,
+      "step": 20557
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.762003633532313e-06,
+      "loss": 0.4474,
+      "step": 20558
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.761426882371602e-06,
+      "loss": 0.4337,
+      "step": 20559
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.76085013121089e-06,
+      "loss": 0.4481,
+      "step": 20560
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.760273380050178e-06,
+      "loss": 0.5559,
+      "step": 20561
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.759696628889465e-06,
+      "loss": 0.4364,
+      "step": 20562
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.759119877728755e-06,
+      "loss": 0.3861,
+      "step": 20563
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.758543126568042e-06,
+      "loss": 0.4413,
+      "step": 20564
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.75796637540733e-06,
+      "loss": 0.4515,
+      "step": 20565
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.75738962424662e-06,
+      "loss": 0.429,
+      "step": 20566
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.756812873085909e-06,
+      "loss": 0.3213,
+      "step": 20567
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.756236121925196e-06,
+      "loss": 0.4909,
+      "step": 20568
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.755659370764484e-06,
+      "loss": 0.4474,
+      "step": 20569
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.755082619603774e-06,
+      "loss": 0.3203,
+      "step": 20570
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.754505868443061e-06,
+      "loss": 0.4439,
+      "step": 20571
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.753929117282349e-06,
+      "loss": 0.3345,
+      "step": 20572
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.753352366121638e-06,
+      "loss": 0.3443,
+      "step": 20573
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.752775614960926e-06,
+      "loss": 0.4016,
+      "step": 20574
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.752198863800214e-06,
+      "loss": 0.4522,
+      "step": 20575
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.751622112639503e-06,
+      "loss": 0.5682,
+      "step": 20576
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.75104536147879e-06,
+      "loss": 0.4303,
+      "step": 20577
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.750468610318078e-06,
+      "loss": 0.4609,
+      "step": 20578
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.749891859157366e-06,
+      "loss": 0.4945,
+      "step": 20579
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.749315107996655e-06,
+      "loss": 0.4383,
+      "step": 20580
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.748738356835945e-06,
+      "loss": 0.3045,
+      "step": 20581
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.748161605675231e-06,
+      "loss": 0.4961,
+      "step": 20582
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74758485451452e-06,
+      "loss": 0.4375,
+      "step": 20583
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74700810335381e-06,
+      "loss": 0.4666,
+      "step": 20584
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.746431352193097e-06,
+      "loss": 0.4092,
+      "step": 20585
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.745854601032385e-06,
+      "loss": 0.3528,
+      "step": 20586
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.745277849871674e-06,
+      "loss": 0.4833,
+      "step": 20587
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.744701098710962e-06,
+      "loss": 0.4063,
+      "step": 20588
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74412434755025e-06,
+      "loss": 0.3875,
+      "step": 20589
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.743547596389539e-06,
+      "loss": 0.5339,
+      "step": 20590
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.742970845228827e-06,
+      "loss": 0.4497,
+      "step": 20591
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.742394094068114e-06,
+      "loss": 0.3853,
+      "step": 20592
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.741817342907404e-06,
+      "loss": 0.436,
+      "step": 20593
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.741240591746692e-06,
+      "loss": 0.3826,
+      "step": 20594
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.74066384058598e-06,
+      "loss": 0.4256,
+      "step": 20595
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.740087089425269e-06,
+      "loss": 0.4349,
+      "step": 20596
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.739510338264556e-06,
+      "loss": 0.4456,
+      "step": 20597
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.738933587103846e-06,
+      "loss": 0.321,
+      "step": 20598
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.738356835943133e-06,
+      "loss": 0.447,
+      "step": 20599
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.737780084782421e-06,
+      "loss": 0.4062,
+      "step": 20600
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.73720333362171e-06,
+      "loss": 0.4504,
+      "step": 20601
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.736626582460998e-06,
+      "loss": 0.4111,
+      "step": 20602
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.736049831300286e-06,
+      "loss": 0.5177,
+      "step": 20603
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.735473080139575e-06,
+      "loss": 0.4251,
+      "step": 20604
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.734896328978863e-06,
+      "loss": 0.3478,
+      "step": 20605
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.73431957781815e-06,
+      "loss": 0.5029,
+      "step": 20606
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.73374282665744e-06,
+      "loss": 0.345,
+      "step": 20607
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.733166075496728e-06,
+      "loss": 0.4359,
+      "step": 20608
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.732589324336015e-06,
+      "loss": 0.3931,
+      "step": 20609
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.732012573175305e-06,
+      "loss": 0.4583,
+      "step": 20610
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.731435822014592e-06,
+      "loss": 0.5125,
+      "step": 20611
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.730859070853882e-06,
+      "loss": 0.5548,
+      "step": 20612
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.73028231969317e-06,
+      "loss": 0.4003,
+      "step": 20613
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.729705568532457e-06,
+      "loss": 0.4614,
+      "step": 20614
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.729128817371746e-06,
+      "loss": 0.4116,
+      "step": 20615
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.728552066211034e-06,
+      "loss": 0.4018,
+      "step": 20616
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.727975315050322e-06,
+      "loss": 0.4803,
+      "step": 20617
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.727398563889611e-06,
+      "loss": 0.4997,
+      "step": 20618
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.726821812728899e-06,
+      "loss": 0.4678,
+      "step": 20619
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.726245061568187e-06,
+      "loss": 0.3944,
+      "step": 20620
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.725668310407476e-06,
+      "loss": 0.4948,
+      "step": 20621
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.725091559246764e-06,
+      "loss": 0.427,
+      "step": 20622
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.724514808086051e-06,
+      "loss": 0.4142,
+      "step": 20623
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.72393805692534e-06,
+      "loss": 0.3836,
+      "step": 20624
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.723361305764628e-06,
+      "loss": 0.5175,
+      "step": 20625
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.722784554603916e-06,
+      "loss": 0.4326,
+      "step": 20626
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.722207803443205e-06,
+      "loss": 0.55,
+      "step": 20627
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.721631052282493e-06,
+      "loss": 0.3719,
+      "step": 20628
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.721054301121783e-06,
+      "loss": 0.3484,
+      "step": 20629
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.72047754996107e-06,
+      "loss": 0.4563,
+      "step": 20630
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.719900798800358e-06,
+      "loss": 0.4126,
+      "step": 20631
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.719324047639647e-06,
+      "loss": 0.4494,
+      "step": 20632
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.718747296478935e-06,
+      "loss": 0.5086,
+      "step": 20633
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.718170545318223e-06,
+      "loss": 0.4154,
+      "step": 20634
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.717593794157512e-06,
+      "loss": 0.4024,
+      "step": 20635
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.7170170429968e-06,
+      "loss": 0.3208,
+      "step": 20636
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.716440291836087e-06,
+      "loss": 0.442,
+      "step": 20637
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.715863540675377e-06,
+      "loss": 0.4985,
+      "step": 20638
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.715286789514664e-06,
+      "loss": 0.3604,
+      "step": 20639
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714710038353952e-06,
+      "loss": 0.4545,
+      "step": 20640
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.714133287193242e-06,
+      "loss": 0.3981,
+      "step": 20641
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.71355653603253e-06,
+      "loss": 0.3623,
+      "step": 20642
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712979784871819e-06,
+      "loss": 0.4442,
+      "step": 20643
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.712403033711106e-06,
+      "loss": 0.4573,
+      "step": 20644
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 8.711826282550394e-06,
+      "loss": 0.3248,
+      "step": 20645
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.711249531389683e-06,
+      "loss": 0.3417,
+      "step": 20646
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.710672780228971e-06,
+      "loss": 0.4328,
+      "step": 20647
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.710096029068259e-06,
+      "loss": 0.445,
+      "step": 20648
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.709519277907548e-06,
+      "loss": 0.4158,
+      "step": 20649
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.708942526746836e-06,
+      "loss": 0.6331,
+      "step": 20650
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.708365775586123e-06,
+      "loss": 0.4864,
+      "step": 20651
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.707789024425413e-06,
+      "loss": 0.5473,
+      "step": 20652
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.7072122732647e-06,
+      "loss": 0.3857,
+      "step": 20653
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.706635522103988e-06,
+      "loss": 0.4678,
+      "step": 20654
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.706058770943278e-06,
+      "loss": 0.4137,
+      "step": 20655
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.705482019782565e-06,
+      "loss": 0.4065,
+      "step": 20656
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.704905268621853e-06,
+      "loss": 0.4186,
+      "step": 20657
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.704328517461142e-06,
+      "loss": 0.3734,
+      "step": 20658
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.70375176630043e-06,
+      "loss": 0.4988,
+      "step": 20659
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.70317501513972e-06,
+      "loss": 0.3508,
+      "step": 20660
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.702598263979007e-06,
+      "loss": 0.4848,
+      "step": 20661
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.702021512818295e-06,
+      "loss": 0.5414,
+      "step": 20662
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.701444761657584e-06,
+      "loss": 0.5581,
+      "step": 20663
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.700868010496872e-06,
+      "loss": 0.5061,
+      "step": 20664
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.70029125933616e-06,
+      "loss": 0.5155,
+      "step": 20665
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.699714508175449e-06,
+      "loss": 0.4953,
+      "step": 20666
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.699137757014737e-06,
+      "loss": 0.5218,
+      "step": 20667
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.698561005854024e-06,
+      "loss": 0.3687,
+      "step": 20668
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.697984254693314e-06,
+      "loss": 0.4297,
+      "step": 20669
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.697407503532601e-06,
+      "loss": 0.3903,
+      "step": 20670
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.696830752371889e-06,
+      "loss": 0.3731,
+      "step": 20671
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.696254001211178e-06,
+      "loss": 0.3741,
+      "step": 20672
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.695677250050466e-06,
+      "loss": 0.4593,
+      "step": 20673
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.695100498889755e-06,
+      "loss": 0.5474,
+      "step": 20674
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.694523747729043e-06,
+      "loss": 0.4351,
+      "step": 20675
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.69394699656833e-06,
+      "loss": 0.4679,
+      "step": 20676
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.69337024540762e-06,
+      "loss": 0.4845,
+      "step": 20677
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.692793494246908e-06,
+      "loss": 0.3703,
+      "step": 20678
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.692216743086196e-06,
+      "loss": 0.4784,
+      "step": 20679
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.691639991925485e-06,
+      "loss": 0.3043,
+      "step": 20680
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.691063240764773e-06,
+      "loss": 0.37,
+      "step": 20681
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.69048648960406e-06,
+      "loss": 0.4311,
+      "step": 20682
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.68990973844335e-06,
+      "loss": 0.5635,
+      "step": 20683
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.689332987282637e-06,
+      "loss": 0.4716,
+      "step": 20684
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.688756236121925e-06,
+      "loss": 0.2964,
+      "step": 20685
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.688179484961214e-06,
+      "loss": 0.4544,
+      "step": 20686
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.687602733800502e-06,
+      "loss": 0.3751,
+      "step": 20687
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.68702598263979e-06,
+      "loss": 0.4266,
+      "step": 20688
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.68644923147908e-06,
+      "loss": 0.4144,
+      "step": 20689
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.685872480318367e-06,
+      "loss": 0.2951,
+      "step": 20690
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.685295729157656e-06,
+      "loss": 0.5209,
+      "step": 20691
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.684718977996944e-06,
+      "loss": 0.3515,
+      "step": 20692
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.684142226836232e-06,
+      "loss": 0.454,
+      "step": 20693
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.683565475675521e-06,
+      "loss": 0.3736,
+      "step": 20694
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.682988724514809e-06,
+      "loss": 0.3826,
+      "step": 20695
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.682411973354096e-06,
+      "loss": 0.4246,
+      "step": 20696
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.681835222193386e-06,
+      "loss": 0.3961,
+      "step": 20697
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.681258471032673e-06,
+      "loss": 0.5179,
+      "step": 20698
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.680681719871961e-06,
+      "loss": 0.4607,
+      "step": 20699
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.68010496871125e-06,
+      "loss": 0.411,
+      "step": 20700
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.679528217550538e-06,
+      "loss": 0.5134,
+      "step": 20701
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.678951466389826e-06,
+      "loss": 0.4257,
+      "step": 20702
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.678374715229115e-06,
+      "loss": 0.4037,
+      "step": 20703
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.677797964068403e-06,
+      "loss": 0.45,
+      "step": 20704
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.677221212907692e-06,
+      "loss": 0.476,
+      "step": 20705
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.67664446174698e-06,
+      "loss": 0.4814,
+      "step": 20706
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.676067710586268e-06,
+      "loss": 0.5168,
+      "step": 20707
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.675490959425557e-06,
+      "loss": 0.3939,
+      "step": 20708
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.674914208264845e-06,
+      "loss": 0.3827,
+      "step": 20709
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.674337457104132e-06,
+      "loss": 0.4971,
+      "step": 20710
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.673760705943422e-06,
+      "loss": 0.3691,
+      "step": 20711
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.67318395478271e-06,
+      "loss": 0.499,
+      "step": 20712
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.672607203621997e-06,
+      "loss": 0.3795,
+      "step": 20713
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.672030452461287e-06,
+      "loss": 0.4465,
+      "step": 20714
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.671453701300574e-06,
+      "loss": 0.4323,
+      "step": 20715
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.670876950139862e-06,
+      "loss": 0.411,
+      "step": 20716
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.670300198979151e-06,
+      "loss": 0.3675,
+      "step": 20717
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66972344781844e-06,
+      "loss": 0.4746,
+      "step": 20718
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.669146696657728e-06,
+      "loss": 0.3784,
+      "step": 20719
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.668569945497016e-06,
+      "loss": 0.5277,
+      "step": 20720
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667993194336304e-06,
+      "loss": 0.4712,
+      "step": 20721
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.667416443175593e-06,
+      "loss": 0.459,
+      "step": 20722
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66683969201488e-06,
+      "loss": 0.3644,
+      "step": 20723
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.666262940854169e-06,
+      "loss": 0.539,
+      "step": 20724
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665686189693458e-06,
+      "loss": 0.3913,
+      "step": 20725
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.665109438532746e-06,
+      "loss": 0.4628,
+      "step": 20726
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.664532687372033e-06,
+      "loss": 0.4177,
+      "step": 20727
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.663955936211323e-06,
+      "loss": 0.3727,
+      "step": 20728
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.66337918505061e-06,
+      "loss": 0.4099,
+      "step": 20729
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662802433889898e-06,
+      "loss": 0.3909,
+      "step": 20730
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.662225682729187e-06,
+      "loss": 0.3735,
+      "step": 20731
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.661648931568477e-06,
+      "loss": 0.5599,
+      "step": 20732
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.661072180407763e-06,
+      "loss": 0.3985,
+      "step": 20733
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.660495429247052e-06,
+      "loss": 0.41,
+      "step": 20734
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.659918678086342e-06,
+      "loss": 0.5821,
+      "step": 20735
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65934192692563e-06,
+      "loss": 0.4008,
+      "step": 20736
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.658765175764917e-06,
+      "loss": 0.4678,
+      "step": 20737
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.658188424604205e-06,
+      "loss": 0.4216,
+      "step": 20738
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.657611673443494e-06,
+      "loss": 0.4553,
+      "step": 20739
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.657034922282782e-06,
+      "loss": 0.5656,
+      "step": 20740
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.65645817112207e-06,
+      "loss": 0.549,
+      "step": 20741
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.655881419961359e-06,
+      "loss": 0.4384,
+      "step": 20742
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.655304668800646e-06,
+      "loss": 0.3977,
+      "step": 20743
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654727917639934e-06,
+      "loss": 0.4336,
+      "step": 20744
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.654151166479223e-06,
+      "loss": 0.394,
+      "step": 20745
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.653574415318511e-06,
+      "loss": 0.3623,
+      "step": 20746
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652997664157799e-06,
+      "loss": 0.3109,
+      "step": 20747
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.652420912997088e-06,
+      "loss": 0.4809,
+      "step": 20748
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.651844161836378e-06,
+      "loss": 0.4516,
+      "step": 20749
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.651267410675665e-06,
+      "loss": 0.3683,
+      "step": 20750
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650690659514953e-06,
+      "loss": 0.4738,
+      "step": 20751
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.650113908354242e-06,
+      "loss": 0.5006,
+      "step": 20752
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64953715719353e-06,
+      "loss": 0.5168,
+      "step": 20753
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648960406032818e-06,
+      "loss": 0.5258,
+      "step": 20754
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.648383654872105e-06,
+      "loss": 0.3735,
+      "step": 20755
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.647806903711395e-06,
+      "loss": 0.4246,
+      "step": 20756
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.647230152550682e-06,
+      "loss": 0.4583,
+      "step": 20757
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64665340138997e-06,
+      "loss": 0.3822,
+      "step": 20758
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.64607665022926e-06,
+      "loss": 0.4679,
+      "step": 20759
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.645499899068547e-06,
+      "loss": 0.4568,
+      "step": 20760
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.644923147907835e-06,
+      "loss": 0.4709,
+      "step": 20761
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.644346396747124e-06,
+      "loss": 0.3102,
+      "step": 20762
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.643769645586414e-06,
+      "loss": 0.4031,
+      "step": 20763
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.6431928944257e-06,
+      "loss": 0.576,
+      "step": 20764
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.642616143264989e-06,
+      "loss": 0.4858,
+      "step": 20765
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.642039392104278e-06,
+      "loss": 0.4201,
+      "step": 20766
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.641462640943566e-06,
+      "loss": 0.5133,
+      "step": 20767
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.640885889782854e-06,
+      "loss": 0.47,
+      "step": 20768
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.640309138622143e-06,
+      "loss": 0.4029,
+      "step": 20769
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63973238746143e-06,
+      "loss": 0.4377,
+      "step": 20770
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.639155636300719e-06,
+      "loss": 0.4795,
+      "step": 20771
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.638578885140006e-06,
+      "loss": 0.3195,
+      "step": 20772
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.638002133979296e-06,
+      "loss": 0.3178,
+      "step": 20773
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.637425382818583e-06,
+      "loss": 0.3812,
+      "step": 20774
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.636848631657871e-06,
+      "loss": 0.521,
+      "step": 20775
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63627188049716e-06,
+      "loss": 0.3988,
+      "step": 20776
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635695129336448e-06,
+      "loss": 0.3676,
+      "step": 20777
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.635118378175736e-06,
+      "loss": 0.4371,
+      "step": 20778
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.634541627015025e-06,
+      "loss": 0.4298,
+      "step": 20779
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.633964875854314e-06,
+      "loss": 0.4609,
+      "step": 20780
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.633388124693602e-06,
+      "loss": 0.387,
+      "step": 20781
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63281137353289e-06,
+      "loss": 0.4937,
+      "step": 20782
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.63223462237218e-06,
+      "loss": 0.446,
+      "step": 20783
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.631657871211467e-06,
+      "loss": 0.4514,
+      "step": 20784
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.631081120050755e-06,
+      "loss": 0.3889,
+      "step": 20785
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.630504368890044e-06,
+      "loss": 0.5491,
+      "step": 20786
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.629927617729332e-06,
+      "loss": 0.4258,
+      "step": 20787
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62935086656862e-06,
+      "loss": 0.4843,
+      "step": 20788
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628774115407907e-06,
+      "loss": 0.4716,
+      "step": 20789
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.628197364247196e-06,
+      "loss": 0.4264,
+      "step": 20790
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627620613086484e-06,
+      "loss": 0.3839,
+      "step": 20791
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.627043861925772e-06,
+      "loss": 0.5138,
+      "step": 20792
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.626467110765061e-06,
+      "loss": 0.5032,
+      "step": 20793
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62589035960435e-06,
+      "loss": 0.6154,
+      "step": 20794
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.625313608443637e-06,
+      "loss": 0.4493,
+      "step": 20795
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.624736857282926e-06,
+      "loss": 0.5005,
+      "step": 20796
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.624160106122215e-06,
+      "loss": 0.46,
+      "step": 20797
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.623583354961503e-06,
+      "loss": 0.3858,
+      "step": 20798
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62300660380079e-06,
+      "loss": 0.3784,
+      "step": 20799
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.62242985264008e-06,
+      "loss": 0.4382,
+      "step": 20800
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621853101479368e-06,
+      "loss": 0.4527,
+      "step": 20801
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.621276350318655e-06,
+      "loss": 0.3185,
+      "step": 20802
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.620699599157945e-06,
+      "loss": 0.4198,
+      "step": 20803
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.620122847997232e-06,
+      "loss": 0.4087,
+      "step": 20804
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.61954609683652e-06,
+      "loss": 0.3364,
+      "step": 20805
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.618969345675808e-06,
+      "loss": 0.6125,
+      "step": 20806
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.618392594515097e-06,
+      "loss": 0.4661,
+      "step": 20807
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.617815843354385e-06,
+      "loss": 0.361,
+      "step": 20808
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.617239092193673e-06,
+      "loss": 0.485,
+      "step": 20809
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.616662341032962e-06,
+      "loss": 0.3513,
+      "step": 20810
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.616085589872251e-06,
+      "loss": 0.4711,
+      "step": 20811
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.615508838711539e-06,
+      "loss": 0.4981,
+      "step": 20812
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.614932087550827e-06,
+      "loss": 0.3553,
+      "step": 20813
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.614355336390116e-06,
+      "loss": 0.3807,
+      "step": 20814
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.613778585229404e-06,
+      "loss": 0.3969,
+      "step": 20815
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.613201834068691e-06,
+      "loss": 0.5121,
+      "step": 20816
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.61262508290798e-06,
+      "loss": 0.4381,
+      "step": 20817
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.612048331747269e-06,
+      "loss": 0.3728,
+      "step": 20818
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.611471580586556e-06,
+      "loss": 0.3953,
+      "step": 20819
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.610894829425846e-06,
+      "loss": 0.4946,
+      "step": 20820
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.610318078265133e-06,
+      "loss": 0.4218,
+      "step": 20821
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.609741327104421e-06,
+      "loss": 0.3631,
+      "step": 20822
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.609164575943709e-06,
+      "loss": 0.3975,
+      "step": 20823
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 8.608587824782998e-06,
+      "loss": 0.41,
+      "step": 20824
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.608011073622287e-06,
+      "loss": 0.3737,
+      "step": 20825
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.607434322461573e-06,
+      "loss": 0.4313,
+      "step": 20826
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.606857571300863e-06,
+      "loss": 0.2818,
+      "step": 20827
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.606280820140152e-06,
+      "loss": 0.5867,
+      "step": 20828
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60570406897944e-06,
+      "loss": 0.294,
+      "step": 20829
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.605127317818728e-06,
+      "loss": 0.379,
+      "step": 20830
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.604550566658017e-06,
+      "loss": 0.3759,
+      "step": 20831
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603973815497305e-06,
+      "loss": 0.4284,
+      "step": 20832
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.603397064336592e-06,
+      "loss": 0.512,
+      "step": 20833
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.602820313175882e-06,
+      "loss": 0.4854,
+      "step": 20834
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.60224356201517e-06,
+      "loss": 0.4502,
+      "step": 20835
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601666810854457e-06,
+      "loss": 0.3822,
+      "step": 20836
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.601090059693746e-06,
+      "loss": 0.5884,
+      "step": 20837
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.600513308533034e-06,
+      "loss": 0.4358,
+      "step": 20838
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.599936557372322e-06,
+      "loss": 0.3782,
+      "step": 20839
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59935980621161e-06,
+      "loss": 0.4942,
+      "step": 20840
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.598783055050899e-06,
+      "loss": 0.5377,
+      "step": 20841
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.598206303890188e-06,
+      "loss": 0.4608,
+      "step": 20842
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.597629552729476e-06,
+      "loss": 0.4183,
+      "step": 20843
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.597052801568764e-06,
+      "loss": 0.5056,
+      "step": 20844
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.596476050408053e-06,
+      "loss": 0.5141,
+      "step": 20845
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59589929924734e-06,
+      "loss": 0.4397,
+      "step": 20846
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.595322548086628e-06,
+      "loss": 0.5016,
+      "step": 20847
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594745796925918e-06,
+      "loss": 0.5159,
+      "step": 20848
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.594169045765205e-06,
+      "loss": 0.337,
+      "step": 20849
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.593592294604493e-06,
+      "loss": 0.5136,
+      "step": 20850
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.593015543443782e-06,
+      "loss": 0.4373,
+      "step": 20851
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.59243879228307e-06,
+      "loss": 0.3466,
+      "step": 20852
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.591862041122358e-06,
+      "loss": 0.3869,
+      "step": 20853
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.591285289961647e-06,
+      "loss": 0.5182,
+      "step": 20854
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.590708538800935e-06,
+      "loss": 0.5332,
+      "step": 20855
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.590131787640224e-06,
+      "loss": 0.3325,
+      "step": 20856
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.589555036479512e-06,
+      "loss": 0.4246,
+      "step": 20857
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5889782853188e-06,
+      "loss": 0.4492,
+      "step": 20858
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.588401534158089e-06,
+      "loss": 0.5284,
+      "step": 20859
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.587824782997377e-06,
+      "loss": 0.3646,
+      "step": 20860
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.587248031836664e-06,
+      "loss": 0.4626,
+      "step": 20861
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.586671280675954e-06,
+      "loss": 0.4184,
+      "step": 20862
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.586094529515241e-06,
+      "loss": 0.5024,
+      "step": 20863
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.585517778354529e-06,
+      "loss": 0.3809,
+      "step": 20864
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.584941027193819e-06,
+      "loss": 0.3914,
+      "step": 20865
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.584364276033106e-06,
+      "loss": 0.4386,
+      "step": 20866
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.583787524872394e-06,
+      "loss": 0.4641,
+      "step": 20867
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.583210773711683e-06,
+      "loss": 0.4905,
+      "step": 20868
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.582634022550971e-06,
+      "loss": 0.4375,
+      "step": 20869
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.582057271390259e-06,
+      "loss": 0.4351,
+      "step": 20870
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.581480520229548e-06,
+      "loss": 0.3811,
+      "step": 20871
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580903769068836e-06,
+      "loss": 0.4443,
+      "step": 20872
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.580327017908125e-06,
+      "loss": 0.5379,
+      "step": 20873
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.579750266747413e-06,
+      "loss": 0.4863,
+      "step": 20874
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5791735155867e-06,
+      "loss": 0.3947,
+      "step": 20875
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.57859676442599e-06,
+      "loss": 0.4315,
+      "step": 20876
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.578020013265278e-06,
+      "loss": 0.5113,
+      "step": 20877
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.577443262104565e-06,
+      "loss": 0.338,
+      "step": 20878
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.576866510943855e-06,
+      "loss": 0.4773,
+      "step": 20879
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.576289759783142e-06,
+      "loss": 0.4154,
+      "step": 20880
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.57571300862243e-06,
+      "loss": 0.4399,
+      "step": 20881
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.57513625746172e-06,
+      "loss": 0.4826,
+      "step": 20882
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.574559506301007e-06,
+      "loss": 0.3912,
+      "step": 20883
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.573982755140295e-06,
+      "loss": 0.5056,
+      "step": 20884
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.573406003979584e-06,
+      "loss": 0.5058,
+      "step": 20885
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.572829252818872e-06,
+      "loss": 0.2959,
+      "step": 20886
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.572252501658161e-06,
+      "loss": 0.318,
+      "step": 20887
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.571675750497449e-06,
+      "loss": 0.5117,
+      "step": 20888
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.571098999336737e-06,
+      "loss": 0.4741,
+      "step": 20889
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.570522248176026e-06,
+      "loss": 0.4816,
+      "step": 20890
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.569945497015314e-06,
+      "loss": 0.4329,
+      "step": 20891
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.569368745854601e-06,
+      "loss": 0.3651,
+      "step": 20892
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.56879199469389e-06,
+      "loss": 0.511,
+      "step": 20893
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.568215243533178e-06,
+      "loss": 0.4476,
+      "step": 20894
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.567638492372466e-06,
+      "loss": 0.4941,
+      "step": 20895
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.567061741211755e-06,
+      "loss": 0.5056,
+      "step": 20896
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.566484990051043e-06,
+      "loss": 0.4,
+      "step": 20897
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.56590823889033e-06,
+      "loss": 0.3964,
+      "step": 20898
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.56533148772962e-06,
+      "loss": 0.4488,
+      "step": 20899
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.564754736568908e-06,
+      "loss": 0.3754,
+      "step": 20900
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.564177985408196e-06,
+      "loss": 0.3944,
+      "step": 20901
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.563601234247485e-06,
+      "loss": 0.4427,
+      "step": 20902
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.563024483086773e-06,
+      "loss": 0.3542,
+      "step": 20903
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.562447731926062e-06,
+      "loss": 0.4268,
+      "step": 20904
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.56187098076535e-06,
+      "loss": 0.3902,
+      "step": 20905
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.561294229604637e-06,
+      "loss": 0.5067,
+      "step": 20906
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.560717478443927e-06,
+      "loss": 0.4864,
+      "step": 20907
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.560140727283214e-06,
+      "loss": 0.4554,
+      "step": 20908
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.559563976122502e-06,
+      "loss": 0.5255,
+      "step": 20909
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.558987224961791e-06,
+      "loss": 0.524,
+      "step": 20910
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.558410473801079e-06,
+      "loss": 0.5081,
+      "step": 20911
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.557833722640367e-06,
+      "loss": 0.4967,
+      "step": 20912
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.557256971479656e-06,
+      "loss": 0.4364,
+      "step": 20913
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.556680220318944e-06,
+      "loss": 0.3872,
+      "step": 20914
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.556103469158232e-06,
+      "loss": 0.5157,
+      "step": 20915
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.555526717997521e-06,
+      "loss": 0.4509,
+      "step": 20916
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.554949966836809e-06,
+      "loss": 0.4424,
+      "step": 20917
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.554373215676098e-06,
+      "loss": 0.3843,
+      "step": 20918
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.553796464515386e-06,
+      "loss": 0.4612,
+      "step": 20919
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.553219713354673e-06,
+      "loss": 0.4975,
+      "step": 20920
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.552642962193963e-06,
+      "loss": 0.4282,
+      "step": 20921
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.55206621103325e-06,
+      "loss": 0.3773,
+      "step": 20922
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.551489459872538e-06,
+      "loss": 0.4383,
+      "step": 20923
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.550912708711828e-06,
+      "loss": 0.3999,
+      "step": 20924
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.550335957551115e-06,
+      "loss": 0.4957,
+      "step": 20925
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.549759206390403e-06,
+      "loss": 0.3757,
+      "step": 20926
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.549182455229692e-06,
+      "loss": 0.3965,
+      "step": 20927
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.54860570406898e-06,
+      "loss": 0.4109,
+      "step": 20928
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.548028952908268e-06,
+      "loss": 0.4545,
+      "step": 20929
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.547452201747557e-06,
+      "loss": 0.3892,
+      "step": 20930
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.546875450586845e-06,
+      "loss": 0.3704,
+      "step": 20931
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.546298699426132e-06,
+      "loss": 0.5485,
+      "step": 20932
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.545721948265422e-06,
+      "loss": 0.4564,
+      "step": 20933
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.54514519710471e-06,
+      "loss": 0.4647,
+      "step": 20934
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.544568445943999e-06,
+      "loss": 0.4269,
+      "step": 20935
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.543991694783287e-06,
+      "loss": 0.4793,
+      "step": 20936
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.543414943622574e-06,
+      "loss": 0.5038,
+      "step": 20937
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.542838192461864e-06,
+      "loss": 0.4554,
+      "step": 20938
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.542261441301151e-06,
+      "loss": 0.5384,
+      "step": 20939
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.541684690140439e-06,
+      "loss": 0.397,
+      "step": 20940
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.541107938979728e-06,
+      "loss": 0.3431,
+      "step": 20941
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.540531187819016e-06,
+      "loss": 0.4529,
+      "step": 20942
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539954436658304e-06,
+      "loss": 0.3907,
+      "step": 20943
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.539377685497593e-06,
+      "loss": 0.4274,
+      "step": 20944
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.53880093433688e-06,
+      "loss": 0.3996,
+      "step": 20945
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.538224183176168e-06,
+      "loss": 0.417,
+      "step": 20946
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.537647432015458e-06,
+      "loss": 0.6034,
+      "step": 20947
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.537070680854746e-06,
+      "loss": 0.4594,
+      "step": 20948
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.536493929694035e-06,
+      "loss": 0.466,
+      "step": 20949
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.535917178533323e-06,
+      "loss": 0.4256,
+      "step": 20950
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.53534042737261e-06,
+      "loss": 0.3777,
+      "step": 20951
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5347636762119e-06,
+      "loss": 0.383,
+      "step": 20952
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.534186925051187e-06,
+      "loss": 0.4538,
+      "step": 20953
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.533610173890475e-06,
+      "loss": 0.3428,
+      "step": 20954
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.533033422729764e-06,
+      "loss": 0.3424,
+      "step": 20955
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.532456671569052e-06,
+      "loss": 0.5132,
+      "step": 20956
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.53187992040834e-06,
+      "loss": 0.4321,
+      "step": 20957
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.531303169247629e-06,
+      "loss": 0.3633,
+      "step": 20958
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.530726418086917e-06,
+      "loss": 0.4233,
+      "step": 20959
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.530149666926205e-06,
+      "loss": 0.5753,
+      "step": 20960
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.529572915765494e-06,
+      "loss": 0.4343,
+      "step": 20961
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.528996164604782e-06,
+      "loss": 0.423,
+      "step": 20962
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.52841941344407e-06,
+      "loss": 0.4935,
+      "step": 20963
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.527842662283359e-06,
+      "loss": 0.4312,
+      "step": 20964
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.527265911122646e-06,
+      "loss": 0.3789,
+      "step": 20965
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.526689159961936e-06,
+      "loss": 0.4265,
+      "step": 20966
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.526112408801223e-06,
+      "loss": 0.5108,
+      "step": 20967
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.525535657640511e-06,
+      "loss": 0.3981,
+      "step": 20968
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.5249589064798e-06,
+      "loss": 0.5014,
+      "step": 20969
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.524382155319088e-06,
+      "loss": 0.4128,
+      "step": 20970
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.523805404158376e-06,
+      "loss": 0.4151,
+      "step": 20971
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.523228652997665e-06,
+      "loss": 0.4737,
+      "step": 20972
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.522651901836953e-06,
+      "loss": 0.408,
+      "step": 20973
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.52207515067624e-06,
+      "loss": 0.5887,
+      "step": 20974
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.52149839951553e-06,
+      "loss": 0.4059,
+      "step": 20975
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.52092164835482e-06,
+      "loss": 0.4292,
+      "step": 20976
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.520344897194105e-06,
+      "loss": 0.4848,
+      "step": 20977
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.519768146033395e-06,
+      "loss": 0.393,
+      "step": 20978
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.519191394872684e-06,
+      "loss": 0.4283,
+      "step": 20979
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.518614643711972e-06,
+      "loss": 0.4856,
+      "step": 20980
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.51803789255126e-06,
+      "loss": 0.4055,
+      "step": 20981
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.517461141390547e-06,
+      "loss": 0.4346,
+      "step": 20982
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.516884390229837e-06,
+      "loss": 0.5076,
+      "step": 20983
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.516307639069124e-06,
+      "loss": 0.5385,
+      "step": 20984
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.515730887908412e-06,
+      "loss": 0.4881,
+      "step": 20985
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.515154136747701e-06,
+      "loss": 0.4402,
+      "step": 20986
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.514577385586989e-06,
+      "loss": 0.4661,
+      "step": 20987
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.514000634426277e-06,
+      "loss": 0.583,
+      "step": 20988
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.513423883265566e-06,
+      "loss": 0.4184,
+      "step": 20989
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.512847132104854e-06,
+      "loss": 0.3396,
+      "step": 20990
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.512270380944141e-06,
+      "loss": 0.5239,
+      "step": 20991
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.51169362978343e-06,
+      "loss": 0.3938,
+      "step": 20992
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.51111687862272e-06,
+      "loss": 0.4077,
+      "step": 20993
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.510540127462006e-06,
+      "loss": 0.4955,
+      "step": 20994
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.509963376301296e-06,
+      "loss": 0.4434,
+      "step": 20995
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.509386625140585e-06,
+      "loss": 0.459,
+      "step": 20996
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.508809873979873e-06,
+      "loss": 0.418,
+      "step": 20997
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.50823312281916e-06,
+      "loss": 0.5499,
+      "step": 20998
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.507656371658448e-06,
+      "loss": 0.3662,
+      "step": 20999
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.507079620497737e-06,
+      "loss": 0.5271,
+      "step": 21000
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.506502869337025e-06,
+      "loss": 0.4316,
+      "step": 21001
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.505926118176313e-06,
+      "loss": 0.5996,
+      "step": 21002
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 8.505349367015602e-06,
+      "loss": 0.4112,
+      "step": 21003
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.50477261585489e-06,
+      "loss": 0.4299,
+      "step": 21004
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.504195864694177e-06,
+      "loss": 0.4498,
+      "step": 21005
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.503619113533467e-06,
+      "loss": 0.512,
+      "step": 21006
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.503042362372756e-06,
+      "loss": 0.4979,
+      "step": 21007
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.502465611212042e-06,
+      "loss": 0.4512,
+      "step": 21008
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.501888860051332e-06,
+      "loss": 0.3765,
+      "step": 21009
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.501312108890621e-06,
+      "loss": 0.3413,
+      "step": 21010
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.500735357729909e-06,
+      "loss": 0.3839,
+      "step": 21011
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.500158606569196e-06,
+      "loss": 0.4496,
+      "step": 21012
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.499581855408486e-06,
+      "loss": 0.3695,
+      "step": 21013
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.499005104247773e-06,
+      "loss": 0.453,
+      "step": 21014
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.498428353087061e-06,
+      "loss": 0.4508,
+      "step": 21015
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497851601926349e-06,
+      "loss": 0.4517,
+      "step": 21016
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.497274850765638e-06,
+      "loss": 0.5134,
+      "step": 21017
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.496698099604926e-06,
+      "loss": 0.4346,
+      "step": 21018
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.496121348444214e-06,
+      "loss": 0.4807,
+      "step": 21019
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.495544597283503e-06,
+      "loss": 0.5202,
+      "step": 21020
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.49496784612279e-06,
+      "loss": 0.5873,
+      "step": 21021
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.494391094962078e-06,
+      "loss": 0.4124,
+      "step": 21022
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.493814343801368e-06,
+      "loss": 0.4441,
+      "step": 21023
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.493237592640657e-06,
+      "loss": 0.4214,
+      "step": 21024
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.492660841479943e-06,
+      "loss": 0.494,
+      "step": 21025
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.492084090319232e-06,
+      "loss": 0.4211,
+      "step": 21026
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.491507339158522e-06,
+      "loss": 0.3952,
+      "step": 21027
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.49093058799781e-06,
+      "loss": 0.3864,
+      "step": 21028
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.490353836837097e-06,
+      "loss": 0.4745,
+      "step": 21029
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.489777085676387e-06,
+      "loss": 0.5173,
+      "step": 21030
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.489200334515674e-06,
+      "loss": 0.3298,
+      "step": 21031
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.488623583354962e-06,
+      "loss": 0.3443,
+      "step": 21032
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.48804683219425e-06,
+      "loss": 0.4126,
+      "step": 21033
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.487470081033539e-06,
+      "loss": 0.3955,
+      "step": 21034
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.486893329872827e-06,
+      "loss": 0.452,
+      "step": 21035
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.486316578712114e-06,
+      "loss": 0.4966,
+      "step": 21036
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.485739827551404e-06,
+      "loss": 0.4293,
+      "step": 21037
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.485163076390693e-06,
+      "loss": 0.4256,
+      "step": 21038
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.484586325229979e-06,
+      "loss": 0.4002,
+      "step": 21039
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.484009574069268e-06,
+      "loss": 0.385,
+      "step": 21040
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.483432822908558e-06,
+      "loss": 0.3985,
+      "step": 21041
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.482856071747846e-06,
+      "loss": 0.4939,
+      "step": 21042
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.482279320587133e-06,
+      "loss": 0.2504,
+      "step": 21043
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.481702569426423e-06,
+      "loss": 0.424,
+      "step": 21044
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.48112581826571e-06,
+      "loss": 0.5253,
+      "step": 21045
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.480549067104998e-06,
+      "loss": 0.3908,
+      "step": 21046
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.479972315944287e-06,
+      "loss": 0.4056,
+      "step": 21047
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.479395564783575e-06,
+      "loss": 0.3684,
+      "step": 21048
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.478818813622863e-06,
+      "loss": 0.4359,
+      "step": 21049
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.47824206246215e-06,
+      "loss": 0.4139,
+      "step": 21050
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.47766531130144e-06,
+      "loss": 0.4622,
+      "step": 21051
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.477088560140727e-06,
+      "loss": 0.3645,
+      "step": 21052
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.476511808980015e-06,
+      "loss": 0.3903,
+      "step": 21053
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.475935057819305e-06,
+      "loss": 0.5124,
+      "step": 21054
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.475358306658594e-06,
+      "loss": 0.4777,
+      "step": 21055
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.47478155549788e-06,
+      "loss": 0.4955,
+      "step": 21056
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.47420480433717e-06,
+      "loss": 0.2768,
+      "step": 21057
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.473628053176459e-06,
+      "loss": 0.4482,
+      "step": 21058
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.473051302015746e-06,
+      "loss": 0.4166,
+      "step": 21059
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.472474550855034e-06,
+      "loss": 0.4645,
+      "step": 21060
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.471897799694323e-06,
+      "loss": 0.5012,
+      "step": 21061
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.471321048533611e-06,
+      "loss": 0.5037,
+      "step": 21062
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.470744297372899e-06,
+      "loss": 0.4632,
+      "step": 21063
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.470167546212188e-06,
+      "loss": 0.4459,
+      "step": 21064
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.469590795051476e-06,
+      "loss": 0.4985,
+      "step": 21065
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.469014043890764e-06,
+      "loss": 0.335,
+      "step": 21066
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.468437292730051e-06,
+      "loss": 0.3749,
+      "step": 21067
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.46786054156934e-06,
+      "loss": 0.4034,
+      "step": 21068
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.46728379040863e-06,
+      "loss": 0.3842,
+      "step": 21069
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.466707039247916e-06,
+      "loss": 0.4455,
+      "step": 21070
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.466130288087205e-06,
+      "loss": 0.3884,
+      "step": 21071
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.465553536926495e-06,
+      "loss": 0.6484,
+      "step": 21072
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.464976785765782e-06,
+      "loss": 0.4469,
+      "step": 21073
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.46440003460507e-06,
+      "loss": 0.485,
+      "step": 21074
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.46382328344436e-06,
+      "loss": 0.4744,
+      "step": 21075
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.463246532283647e-06,
+      "loss": 0.4316,
+      "step": 21076
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.462669781122935e-06,
+      "loss": 0.4717,
+      "step": 21077
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.462093029962224e-06,
+      "loss": 0.5058,
+      "step": 21078
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.461516278801512e-06,
+      "loss": 0.4425,
+      "step": 21079
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.4609395276408e-06,
+      "loss": 0.2583,
+      "step": 21080
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.460362776480089e-06,
+      "loss": 0.4655,
+      "step": 21081
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.459786025319377e-06,
+      "loss": 0.4726,
+      "step": 21082
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.459209274158664e-06,
+      "loss": 0.3471,
+      "step": 21083
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.458632522997952e-06,
+      "loss": 0.3863,
+      "step": 21084
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.458055771837241e-06,
+      "loss": 0.3008,
+      "step": 21085
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.45747902067653e-06,
+      "loss": 0.4269,
+      "step": 21086
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456902269515818e-06,
+      "loss": 0.4882,
+      "step": 21087
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.456325518355106e-06,
+      "loss": 0.4113,
+      "step": 21088
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.455748767194396e-06,
+      "loss": 0.4668,
+      "step": 21089
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.455172016033683e-06,
+      "loss": 0.3864,
+      "step": 21090
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.454595264872971e-06,
+      "loss": 0.4818,
+      "step": 21091
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.45401851371226e-06,
+      "loss": 0.5406,
+      "step": 21092
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.453441762551548e-06,
+      "loss": 0.4188,
+      "step": 21093
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.452865011390836e-06,
+      "loss": 0.481,
+      "step": 21094
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.452288260230125e-06,
+      "loss": 0.4862,
+      "step": 21095
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.451711509069413e-06,
+      "loss": 0.365,
+      "step": 21096
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.4511347579087e-06,
+      "loss": 0.4291,
+      "step": 21097
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.45055800674799e-06,
+      "loss": 0.4331,
+      "step": 21098
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.449981255587277e-06,
+      "loss": 0.442,
+      "step": 21099
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.449404504426567e-06,
+      "loss": 0.4364,
+      "step": 21100
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.448827753265855e-06,
+      "loss": 0.5331,
+      "step": 21101
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.448251002105142e-06,
+      "loss": 0.4638,
+      "step": 21102
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.447674250944432e-06,
+      "loss": 0.5126,
+      "step": 21103
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.44709749978372e-06,
+      "loss": 0.397,
+      "step": 21104
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.446520748623007e-06,
+      "loss": 0.381,
+      "step": 21105
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.445943997462296e-06,
+      "loss": 0.366,
+      "step": 21106
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.445367246301584e-06,
+      "loss": 0.392,
+      "step": 21107
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.444790495140872e-06,
+      "loss": 0.4826,
+      "step": 21108
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.444213743980161e-06,
+      "loss": 0.4434,
+      "step": 21109
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.443636992819449e-06,
+      "loss": 0.5002,
+      "step": 21110
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.443060241658736e-06,
+      "loss": 0.4159,
+      "step": 21111
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.442483490498026e-06,
+      "loss": 0.5335,
+      "step": 21112
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.441906739337314e-06,
+      "loss": 0.5343,
+      "step": 21113
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.441329988176601e-06,
+      "loss": 0.5864,
+      "step": 21114
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.44075323701589e-06,
+      "loss": 0.4909,
+      "step": 21115
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.440176485855178e-06,
+      "loss": 0.3774,
+      "step": 21116
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.439599734694468e-06,
+      "loss": 0.4555,
+      "step": 21117
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.439022983533755e-06,
+      "loss": 0.4675,
+      "step": 21118
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.438446232373043e-06,
+      "loss": 0.5276,
+      "step": 21119
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.437869481212332e-06,
+      "loss": 0.5033,
+      "step": 21120
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.43729273005162e-06,
+      "loss": 0.5083,
+      "step": 21121
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436715978890908e-06,
+      "loss": 0.3953,
+      "step": 21122
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.436139227730197e-06,
+      "loss": 0.3829,
+      "step": 21123
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.435562476569485e-06,
+      "loss": 0.4306,
+      "step": 21124
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.434985725408773e-06,
+      "loss": 0.4881,
+      "step": 21125
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.434408974248062e-06,
+      "loss": 0.4849,
+      "step": 21126
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.43383222308735e-06,
+      "loss": 0.4455,
+      "step": 21127
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.433255471926637e-06,
+      "loss": 0.4099,
+      "step": 21128
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.432678720765927e-06,
+      "loss": 0.4982,
+      "step": 21129
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.432101969605214e-06,
+      "loss": 0.4939,
+      "step": 21130
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.431525218444504e-06,
+      "loss": 0.314,
+      "step": 21131
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.430948467283791e-06,
+      "loss": 0.4732,
+      "step": 21132
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.430371716123079e-06,
+      "loss": 0.4062,
+      "step": 21133
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.429794964962368e-06,
+      "loss": 0.4947,
+      "step": 21134
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.429218213801656e-06,
+      "loss": 0.4618,
+      "step": 21135
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.428641462640944e-06,
+      "loss": 0.4251,
+      "step": 21136
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.428064711480233e-06,
+      "loss": 0.3983,
+      "step": 21137
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.427487960319521e-06,
+      "loss": 0.4241,
+      "step": 21138
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.426911209158809e-06,
+      "loss": 0.4248,
+      "step": 21139
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.426334457998098e-06,
+      "loss": 0.4839,
+      "step": 21140
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.425757706837386e-06,
+      "loss": 0.404,
+      "step": 21141
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.425180955676673e-06,
+      "loss": 0.478,
+      "step": 21142
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.424604204515963e-06,
+      "loss": 0.4994,
+      "step": 21143
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.42402745335525e-06,
+      "loss": 0.5009,
+      "step": 21144
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.423450702194538e-06,
+      "loss": 0.3235,
+      "step": 21145
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.422873951033827e-06,
+      "loss": 0.4466,
+      "step": 21146
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.422297199873115e-06,
+      "loss": 0.4934,
+      "step": 21147
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.421720448712405e-06,
+      "loss": 0.4417,
+      "step": 21148
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.421143697551692e-06,
+      "loss": 0.4932,
+      "step": 21149
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.42056694639098e-06,
+      "loss": 0.4496,
+      "step": 21150
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.41999019523027e-06,
+      "loss": 0.5056,
+      "step": 21151
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.419413444069557e-06,
+      "loss": 0.3006,
+      "step": 21152
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.418836692908845e-06,
+      "loss": 0.3967,
+      "step": 21153
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.418259941748134e-06,
+      "loss": 0.4847,
+      "step": 21154
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.417683190587422e-06,
+      "loss": 0.4332,
+      "step": 21155
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.41710643942671e-06,
+      "loss": 0.3868,
+      "step": 21156
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.416529688265999e-06,
+      "loss": 0.4405,
+      "step": 21157
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415952937105286e-06,
+      "loss": 0.3318,
+      "step": 21158
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.415376185944574e-06,
+      "loss": 0.4163,
+      "step": 21159
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.414799434783864e-06,
+      "loss": 0.4662,
+      "step": 21160
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.414222683623151e-06,
+      "loss": 0.6032,
+      "step": 21161
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.41364593246244e-06,
+      "loss": 0.4426,
+      "step": 21162
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.413069181301728e-06,
+      "loss": 0.4437,
+      "step": 21163
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.412492430141016e-06,
+      "loss": 0.4258,
+      "step": 21164
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.411915678980305e-06,
+      "loss": 0.4275,
+      "step": 21165
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.411338927819593e-06,
+      "loss": 0.389,
+      "step": 21166
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.41076217665888e-06,
+      "loss": 0.5149,
+      "step": 21167
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.41018542549817e-06,
+      "loss": 0.4593,
+      "step": 21168
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.409608674337458e-06,
+      "loss": 0.3059,
+      "step": 21169
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.409031923176745e-06,
+      "loss": 0.3705,
+      "step": 21170
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.408455172016035e-06,
+      "loss": 0.41,
+      "step": 21171
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.407878420855323e-06,
+      "loss": 0.4616,
+      "step": 21172
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.40730166969461e-06,
+      "loss": 0.4776,
+      "step": 21173
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.4067249185339e-06,
+      "loss": 0.4851,
+      "step": 21174
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.406148167373187e-06,
+      "loss": 0.373,
+      "step": 21175
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.405571416212475e-06,
+      "loss": 0.5097,
+      "step": 21176
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.404994665051764e-06,
+      "loss": 0.4575,
+      "step": 21177
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.404417913891052e-06,
+      "loss": 0.4864,
+      "step": 21178
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.403841162730341e-06,
+      "loss": 0.3945,
+      "step": 21179
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.403264411569629e-06,
+      "loss": 0.3682,
+      "step": 21180
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 8.402687660408917e-06,
+      "loss": 0.4277,
+      "step": 21181
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.402110909248206e-06,
+      "loss": 0.4628,
+      "step": 21182
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.401534158087494e-06,
+      "loss": 0.3486,
+      "step": 21183
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.400957406926782e-06,
+      "loss": 0.4541,
+      "step": 21184
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.400380655766071e-06,
+      "loss": 0.4073,
+      "step": 21185
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.399803904605359e-06,
+      "loss": 0.3754,
+      "step": 21186
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.399227153444646e-06,
+      "loss": 0.4836,
+      "step": 21187
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.398650402283936e-06,
+      "loss": 0.352,
+      "step": 21188
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.398073651123223e-06,
+      "loss": 0.52,
+      "step": 21189
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.397496899962511e-06,
+      "loss": 0.3977,
+      "step": 21190
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.3969201488018e-06,
+      "loss": 0.5016,
+      "step": 21191
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.396343397641088e-06,
+      "loss": 0.3265,
+      "step": 21192
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395766646480377e-06,
+      "loss": 0.5003,
+      "step": 21193
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.395189895319665e-06,
+      "loss": 0.4296,
+      "step": 21194
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.394613144158953e-06,
+      "loss": 0.5429,
+      "step": 21195
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.394036392998242e-06,
+      "loss": 0.4326,
+      "step": 21196
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.39345964183753e-06,
+      "loss": 0.4164,
+      "step": 21197
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.392882890676818e-06,
+      "loss": 0.3816,
+      "step": 21198
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.392306139516107e-06,
+      "loss": 0.5092,
+      "step": 21199
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.391729388355395e-06,
+      "loss": 0.468,
+      "step": 21200
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.391152637194682e-06,
+      "loss": 0.455,
+      "step": 21201
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.390575886033972e-06,
+      "loss": 0.4699,
+      "step": 21202
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.38999913487326e-06,
+      "loss": 0.4659,
+      "step": 21203
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.389422383712547e-06,
+      "loss": 0.459,
+      "step": 21204
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.388845632551836e-06,
+      "loss": 0.4106,
+      "step": 21205
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.388268881391124e-06,
+      "loss": 0.4289,
+      "step": 21206
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.387692130230412e-06,
+      "loss": 0.3708,
+      "step": 21207
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.387115379069701e-06,
+      "loss": 0.5217,
+      "step": 21208
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.386538627908989e-06,
+      "loss": 0.4207,
+      "step": 21209
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.385961876748278e-06,
+      "loss": 0.4049,
+      "step": 21210
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.385385125587566e-06,
+      "loss": 0.4338,
+      "step": 21211
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.384808374426854e-06,
+      "loss": 0.4065,
+      "step": 21212
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.384231623266143e-06,
+      "loss": 0.5444,
+      "step": 21213
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.38365487210543e-06,
+      "loss": 0.543,
+      "step": 21214
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.383078120944718e-06,
+      "loss": 0.4524,
+      "step": 21215
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.382501369784008e-06,
+      "loss": 0.4888,
+      "step": 21216
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.381924618623295e-06,
+      "loss": 0.458,
+      "step": 21217
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.381347867462583e-06,
+      "loss": 0.3835,
+      "step": 21218
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.380771116301873e-06,
+      "loss": 0.538,
+      "step": 21219
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.38019436514116e-06,
+      "loss": 0.5055,
+      "step": 21220
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.379617613980448e-06,
+      "loss": 0.4184,
+      "step": 21221
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.379040862819737e-06,
+      "loss": 0.4741,
+      "step": 21222
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.378464111659025e-06,
+      "loss": 0.4412,
+      "step": 21223
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.377887360498314e-06,
+      "loss": 0.3298,
+      "step": 21224
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.377310609337602e-06,
+      "loss": 0.384,
+      "step": 21225
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37673385817689e-06,
+      "loss": 0.4712,
+      "step": 21226
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.376157107016179e-06,
+      "loss": 0.4374,
+      "step": 21227
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.375580355855467e-06,
+      "loss": 0.4194,
+      "step": 21228
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.375003604694754e-06,
+      "loss": 0.4875,
+      "step": 21229
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.374426853534044e-06,
+      "loss": 0.352,
+      "step": 21230
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.373850102373332e-06,
+      "loss": 0.419,
+      "step": 21231
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.37327335121262e-06,
+      "loss": 0.4024,
+      "step": 21232
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.372696600051909e-06,
+      "loss": 0.3338,
+      "step": 21233
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.372119848891196e-06,
+      "loss": 0.4488,
+      "step": 21234
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.371543097730484e-06,
+      "loss": 0.4899,
+      "step": 21235
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.370966346569773e-06,
+      "loss": 0.4228,
+      "step": 21236
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.370389595409063e-06,
+      "loss": 0.4966,
+      "step": 21237
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.369812844248349e-06,
+      "loss": 0.4773,
+      "step": 21238
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.369236093087638e-06,
+      "loss": 0.5346,
+      "step": 21239
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.368659341926927e-06,
+      "loss": 0.388,
+      "step": 21240
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.368082590766215e-06,
+      "loss": 0.4861,
+      "step": 21241
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.367505839605503e-06,
+      "loss": 0.4968,
+      "step": 21242
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.36692908844479e-06,
+      "loss": 0.4472,
+      "step": 21243
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.36635233728408e-06,
+      "loss": 0.425,
+      "step": 21244
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.365775586123368e-06,
+      "loss": 0.423,
+      "step": 21245
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.365198834962655e-06,
+      "loss": 0.3481,
+      "step": 21246
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.364622083801945e-06,
+      "loss": 0.4601,
+      "step": 21247
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.364045332641232e-06,
+      "loss": 0.5315,
+      "step": 21248
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.36346858148052e-06,
+      "loss": 0.3389,
+      "step": 21249
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.36289183031981e-06,
+      "loss": 0.4225,
+      "step": 21250
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.362315079159097e-06,
+      "loss": 0.4601,
+      "step": 21251
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.361738327998385e-06,
+      "loss": 0.543,
+      "step": 21252
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.361161576837674e-06,
+      "loss": 0.5426,
+      "step": 21253
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.360584825676964e-06,
+      "loss": 0.5174,
+      "step": 21254
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.360008074516251e-06,
+      "loss": 0.4588,
+      "step": 21255
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.359431323355539e-06,
+      "loss": 0.5917,
+      "step": 21256
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.358854572194828e-06,
+      "loss": 0.4101,
+      "step": 21257
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.358277821034116e-06,
+      "loss": 0.4863,
+      "step": 21258
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.357701069873404e-06,
+      "loss": 0.4426,
+      "step": 21259
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.357124318712691e-06,
+      "loss": 0.3199,
+      "step": 21260
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.35654756755198e-06,
+      "loss": 0.2839,
+      "step": 21261
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.355970816391268e-06,
+      "loss": 0.3965,
+      "step": 21262
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.355394065230556e-06,
+      "loss": 0.3761,
+      "step": 21263
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.354817314069845e-06,
+      "loss": 0.4197,
+      "step": 21264
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.354240562909133e-06,
+      "loss": 0.4366,
+      "step": 21265
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.35366381174842e-06,
+      "loss": 0.4343,
+      "step": 21266
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.35308706058771e-06,
+      "loss": 0.4065,
+      "step": 21267
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.352510309427e-06,
+      "loss": 0.5254,
+      "step": 21268
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.351933558266286e-06,
+      "loss": 0.5606,
+      "step": 21269
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.351356807105575e-06,
+      "loss": 0.4013,
+      "step": 21270
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.350780055944864e-06,
+      "loss": 0.3173,
+      "step": 21271
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.350203304784152e-06,
+      "loss": 0.4875,
+      "step": 21272
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.34962655362344e-06,
+      "loss": 0.5028,
+      "step": 21273
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.349049802462729e-06,
+      "loss": 0.4797,
+      "step": 21274
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.348473051302017e-06,
+      "loss": 0.3705,
+      "step": 21275
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.347896300141304e-06,
+      "loss": 0.3962,
+      "step": 21276
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.347319548980592e-06,
+      "loss": 0.4298,
+      "step": 21277
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.346742797819882e-06,
+      "loss": 0.4617,
+      "step": 21278
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.34616604665917e-06,
+      "loss": 0.4472,
+      "step": 21279
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.345589295498457e-06,
+      "loss": 0.4684,
+      "step": 21280
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.345012544337746e-06,
+      "loss": 0.4559,
+      "step": 21281
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.344435793177034e-06,
+      "loss": 0.4212,
+      "step": 21282
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.343859042016322e-06,
+      "loss": 0.4097,
+      "step": 21283
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.343282290855611e-06,
+      "loss": 0.3012,
+      "step": 21284
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.3427055396949e-06,
+      "loss": 0.4638,
+      "step": 21285
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.342128788534188e-06,
+      "loss": 0.3499,
+      "step": 21286
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.341552037373476e-06,
+      "loss": 0.3427,
+      "step": 21287
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.340975286212765e-06,
+      "loss": 0.5195,
+      "step": 21288
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.340398535052053e-06,
+      "loss": 0.4253,
+      "step": 21289
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.33982178389134e-06,
+      "loss": 0.4093,
+      "step": 21290
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.33924503273063e-06,
+      "loss": 0.4072,
+      "step": 21291
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.338668281569918e-06,
+      "loss": 0.3343,
+      "step": 21292
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.338091530409205e-06,
+      "loss": 0.4493,
+      "step": 21293
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.337514779248493e-06,
+      "loss": 0.4482,
+      "step": 21294
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.336938028087782e-06,
+      "loss": 0.41,
+      "step": 21295
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.33636127692707e-06,
+      "loss": 0.3623,
+      "step": 21296
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.335784525766358e-06,
+      "loss": 0.4717,
+      "step": 21297
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.335207774605647e-06,
+      "loss": 0.4845,
+      "step": 21298
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.334631023444936e-06,
+      "loss": 0.3935,
+      "step": 21299
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.334054272284222e-06,
+      "loss": 0.5051,
+      "step": 21300
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.333477521123512e-06,
+      "loss": 0.3256,
+      "step": 21301
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.332900769962801e-06,
+      "loss": 0.5116,
+      "step": 21302
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.332324018802089e-06,
+      "loss": 0.5054,
+      "step": 21303
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.331747267641377e-06,
+      "loss": 0.3014,
+      "step": 21304
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.331170516480666e-06,
+      "loss": 0.4845,
+      "step": 21305
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.330593765319954e-06,
+      "loss": 0.5298,
+      "step": 21306
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.330017014159241e-06,
+      "loss": 0.4288,
+      "step": 21307
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.32944026299853e-06,
+      "loss": 0.4794,
+      "step": 21308
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.328863511837818e-06,
+      "loss": 0.5139,
+      "step": 21309
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.328286760677106e-06,
+      "loss": 0.4243,
+      "step": 21310
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.327710009516394e-06,
+      "loss": 0.3649,
+      "step": 21311
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.327133258355683e-06,
+      "loss": 0.5049,
+      "step": 21312
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.326556507194973e-06,
+      "loss": 0.2641,
+      "step": 21313
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.325979756034258e-06,
+      "loss": 0.4697,
+      "step": 21314
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.325403004873548e-06,
+      "loss": 0.4419,
+      "step": 21315
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.324826253712837e-06,
+      "loss": 0.4748,
+      "step": 21316
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.324249502552125e-06,
+      "loss": 0.4801,
+      "step": 21317
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.323672751391413e-06,
+      "loss": 0.2647,
+      "step": 21318
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.323096000230702e-06,
+      "loss": 0.4559,
+      "step": 21319
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.32251924906999e-06,
+      "loss": 0.3849,
+      "step": 21320
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.321942497909277e-06,
+      "loss": 0.3419,
+      "step": 21321
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.321365746748567e-06,
+      "loss": 0.5287,
+      "step": 21322
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.320788995587854e-06,
+      "loss": 0.4105,
+      "step": 21323
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.320212244427142e-06,
+      "loss": 0.3418,
+      "step": 21324
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.319635493266432e-06,
+      "loss": 0.4353,
+      "step": 21325
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.31905874210572e-06,
+      "loss": 0.3382,
+      "step": 21326
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.318481990945007e-06,
+      "loss": 0.4421,
+      "step": 21327
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.317905239784295e-06,
+      "loss": 0.445,
+      "step": 21328
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.317328488623584e-06,
+      "loss": 0.408,
+      "step": 21329
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.316751737462873e-06,
+      "loss": 0.3042,
+      "step": 21330
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.31617498630216e-06,
+      "loss": 0.4416,
+      "step": 21331
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.315598235141449e-06,
+      "loss": 0.4186,
+      "step": 21332
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.315021483980738e-06,
+      "loss": 0.4135,
+      "step": 21333
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.314444732820026e-06,
+      "loss": 0.4789,
+      "step": 21334
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.313867981659313e-06,
+      "loss": 0.4871,
+      "step": 21335
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.313291230498603e-06,
+      "loss": 0.4706,
+      "step": 21336
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.31271447933789e-06,
+      "loss": 0.5946,
+      "step": 21337
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.312137728177178e-06,
+      "loss": 0.4259,
+      "step": 21338
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.311560977016468e-06,
+      "loss": 0.4092,
+      "step": 21339
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.310984225855755e-06,
+      "loss": 0.5172,
+      "step": 21340
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.310407474695043e-06,
+      "loss": 0.4903,
+      "step": 21341
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.309830723534332e-06,
+      "loss": 0.4728,
+      "step": 21342
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.30925397237362e-06,
+      "loss": 0.453,
+      "step": 21343
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.30867722121291e-06,
+      "loss": 0.4724,
+      "step": 21344
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.308100470052195e-06,
+      "loss": 0.4588,
+      "step": 21345
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.307523718891485e-06,
+      "loss": 0.3789,
+      "step": 21346
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.306946967730774e-06,
+      "loss": 0.2604,
+      "step": 21347
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.306370216570062e-06,
+      "loss": 0.3542,
+      "step": 21348
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.30579346540935e-06,
+      "loss": 0.4439,
+      "step": 21349
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.305216714248639e-06,
+      "loss": 0.5284,
+      "step": 21350
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.304639963087927e-06,
+      "loss": 0.4682,
+      "step": 21351
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.304063211927214e-06,
+      "loss": 0.4177,
+      "step": 21352
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.303486460766504e-06,
+      "loss": 0.4529,
+      "step": 21353
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.302909709605791e-06,
+      "loss": 0.4157,
+      "step": 21354
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.302332958445079e-06,
+      "loss": 0.4353,
+      "step": 21355
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.301756207284368e-06,
+      "loss": 0.4019,
+      "step": 21356
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.301179456123656e-06,
+      "loss": 0.5955,
+      "step": 21357
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.300602704962944e-06,
+      "loss": 0.4768,
+      "step": 21358
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.300025953802233e-06,
+      "loss": 0.4778,
+      "step": 21359
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 8.29944920264152e-06,
+      "loss": 0.5911,
+      "step": 21360
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.29887245148081e-06,
+      "loss": 0.5962,
+      "step": 21361
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.298295700320098e-06,
+      "loss": 0.4187,
+      "step": 21362
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.297718949159386e-06,
+      "loss": 0.4462,
+      "step": 21363
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.297142197998675e-06,
+      "loss": 0.267,
+      "step": 21364
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.296565446837963e-06,
+      "loss": 0.3774,
+      "step": 21365
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.29598869567725e-06,
+      "loss": 0.4465,
+      "step": 21366
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.29541194451654e-06,
+      "loss": 0.4715,
+      "step": 21367
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.294835193355827e-06,
+      "loss": 0.3342,
+      "step": 21368
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.294258442195115e-06,
+      "loss": 0.4251,
+      "step": 21369
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.293681691034404e-06,
+      "loss": 0.4328,
+      "step": 21370
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.293104939873692e-06,
+      "loss": 0.3663,
+      "step": 21371
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.29252818871298e-06,
+      "loss": 0.3867,
+      "step": 21372
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.29195143755227e-06,
+      "loss": 0.4868,
+      "step": 21373
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.291374686391557e-06,
+      "loss": 0.3992,
+      "step": 21374
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.290797935230846e-06,
+      "loss": 0.4276,
+      "step": 21375
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.290221184070134e-06,
+      "loss": 0.4181,
+      "step": 21376
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.289644432909422e-06,
+      "loss": 0.481,
+      "step": 21377
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.289067681748711e-06,
+      "loss": 0.4905,
+      "step": 21378
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.288490930587999e-06,
+      "loss": 0.4594,
+      "step": 21379
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.287914179427286e-06,
+      "loss": 0.4699,
+      "step": 21380
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.287337428266576e-06,
+      "loss": 0.5045,
+      "step": 21381
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.286760677105863e-06,
+      "loss": 0.4707,
+      "step": 21382
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.286183925945151e-06,
+      "loss": 0.4176,
+      "step": 21383
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.28560717478444e-06,
+      "loss": 0.4577,
+      "step": 21384
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.285030423623728e-06,
+      "loss": 0.4608,
+      "step": 21385
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.284453672463016e-06,
+      "loss": 0.3551,
+      "step": 21386
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.283876921302305e-06,
+      "loss": 0.3813,
+      "step": 21387
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.283300170141593e-06,
+      "loss": 0.4214,
+      "step": 21388
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.28272341898088e-06,
+      "loss": 0.3746,
+      "step": 21389
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.28214666782017e-06,
+      "loss": 0.4158,
+      "step": 21390
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.281569916659458e-06,
+      "loss": 0.4316,
+      "step": 21391
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.280993165498747e-06,
+      "loss": 0.4689,
+      "step": 21392
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.280416414338035e-06,
+      "loss": 0.4897,
+      "step": 21393
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.279839663177322e-06,
+      "loss": 0.5291,
+      "step": 21394
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.279262912016612e-06,
+      "loss": 0.4871,
+      "step": 21395
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.2786861608559e-06,
+      "loss": 0.3933,
+      "step": 21396
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.278109409695187e-06,
+      "loss": 0.3535,
+      "step": 21397
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.277532658534477e-06,
+      "loss": 0.3792,
+      "step": 21398
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.276955907373764e-06,
+      "loss": 0.4886,
+      "step": 21399
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.276379156213052e-06,
+      "loss": 0.4415,
+      "step": 21400
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.275802405052341e-06,
+      "loss": 0.3221,
+      "step": 21401
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.275225653891629e-06,
+      "loss": 0.4496,
+      "step": 21402
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.274648902730917e-06,
+      "loss": 0.4231,
+      "step": 21403
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.274072151570206e-06,
+      "loss": 0.4201,
+      "step": 21404
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.273495400409494e-06,
+      "loss": 0.3011,
+      "step": 21405
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.272918649248783e-06,
+      "loss": 0.3078,
+      "step": 21406
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.27234189808807e-06,
+      "loss": 0.4926,
+      "step": 21407
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271765146927358e-06,
+      "loss": 0.3712,
+      "step": 21408
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.271188395766648e-06,
+      "loss": 0.5091,
+      "step": 21409
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.270611644605936e-06,
+      "loss": 0.3489,
+      "step": 21410
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.270034893445223e-06,
+      "loss": 0.3278,
+      "step": 21411
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.269458142284513e-06,
+      "loss": 0.4449,
+      "step": 21412
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.2688813911238e-06,
+      "loss": 0.498,
+      "step": 21413
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.268304639963088e-06,
+      "loss": 0.4993,
+      "step": 21414
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.267727888802377e-06,
+      "loss": 0.3577,
+      "step": 21415
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.267151137641665e-06,
+      "loss": 0.5184,
+      "step": 21416
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.266574386480953e-06,
+      "loss": 0.4816,
+      "step": 21417
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.265997635320242e-06,
+      "loss": 0.3843,
+      "step": 21418
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.26542088415953e-06,
+      "loss": 0.4155,
+      "step": 21419
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.264844132998817e-06,
+      "loss": 0.4111,
+      "step": 21420
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.264267381838107e-06,
+      "loss": 0.3342,
+      "step": 21421
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.263690630677395e-06,
+      "loss": 0.4156,
+      "step": 21422
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.263113879516684e-06,
+      "loss": 0.4503,
+      "step": 21423
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.262537128355972e-06,
+      "loss": 0.2214,
+      "step": 21424
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.26196037719526e-06,
+      "loss": 0.4657,
+      "step": 21425
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.261383626034549e-06,
+      "loss": 0.4322,
+      "step": 21426
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.260806874873836e-06,
+      "loss": 0.5582,
+      "step": 21427
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.260230123713124e-06,
+      "loss": 0.444,
+      "step": 21428
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.259653372552413e-06,
+      "loss": 0.4219,
+      "step": 21429
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.259076621391701e-06,
+      "loss": 0.4759,
+      "step": 21430
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.258499870230989e-06,
+      "loss": 0.4655,
+      "step": 21431
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.257923119070278e-06,
+      "loss": 0.5173,
+      "step": 21432
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.257346367909566e-06,
+      "loss": 0.5594,
+      "step": 21433
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.256769616748854e-06,
+      "loss": 0.3795,
+      "step": 21434
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.256192865588143e-06,
+      "loss": 0.3861,
+      "step": 21435
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.25561611442743e-06,
+      "loss": 0.3847,
+      "step": 21436
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.25503936326672e-06,
+      "loss": 0.4837,
+      "step": 21437
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.254462612106008e-06,
+      "loss": 0.3576,
+      "step": 21438
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.253885860945295e-06,
+      "loss": 0.5552,
+      "step": 21439
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.253309109784585e-06,
+      "loss": 0.3689,
+      "step": 21440
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.252732358623872e-06,
+      "loss": 0.5129,
+      "step": 21441
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.25215560746316e-06,
+      "loss": 0.2745,
+      "step": 21442
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.25157885630245e-06,
+      "loss": 0.5004,
+      "step": 21443
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.251002105141737e-06,
+      "loss": 0.3459,
+      "step": 21444
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.250425353981025e-06,
+      "loss": 0.4486,
+      "step": 21445
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.249848602820314e-06,
+      "loss": 0.4611,
+      "step": 21446
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.249271851659602e-06,
+      "loss": 0.4092,
+      "step": 21447
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.24869510049889e-06,
+      "loss": 0.5296,
+      "step": 21448
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.248118349338179e-06,
+      "loss": 0.4326,
+      "step": 21449
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.247541598177467e-06,
+      "loss": 0.4154,
+      "step": 21450
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.246964847016754e-06,
+      "loss": 0.4344,
+      "step": 21451
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.246388095856044e-06,
+      "loss": 0.4032,
+      "step": 21452
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.245811344695331e-06,
+      "loss": 0.4509,
+      "step": 21453
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.24523459353462e-06,
+      "loss": 0.5511,
+      "step": 21454
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.244657842373908e-06,
+      "loss": 0.4211,
+      "step": 21455
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.244081091213196e-06,
+      "loss": 0.4342,
+      "step": 21456
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.243504340052486e-06,
+      "loss": 0.4626,
+      "step": 21457
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.242927588891773e-06,
+      "loss": 0.3373,
+      "step": 21458
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.242350837731061e-06,
+      "loss": 0.4261,
+      "step": 21459
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.24177408657035e-06,
+      "loss": 0.4673,
+      "step": 21460
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.241197335409638e-06,
+      "loss": 0.5048,
+      "step": 21461
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.240620584248926e-06,
+      "loss": 0.4761,
+      "step": 21462
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.240043833088215e-06,
+      "loss": 0.4432,
+      "step": 21463
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.239467081927503e-06,
+      "loss": 0.4284,
+      "step": 21464
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23889033076679e-06,
+      "loss": 0.3073,
+      "step": 21465
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23831357960608e-06,
+      "loss": 0.5661,
+      "step": 21466
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.237736828445367e-06,
+      "loss": 0.5051,
+      "step": 21467
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.237160077284657e-06,
+      "loss": 0.323,
+      "step": 21468
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.236583326123945e-06,
+      "loss": 0.4042,
+      "step": 21469
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.236006574963232e-06,
+      "loss": 0.4821,
+      "step": 21470
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.235429823802522e-06,
+      "loss": 0.4148,
+      "step": 21471
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.23485307264181e-06,
+      "loss": 0.4585,
+      "step": 21472
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.234276321481097e-06,
+      "loss": 0.456,
+      "step": 21473
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.233699570320386e-06,
+      "loss": 0.3381,
+      "step": 21474
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.233122819159674e-06,
+      "loss": 0.4543,
+      "step": 21475
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.232546067998962e-06,
+      "loss": 0.522,
+      "step": 21476
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.231969316838251e-06,
+      "loss": 0.4546,
+      "step": 21477
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.231392565677539e-06,
+      "loss": 0.3793,
+      "step": 21478
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.230815814516826e-06,
+      "loss": 0.3789,
+      "step": 21479
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.230239063356116e-06,
+      "loss": 0.5269,
+      "step": 21480
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.229662312195405e-06,
+      "loss": 0.4589,
+      "step": 21481
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.229085561034691e-06,
+      "loss": 0.4478,
+      "step": 21482
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.22850880987398e-06,
+      "loss": 0.3498,
+      "step": 21483
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.22793205871327e-06,
+      "loss": 0.2464,
+      "step": 21484
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.227355307552558e-06,
+      "loss": 0.4578,
+      "step": 21485
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.226778556391845e-06,
+      "loss": 0.4974,
+      "step": 21486
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.226201805231133e-06,
+      "loss": 0.4601,
+      "step": 21487
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.225625054070422e-06,
+      "loss": 0.4268,
+      "step": 21488
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.22504830290971e-06,
+      "loss": 0.5265,
+      "step": 21489
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.224471551748998e-06,
+      "loss": 0.4171,
+      "step": 21490
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.223894800588287e-06,
+      "loss": 0.3361,
+      "step": 21491
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.223318049427575e-06,
+      "loss": 0.5466,
+      "step": 21492
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.222741298266863e-06,
+      "loss": 0.3875,
+      "step": 21493
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.222164547106152e-06,
+      "loss": 0.4512,
+      "step": 21494
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.22158779594544e-06,
+      "loss": 0.4458,
+      "step": 21495
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.221011044784727e-06,
+      "loss": 0.4198,
+      "step": 21496
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.220434293624017e-06,
+      "loss": 0.3986,
+      "step": 21497
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.219857542463306e-06,
+      "loss": 0.3963,
+      "step": 21498
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.219280791302594e-06,
+      "loss": 0.3761,
+      "step": 21499
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.218704040141881e-06,
+      "loss": 0.3952,
+      "step": 21500
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.21812728898117e-06,
+      "loss": 0.5787,
+      "step": 21501
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.217550537820458e-06,
+      "loss": 0.4343,
+      "step": 21502
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.216973786659746e-06,
+      "loss": 0.4496,
+      "step": 21503
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.216397035499034e-06,
+      "loss": 0.3376,
+      "step": 21504
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.215820284338323e-06,
+      "loss": 0.3118,
+      "step": 21505
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.215243533177611e-06,
+      "loss": 0.439,
+      "step": 21506
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.214666782016899e-06,
+      "loss": 0.4533,
+      "step": 21507
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.214090030856188e-06,
+      "loss": 0.3221,
+      "step": 21508
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.213513279695476e-06,
+      "loss": 0.4142,
+      "step": 21509
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.212936528534763e-06,
+      "loss": 0.3699,
+      "step": 21510
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.212359777374053e-06,
+      "loss": 0.4279,
+      "step": 21511
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.211783026213342e-06,
+      "loss": 0.5011,
+      "step": 21512
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.211206275052628e-06,
+      "loss": 0.3987,
+      "step": 21513
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.210629523891917e-06,
+      "loss": 0.6447,
+      "step": 21514
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.210052772731207e-06,
+      "loss": 0.3638,
+      "step": 21515
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.209476021570495e-06,
+      "loss": 0.5073,
+      "step": 21516
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.208899270409782e-06,
+      "loss": 0.4479,
+      "step": 21517
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.208322519249072e-06,
+      "loss": 0.5497,
+      "step": 21518
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.20774576808836e-06,
+      "loss": 0.4045,
+      "step": 21519
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.207169016927647e-06,
+      "loss": 0.5277,
+      "step": 21520
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.206592265766935e-06,
+      "loss": 0.3235,
+      "step": 21521
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.206015514606224e-06,
+      "loss": 0.3831,
+      "step": 21522
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.205438763445512e-06,
+      "loss": 0.4609,
+      "step": 21523
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.2048620122848e-06,
+      "loss": 0.3862,
+      "step": 21524
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.204285261124089e-06,
+      "loss": 0.4649,
+      "step": 21525
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.203708509963376e-06,
+      "loss": 0.4276,
+      "step": 21526
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.203131758802664e-06,
+      "loss": 0.4247,
+      "step": 21527
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.202555007641954e-06,
+      "loss": 0.4085,
+      "step": 21528
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.201978256481243e-06,
+      "loss": 0.3612,
+      "step": 21529
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.20140150532053e-06,
+      "loss": 0.3518,
+      "step": 21530
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.200824754159818e-06,
+      "loss": 0.407,
+      "step": 21531
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.200248002999108e-06,
+      "loss": 0.4531,
+      "step": 21532
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.199671251838395e-06,
+      "loss": 0.4391,
+      "step": 21533
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.199094500677683e-06,
+      "loss": 0.4383,
+      "step": 21534
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.198517749516972e-06,
+      "loss": 0.5107,
+      "step": 21535
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.19794099835626e-06,
+      "loss": 0.4319,
+      "step": 21536
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.197364247195548e-06,
+      "loss": 0.4447,
+      "step": 21537
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.196787496034835e-06,
+      "loss": 0.4238,
+      "step": 21538
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 8.196210744874125e-06,
+      "loss": 0.5112,
+      "step": 21539
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.195633993713413e-06,
+      "loss": 0.415,
+      "step": 21540
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.1950572425527e-06,
+      "loss": 0.4233,
+      "step": 21541
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.19448049139199e-06,
+      "loss": 0.4017,
+      "step": 21542
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.193903740231279e-06,
+      "loss": 0.4405,
+      "step": 21543
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.193326989070565e-06,
+      "loss": 0.5728,
+      "step": 21544
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.192750237909854e-06,
+      "loss": 0.3637,
+      "step": 21545
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.192173486749144e-06,
+      "loss": 0.3786,
+      "step": 21546
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.191596735588431e-06,
+      "loss": 0.4686,
+      "step": 21547
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.191019984427719e-06,
+      "loss": 0.4713,
+      "step": 21548
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.190443233267008e-06,
+      "loss": 0.4906,
+      "step": 21549
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189866482106296e-06,
+      "loss": 0.5103,
+      "step": 21550
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.189289730945584e-06,
+      "loss": 0.587,
+      "step": 21551
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.188712979784873e-06,
+      "loss": 0.3844,
+      "step": 21552
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.188136228624161e-06,
+      "loss": 0.6041,
+      "step": 21553
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.187559477463449e-06,
+      "loss": 0.447,
+      "step": 21554
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.186982726302736e-06,
+      "loss": 0.3949,
+      "step": 21555
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.186405975142026e-06,
+      "loss": 0.4805,
+      "step": 21556
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.185829223981313e-06,
+      "loss": 0.4468,
+      "step": 21557
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.185252472820601e-06,
+      "loss": 0.5329,
+      "step": 21558
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.18467572165989e-06,
+      "loss": 0.5226,
+      "step": 21559
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.18409897049918e-06,
+      "loss": 0.4086,
+      "step": 21560
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.183522219338467e-06,
+      "loss": 0.3446,
+      "step": 21561
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.182945468177755e-06,
+      "loss": 0.4223,
+      "step": 21562
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.182368717017045e-06,
+      "loss": 0.4053,
+      "step": 21563
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.181791965856332e-06,
+      "loss": 0.4123,
+      "step": 21564
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.18121521469562e-06,
+      "loss": 0.3782,
+      "step": 21565
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.18063846353491e-06,
+      "loss": 0.4301,
+      "step": 21566
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.180061712374197e-06,
+      "loss": 0.48,
+      "step": 21567
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.179484961213485e-06,
+      "loss": 0.5228,
+      "step": 21568
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.178908210052774e-06,
+      "loss": 0.4597,
+      "step": 21569
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.178331458892062e-06,
+      "loss": 0.5147,
+      "step": 21570
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.17775470773135e-06,
+      "loss": 0.3618,
+      "step": 21571
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.177177956570637e-06,
+      "loss": 0.4776,
+      "step": 21572
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.176601205409926e-06,
+      "loss": 0.4016,
+      "step": 21573
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.176024454249216e-06,
+      "loss": 0.2584,
+      "step": 21574
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.175447703088502e-06,
+      "loss": 0.4235,
+      "step": 21575
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.174870951927791e-06,
+      "loss": 0.5354,
+      "step": 21576
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.17429420076708e-06,
+      "loss": 0.4658,
+      "step": 21577
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.173717449606368e-06,
+      "loss": 0.434,
+      "step": 21578
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.173140698445656e-06,
+      "loss": 0.3651,
+      "step": 21579
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.172563947284945e-06,
+      "loss": 0.469,
+      "step": 21580
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.171987196124233e-06,
+      "loss": 0.4451,
+      "step": 21581
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.17141044496352e-06,
+      "loss": 0.4171,
+      "step": 21582
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.17083369380281e-06,
+      "loss": 0.3804,
+      "step": 21583
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.170256942642098e-06,
+      "loss": 0.5209,
+      "step": 21584
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.169680191481385e-06,
+      "loss": 0.2552,
+      "step": 21585
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.169103440320675e-06,
+      "loss": 0.442,
+      "step": 21586
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.168526689159963e-06,
+      "loss": 0.5499,
+      "step": 21587
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.16794993799925e-06,
+      "loss": 0.4343,
+      "step": 21588
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.167373186838538e-06,
+      "loss": 0.4778,
+      "step": 21589
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.166796435677827e-06,
+      "loss": 0.5315,
+      "step": 21590
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.166219684517117e-06,
+      "loss": 0.3371,
+      "step": 21591
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.165642933356404e-06,
+      "loss": 0.5499,
+      "step": 21592
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.165066182195692e-06,
+      "loss": 0.3827,
+      "step": 21593
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.164489431034981e-06,
+      "loss": 0.3679,
+      "step": 21594
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.163912679874269e-06,
+      "loss": 0.3227,
+      "step": 21595
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.163335928713557e-06,
+      "loss": 0.5458,
+      "step": 21596
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.162759177552846e-06,
+      "loss": 0.5097,
+      "step": 21597
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.162182426392134e-06,
+      "loss": 0.4256,
+      "step": 21598
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.161605675231422e-06,
+      "loss": 0.4531,
+      "step": 21599
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.161028924070711e-06,
+      "loss": 0.3794,
+      "step": 21600
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.160452172909999e-06,
+      "loss": 0.4295,
+      "step": 21601
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.159875421749286e-06,
+      "loss": 0.4243,
+      "step": 21602
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.159298670588576e-06,
+      "loss": 0.578,
+      "step": 21603
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.158721919427863e-06,
+      "loss": 0.4579,
+      "step": 21604
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.158145168267153e-06,
+      "loss": 0.4116,
+      "step": 21605
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.157568417106439e-06,
+      "loss": 0.4304,
+      "step": 21606
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.156991665945728e-06,
+      "loss": 0.4573,
+      "step": 21607
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.156414914785017e-06,
+      "loss": 0.4875,
+      "step": 21608
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.155838163624305e-06,
+      "loss": 0.4423,
+      "step": 21609
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.155261412463593e-06,
+      "loss": 0.423,
+      "step": 21610
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.154684661302882e-06,
+      "loss": 0.4262,
+      "step": 21611
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.15410791014217e-06,
+      "loss": 0.4466,
+      "step": 21612
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.153531158981458e-06,
+      "loss": 0.4244,
+      "step": 21613
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.152954407820747e-06,
+      "loss": 0.5101,
+      "step": 21614
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.152377656660035e-06,
+      "loss": 0.4964,
+      "step": 21615
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.151800905499322e-06,
+      "loss": 0.3705,
+      "step": 21616
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.151224154338612e-06,
+      "loss": 0.4411,
+      "step": 21617
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.1506474031779e-06,
+      "loss": 0.4485,
+      "step": 21618
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.150070652017187e-06,
+      "loss": 0.4086,
+      "step": 21619
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.149493900856476e-06,
+      "loss": 0.3961,
+      "step": 21620
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148917149695764e-06,
+      "loss": 0.4505,
+      "step": 21621
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.148340398535054e-06,
+      "loss": 0.5813,
+      "step": 21622
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.147763647374341e-06,
+      "loss": 0.4587,
+      "step": 21623
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.147186896213629e-06,
+      "loss": 0.4241,
+      "step": 21624
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.146610145052918e-06,
+      "loss": 0.4924,
+      "step": 21625
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.146033393892206e-06,
+      "loss": 0.3968,
+      "step": 21626
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.145456642731494e-06,
+      "loss": 0.4827,
+      "step": 21627
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.144879891570783e-06,
+      "loss": 0.3019,
+      "step": 21628
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.14430314041007e-06,
+      "loss": 0.4018,
+      "step": 21629
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.143726389249358e-06,
+      "loss": 0.3341,
+      "step": 21630
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.143149638088648e-06,
+      "loss": 0.4312,
+      "step": 21631
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.142572886927935e-06,
+      "loss": 0.4167,
+      "step": 21632
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.141996135767223e-06,
+      "loss": 0.564,
+      "step": 21633
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.141419384606513e-06,
+      "loss": 0.5038,
+      "step": 21634
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.1408426334458e-06,
+      "loss": 0.4596,
+      "step": 21635
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.14026588228509e-06,
+      "loss": 0.4416,
+      "step": 21636
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.139689131124377e-06,
+      "loss": 0.4152,
+      "step": 21637
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.139112379963665e-06,
+      "loss": 0.4433,
+      "step": 21638
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.138535628802954e-06,
+      "loss": 0.3772,
+      "step": 21639
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.137958877642242e-06,
+      "loss": 0.4877,
+      "step": 21640
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.13738212648153e-06,
+      "loss": 0.4938,
+      "step": 21641
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.136805375320819e-06,
+      "loss": 0.3932,
+      "step": 21642
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.136228624160107e-06,
+      "loss": 0.4714,
+      "step": 21643
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.135651872999394e-06,
+      "loss": 0.3669,
+      "step": 21644
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.135075121838684e-06,
+      "loss": 0.4306,
+      "step": 21645
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.134498370677972e-06,
+      "loss": 0.3399,
+      "step": 21646
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.13392161951726e-06,
+      "loss": 0.5216,
+      "step": 21647
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.133344868356549e-06,
+      "loss": 0.4627,
+      "step": 21648
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.132768117195836e-06,
+      "loss": 0.335,
+      "step": 21649
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.132191366035124e-06,
+      "loss": 0.3562,
+      "step": 21650
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.131614614874413e-06,
+      "loss": 0.5217,
+      "step": 21651
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.131037863713701e-06,
+      "loss": 0.4581,
+      "step": 21652
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.13046111255299e-06,
+      "loss": 0.3425,
+      "step": 21653
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.129884361392278e-06,
+      "loss": 0.3855,
+      "step": 21654
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.129307610231566e-06,
+      "loss": 0.485,
+      "step": 21655
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.128730859070855e-06,
+      "loss": 0.461,
+      "step": 21656
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.128154107910143e-06,
+      "loss": 0.403,
+      "step": 21657
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.12757735674943e-06,
+      "loss": 0.3903,
+      "step": 21658
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.12700060558872e-06,
+      "loss": 0.4539,
+      "step": 21659
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.126423854428008e-06,
+      "loss": 0.5269,
+      "step": 21660
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.125847103267295e-06,
+      "loss": 0.4224,
+      "step": 21661
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.125270352106585e-06,
+      "loss": 0.5109,
+      "step": 21662
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.124693600945872e-06,
+      "loss": 0.3758,
+      "step": 21663
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.12411684978516e-06,
+      "loss": 0.4628,
+      "step": 21664
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.12354009862445e-06,
+      "loss": 0.4803,
+      "step": 21665
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.122963347463737e-06,
+      "loss": 0.3977,
+      "step": 21666
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.122386596303026e-06,
+      "loss": 0.4009,
+      "step": 21667
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.121809845142314e-06,
+      "loss": 0.4887,
+      "step": 21668
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.121233093981602e-06,
+      "loss": 0.3206,
+      "step": 21669
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.120656342820891e-06,
+      "loss": 0.5027,
+      "step": 21670
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.120079591660179e-06,
+      "loss": 0.4866,
+      "step": 21671
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.119502840499467e-06,
+      "loss": 0.4231,
+      "step": 21672
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.118926089338756e-06,
+      "loss": 0.4563,
+      "step": 21673
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.118349338178044e-06,
+      "loss": 0.4601,
+      "step": 21674
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.117772587017331e-06,
+      "loss": 0.4899,
+      "step": 21675
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.11719583585662e-06,
+      "loss": 0.4053,
+      "step": 21676
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.116619084695908e-06,
+      "loss": 0.3146,
+      "step": 21677
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.116042333535196e-06,
+      "loss": 0.4212,
+      "step": 21678
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.115465582374485e-06,
+      "loss": 0.5291,
+      "step": 21679
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.114888831213773e-06,
+      "loss": 0.4453,
+      "step": 21680
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.114312080053063e-06,
+      "loss": 0.5401,
+      "step": 21681
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.11373532889235e-06,
+      "loss": 0.4914,
+      "step": 21682
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.113158577731638e-06,
+      "loss": 0.3942,
+      "step": 21683
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.112581826570927e-06,
+      "loss": 0.4062,
+      "step": 21684
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.112005075410215e-06,
+      "loss": 0.3868,
+      "step": 21685
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.111428324249503e-06,
+      "loss": 0.373,
+      "step": 21686
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.110851573088792e-06,
+      "loss": 0.4498,
+      "step": 21687
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.11027482192808e-06,
+      "loss": 0.3376,
+      "step": 21688
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.109698070767367e-06,
+      "loss": 0.3856,
+      "step": 21689
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.109121319606657e-06,
+      "loss": 0.4003,
+      "step": 21690
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.108544568445944e-06,
+      "loss": 0.4046,
+      "step": 21691
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.107967817285232e-06,
+      "loss": 0.4464,
+      "step": 21692
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.107391066124522e-06,
+      "loss": 0.4644,
+      "step": 21693
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.10681431496381e-06,
+      "loss": 0.5502,
+      "step": 21694
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.106237563803097e-06,
+      "loss": 0.5365,
+      "step": 21695
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.105660812642386e-06,
+      "loss": 0.4558,
+      "step": 21696
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.105084061481674e-06,
+      "loss": 0.4063,
+      "step": 21697
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.104507310320963e-06,
+      "loss": 0.4532,
+      "step": 21698
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.103930559160251e-06,
+      "loss": 0.3694,
+      "step": 21699
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.103353807999539e-06,
+      "loss": 0.3269,
+      "step": 21700
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.102777056838828e-06,
+      "loss": 0.4068,
+      "step": 21701
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.102200305678116e-06,
+      "loss": 0.4452,
+      "step": 21702
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.101623554517403e-06,
+      "loss": 0.446,
+      "step": 21703
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.101046803356693e-06,
+      "loss": 0.4304,
+      "step": 21704
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.10047005219598e-06,
+      "loss": 0.4971,
+      "step": 21705
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.099893301035268e-06,
+      "loss": 0.3273,
+      "step": 21706
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.099316549874558e-06,
+      "loss": 0.484,
+      "step": 21707
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.098739798713845e-06,
+      "loss": 0.4579,
+      "step": 21708
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.098163047553133e-06,
+      "loss": 0.4475,
+      "step": 21709
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.097586296392422e-06,
+      "loss": 0.4216,
+      "step": 21710
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.09700954523171e-06,
+      "loss": 0.48,
+      "step": 21711
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.096432794071e-06,
+      "loss": 0.4578,
+      "step": 21712
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.095856042910287e-06,
+      "loss": 0.437,
+      "step": 21713
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.095279291749575e-06,
+      "loss": 0.5149,
+      "step": 21714
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.094702540588864e-06,
+      "loss": 0.4305,
+      "step": 21715
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.094125789428152e-06,
+      "loss": 0.5386,
+      "step": 21716
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.09354903826744e-06,
+      "loss": 0.3755,
+      "step": 21717
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 8.092972287106729e-06,
+      "loss": 0.3577,
+      "step": 21718
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.092395535946017e-06,
+      "loss": 0.5075,
+      "step": 21719
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.091818784785304e-06,
+      "loss": 0.4753,
+      "step": 21720
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.091242033624594e-06,
+      "loss": 0.602,
+      "step": 21721
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.090665282463881e-06,
+      "loss": 0.5703,
+      "step": 21722
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.090088531303169e-06,
+      "loss": 0.5112,
+      "step": 21723
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.089511780142458e-06,
+      "loss": 0.4822,
+      "step": 21724
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.088935028981748e-06,
+      "loss": 0.4022,
+      "step": 21725
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.088358277821034e-06,
+      "loss": 0.4563,
+      "step": 21726
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.087781526660323e-06,
+      "loss": 0.4115,
+      "step": 21727
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.087204775499611e-06,
+      "loss": 0.3396,
+      "step": 21728
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.0866280243389e-06,
+      "loss": 0.4767,
+      "step": 21729
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.086051273178188e-06,
+      "loss": 0.3421,
+      "step": 21730
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.085474522017476e-06,
+      "loss": 0.4639,
+      "step": 21731
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.084897770856765e-06,
+      "loss": 0.3934,
+      "step": 21732
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.084321019696053e-06,
+      "loss": 0.4588,
+      "step": 21733
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08374426853534e-06,
+      "loss": 0.4327,
+      "step": 21734
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08316751737463e-06,
+      "loss": 0.4023,
+      "step": 21735
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.082590766213917e-06,
+      "loss": 0.5003,
+      "step": 21736
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.082014015053205e-06,
+      "loss": 0.4165,
+      "step": 21737
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.081437263892494e-06,
+      "loss": 0.3347,
+      "step": 21738
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.080860512731782e-06,
+      "loss": 0.5352,
+      "step": 21739
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.08028376157107e-06,
+      "loss": 0.3986,
+      "step": 21740
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.07970701041036e-06,
+      "loss": 0.4434,
+      "step": 21741
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.079130259249649e-06,
+      "loss": 0.5107,
+      "step": 21742
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.078553508088936e-06,
+      "loss": 0.2547,
+      "step": 21743
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.077976756928224e-06,
+      "loss": 0.5088,
+      "step": 21744
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.077400005767513e-06,
+      "loss": 0.4543,
+      "step": 21745
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.076823254606801e-06,
+      "loss": 0.376,
+      "step": 21746
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.076246503446089e-06,
+      "loss": 0.3795,
+      "step": 21747
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.075669752285376e-06,
+      "loss": 0.2494,
+      "step": 21748
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.075093001124666e-06,
+      "loss": 0.3789,
+      "step": 21749
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.074516249963953e-06,
+      "loss": 0.4506,
+      "step": 21750
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.073939498803241e-06,
+      "loss": 0.3692,
+      "step": 21751
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.07336274764253e-06,
+      "loss": 0.4535,
+      "step": 21752
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.072785996481818e-06,
+      "loss": 0.3913,
+      "step": 21753
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.072209245321106e-06,
+      "loss": 0.6219,
+      "step": 21754
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.071632494160395e-06,
+      "loss": 0.4488,
+      "step": 21755
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.071055742999685e-06,
+      "loss": 0.508,
+      "step": 21756
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.07047899183897e-06,
+      "loss": 0.3537,
+      "step": 21757
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.06990224067826e-06,
+      "loss": 0.4533,
+      "step": 21758
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.06932548951755e-06,
+      "loss": 0.3525,
+      "step": 21759
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.068748738356837e-06,
+      "loss": 0.415,
+      "step": 21760
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.068171987196125e-06,
+      "loss": 0.4809,
+      "step": 21761
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.067595236035414e-06,
+      "loss": 0.4771,
+      "step": 21762
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.067018484874702e-06,
+      "loss": 0.4135,
+      "step": 21763
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.06644173371399e-06,
+      "loss": 0.3538,
+      "step": 21764
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065864982553277e-06,
+      "loss": 0.4553,
+      "step": 21765
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.065288231392567e-06,
+      "loss": 0.5567,
+      "step": 21766
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.064711480231854e-06,
+      "loss": 0.5802,
+      "step": 21767
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.064134729071142e-06,
+      "loss": 0.453,
+      "step": 21768
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.063557977910431e-06,
+      "loss": 0.5453,
+      "step": 21769
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.062981226749719e-06,
+      "loss": 0.4173,
+      "step": 21770
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.062404475589007e-06,
+      "loss": 0.4106,
+      "step": 21771
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.061827724428296e-06,
+      "loss": 0.3226,
+      "step": 21772
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.061250973267585e-06,
+      "loss": 0.5394,
+      "step": 21773
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.060674222106873e-06,
+      "loss": 0.3484,
+      "step": 21774
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.060097470946161e-06,
+      "loss": 0.4959,
+      "step": 21775
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.05952071978545e-06,
+      "loss": 0.5418,
+      "step": 21776
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.058943968624738e-06,
+      "loss": 0.3785,
+      "step": 21777
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.058367217464026e-06,
+      "loss": 0.4682,
+      "step": 21778
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.057790466303315e-06,
+      "loss": 0.4096,
+      "step": 21779
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.057213715142603e-06,
+      "loss": 0.4004,
+      "step": 21780
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.05663696398189e-06,
+      "loss": 0.3272,
+      "step": 21781
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.056060212821178e-06,
+      "loss": 0.376,
+      "step": 21782
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.055483461660467e-06,
+      "loss": 0.448,
+      "step": 21783
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.054906710499755e-06,
+      "loss": 0.4782,
+      "step": 21784
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.054329959339043e-06,
+      "loss": 0.4007,
+      "step": 21785
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.053753208178332e-06,
+      "loss": 0.4123,
+      "step": 21786
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.053176457017622e-06,
+      "loss": 0.4164,
+      "step": 21787
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.052599705856908e-06,
+      "loss": 0.4457,
+      "step": 21788
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.052022954696197e-06,
+      "loss": 0.3968,
+      "step": 21789
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.051446203535486e-06,
+      "loss": 0.4355,
+      "step": 21790
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.050869452374774e-06,
+      "loss": 0.4866,
+      "step": 21791
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.050292701214062e-06,
+      "loss": 0.3762,
+      "step": 21792
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.049715950053351e-06,
+      "loss": 0.397,
+      "step": 21793
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.049139198892639e-06,
+      "loss": 0.5607,
+      "step": 21794
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.048562447731926e-06,
+      "loss": 0.5493,
+      "step": 21795
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.047985696571216e-06,
+      "loss": 0.3421,
+      "step": 21796
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.047408945410503e-06,
+      "loss": 0.5389,
+      "step": 21797
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.046832194249791e-06,
+      "loss": 0.4524,
+      "step": 21798
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.046255443089079e-06,
+      "loss": 0.4486,
+      "step": 21799
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045678691928368e-06,
+      "loss": 0.3744,
+      "step": 21800
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.045101940767656e-06,
+      "loss": 0.4664,
+      "step": 21801
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.044525189606944e-06,
+      "loss": 0.4627,
+      "step": 21802
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.043948438446233e-06,
+      "loss": 0.4144,
+      "step": 21803
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.043371687285522e-06,
+      "loss": 0.5409,
+      "step": 21804
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.04279493612481e-06,
+      "loss": 0.4803,
+      "step": 21805
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.042218184964098e-06,
+      "loss": 0.4324,
+      "step": 21806
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.041641433803387e-06,
+      "loss": 0.3923,
+      "step": 21807
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.041064682642675e-06,
+      "loss": 0.3637,
+      "step": 21808
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.040487931481962e-06,
+      "loss": 0.3943,
+      "step": 21809
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.039911180321252e-06,
+      "loss": 0.4104,
+      "step": 21810
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.03933442916054e-06,
+      "loss": 0.533,
+      "step": 21811
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.038757677999827e-06,
+      "loss": 0.4146,
+      "step": 21812
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.038180926839117e-06,
+      "loss": 0.4449,
+      "step": 21813
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.037604175678404e-06,
+      "loss": 0.3747,
+      "step": 21814
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.037027424517692e-06,
+      "loss": 0.518,
+      "step": 21815
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.03645067335698e-06,
+      "loss": 0.3844,
+      "step": 21816
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.035873922196269e-06,
+      "loss": 0.3824,
+      "step": 21817
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.035297171035558e-06,
+      "loss": 0.4443,
+      "step": 21818
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.034720419874844e-06,
+      "loss": 0.4231,
+      "step": 21819
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.034143668714134e-06,
+      "loss": 0.5696,
+      "step": 21820
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.033566917553423e-06,
+      "loss": 0.5336,
+      "step": 21821
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.03299016639271e-06,
+      "loss": 0.4006,
+      "step": 21822
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.032413415231999e-06,
+      "loss": 0.5903,
+      "step": 21823
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.031836664071288e-06,
+      "loss": 0.431,
+      "step": 21824
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.031259912910576e-06,
+      "loss": 0.4874,
+      "step": 21825
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.030683161749863e-06,
+      "loss": 0.4937,
+      "step": 21826
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.030106410589153e-06,
+      "loss": 0.4135,
+      "step": 21827
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.02952965942844e-06,
+      "loss": 0.4274,
+      "step": 21828
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.028952908267728e-06,
+      "loss": 0.3719,
+      "step": 21829
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.028376157107017e-06,
+      "loss": 0.5334,
+      "step": 21830
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.027799405946305e-06,
+      "loss": 0.3279,
+      "step": 21831
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.027222654785593e-06,
+      "loss": 0.51,
+      "step": 21832
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.02664590362488e-06,
+      "loss": 0.4074,
+      "step": 21833
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.02606915246417e-06,
+      "loss": 0.4035,
+      "step": 21834
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.02549240130346e-06,
+      "loss": 0.465,
+      "step": 21835
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024915650142747e-06,
+      "loss": 0.3978,
+      "step": 21836
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.024338898982035e-06,
+      "loss": 0.5361,
+      "step": 21837
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.023762147821324e-06,
+      "loss": 0.4688,
+      "step": 21838
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.023185396660612e-06,
+      "loss": 0.5176,
+      "step": 21839
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.0226086454999e-06,
+      "loss": 0.3964,
+      "step": 21840
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.022031894339189e-06,
+      "loss": 0.5238,
+      "step": 21841
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.021455143178476e-06,
+      "loss": 0.4718,
+      "step": 21842
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.020878392017764e-06,
+      "loss": 0.4391,
+      "step": 21843
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.020301640857053e-06,
+      "loss": 0.4542,
+      "step": 21844
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.019724889696341e-06,
+      "loss": 0.4522,
+      "step": 21845
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.019148138535629e-06,
+      "loss": 0.4115,
+      "step": 21846
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.018571387374918e-06,
+      "loss": 0.4728,
+      "step": 21847
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.017994636214206e-06,
+      "loss": 0.3747,
+      "step": 21848
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.017417885053495e-06,
+      "loss": 0.4086,
+      "step": 21849
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.016841133892781e-06,
+      "loss": 0.3358,
+      "step": 21850
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.01626438273207e-06,
+      "loss": 0.4907,
+      "step": 21851
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.01568763157136e-06,
+      "loss": 0.5706,
+      "step": 21852
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.015110880410648e-06,
+      "loss": 0.4141,
+      "step": 21853
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.014534129249935e-06,
+      "loss": 0.361,
+      "step": 21854
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.013957378089225e-06,
+      "loss": 0.2995,
+      "step": 21855
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.013380626928512e-06,
+      "loss": 0.4235,
+      "step": 21856
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.0128038757678e-06,
+      "loss": 0.5214,
+      "step": 21857
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.01222712460709e-06,
+      "loss": 0.3878,
+      "step": 21858
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.011650373446377e-06,
+      "loss": 0.4723,
+      "step": 21859
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.011073622285665e-06,
+      "loss": 0.4461,
+      "step": 21860
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.010496871124954e-06,
+      "loss": 0.4699,
+      "step": 21861
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.009920119964242e-06,
+      "loss": 0.4308,
+      "step": 21862
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.00934336880353e-06,
+      "loss": 0.5655,
+      "step": 21863
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.008766617642819e-06,
+      "loss": 0.3233,
+      "step": 21864
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.008189866482107e-06,
+      "loss": 0.4442,
+      "step": 21865
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.007613115321396e-06,
+      "loss": 0.3203,
+      "step": 21866
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.007036364160684e-06,
+      "loss": 0.4037,
+      "step": 21867
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.006459612999971e-06,
+      "loss": 0.3445,
+      "step": 21868
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.00588286183926e-06,
+      "loss": 0.4723,
+      "step": 21869
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.005306110678549e-06,
+      "loss": 0.3765,
+      "step": 21870
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004729359517836e-06,
+      "loss": 0.4604,
+      "step": 21871
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.004152608357126e-06,
+      "loss": 0.4717,
+      "step": 21872
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.003575857196413e-06,
+      "loss": 0.4096,
+      "step": 21873
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.002999106035701e-06,
+      "loss": 0.4647,
+      "step": 21874
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.00242235487499e-06,
+      "loss": 0.37,
+      "step": 21875
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.001845603714278e-06,
+      "loss": 0.558,
+      "step": 21876
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.001268852553566e-06,
+      "loss": 0.4125,
+      "step": 21877
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.000692101392855e-06,
+      "loss": 0.4383,
+      "step": 21878
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 8.000115350232143e-06,
+      "loss": 0.3539,
+      "step": 21879
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.999538599071432e-06,
+      "loss": 0.4335,
+      "step": 21880
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.99896184791072e-06,
+      "loss": 0.3639,
+      "step": 21881
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.998385096750008e-06,
+      "loss": 0.4988,
+      "step": 21882
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.997808345589297e-06,
+      "loss": 0.4169,
+      "step": 21883
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.997231594428585e-06,
+      "loss": 0.4812,
+      "step": 21884
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.996654843267872e-06,
+      "loss": 0.4524,
+      "step": 21885
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.996078092107162e-06,
+      "loss": 0.3245,
+      "step": 21886
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.99550134094645e-06,
+      "loss": 0.4987,
+      "step": 21887
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.994924589785737e-06,
+      "loss": 0.5229,
+      "step": 21888
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.994347838625026e-06,
+      "loss": 0.474,
+      "step": 21889
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.993771087464314e-06,
+      "loss": 0.4056,
+      "step": 21890
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.993194336303602e-06,
+      "loss": 0.452,
+      "step": 21891
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.992617585142891e-06,
+      "loss": 0.3086,
+      "step": 21892
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.992040833982179e-06,
+      "loss": 0.4006,
+      "step": 21893
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.991464082821467e-06,
+      "loss": 0.4473,
+      "step": 21894
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.990887331660756e-06,
+      "loss": 0.4553,
+      "step": 21895
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 7.990310580500044e-06,
+      "loss": 0.4837,
+      "step": 21896
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.989733829339333e-06,
+      "loss": 0.4366,
+      "step": 21897
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.98915707817862e-06,
+      "loss": 0.4644,
+      "step": 21898
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.988580327017908e-06,
+      "loss": 0.3495,
+      "step": 21899
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.988003575857198e-06,
+      "loss": 0.4043,
+      "step": 21900
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.987426824696485e-06,
+      "loss": 0.5019,
+      "step": 21901
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.986850073535773e-06,
+      "loss": 0.4713,
+      "step": 21902
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.986273322375062e-06,
+      "loss": 0.4451,
+      "step": 21903
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.98569657121435e-06,
+      "loss": 0.3473,
+      "step": 21904
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.985119820053638e-06,
+      "loss": 0.3642,
+      "step": 21905
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.984543068892927e-06,
+      "loss": 0.4858,
+      "step": 21906
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983966317732215e-06,
+      "loss": 0.4282,
+      "step": 21907
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.983389566571503e-06,
+      "loss": 0.4205,
+      "step": 21908
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.982812815410792e-06,
+      "loss": 0.538,
+      "step": 21909
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.98223606425008e-06,
+      "loss": 0.457,
+      "step": 21910
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.981659313089369e-06,
+      "loss": 0.4571,
+      "step": 21911
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.981082561928657e-06,
+      "loss": 0.4387,
+      "step": 21912
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.980505810767944e-06,
+      "loss": 0.5084,
+      "step": 21913
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.979929059607234e-06,
+      "loss": 0.527,
+      "step": 21914
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.979352308446521e-06,
+      "loss": 0.4552,
+      "step": 21915
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.978775557285809e-06,
+      "loss": 0.4371,
+      "step": 21916
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.978198806125099e-06,
+      "loss": 0.3948,
+      "step": 21917
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.977622054964386e-06,
+      "loss": 0.5645,
+      "step": 21918
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.977045303803674e-06,
+      "loss": 0.4627,
+      "step": 21919
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.976468552642963e-06,
+      "loss": 0.2817,
+      "step": 21920
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.975891801482251e-06,
+      "loss": 0.4301,
+      "step": 21921
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.975315050321539e-06,
+      "loss": 0.3043,
+      "step": 21922
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.974738299160828e-06,
+      "loss": 0.4655,
+      "step": 21923
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.974161548000116e-06,
+      "loss": 0.3815,
+      "step": 21924
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.973584796839403e-06,
+      "loss": 0.3981,
+      "step": 21925
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.973008045678693e-06,
+      "loss": 0.4205,
+      "step": 21926
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.97243129451798e-06,
+      "loss": 0.4023,
+      "step": 21927
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.97185454335727e-06,
+      "loss": 0.4839,
+      "step": 21928
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.971277792196558e-06,
+      "loss": 0.3922,
+      "step": 21929
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.970701041035845e-06,
+      "loss": 0.4905,
+      "step": 21930
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.970124289875135e-06,
+      "loss": 0.4112,
+      "step": 21931
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.969547538714422e-06,
+      "loss": 0.3509,
+      "step": 21932
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.96897078755371e-06,
+      "loss": 0.4488,
+      "step": 21933
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.968394036393e-06,
+      "loss": 0.4578,
+      "step": 21934
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.967817285232287e-06,
+      "loss": 0.3645,
+      "step": 21935
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.967240534071575e-06,
+      "loss": 0.3766,
+      "step": 21936
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.966663782910864e-06,
+      "loss": 0.439,
+      "step": 21937
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.966087031750152e-06,
+      "loss": 0.4225,
+      "step": 21938
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.96551028058944e-06,
+      "loss": 0.3832,
+      "step": 21939
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.964933529428729e-06,
+      "loss": 0.4223,
+      "step": 21940
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.964356778268017e-06,
+      "loss": 0.4583,
+      "step": 21941
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.963780027107306e-06,
+      "loss": 0.4342,
+      "step": 21942
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.963203275946594e-06,
+      "loss": 0.406,
+      "step": 21943
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.962626524785881e-06,
+      "loss": 0.3339,
+      "step": 21944
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.96204977362517e-06,
+      "loss": 0.4185,
+      "step": 21945
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.961473022464458e-06,
+      "loss": 0.4143,
+      "step": 21946
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.960896271303746e-06,
+      "loss": 0.5338,
+      "step": 21947
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.960319520143035e-06,
+      "loss": 0.4471,
+      "step": 21948
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.959742768982323e-06,
+      "loss": 0.3889,
+      "step": 21949
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.95916601782161e-06,
+      "loss": 0.4116,
+      "step": 21950
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.9585892666609e-06,
+      "loss": 0.4339,
+      "step": 21951
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.958012515500188e-06,
+      "loss": 0.486,
+      "step": 21952
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.957435764339476e-06,
+      "loss": 0.477,
+      "step": 21953
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.956859013178765e-06,
+      "loss": 0.433,
+      "step": 21954
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.956282262018053e-06,
+      "loss": 0.4372,
+      "step": 21955
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.95570551085734e-06,
+      "loss": 0.5544,
+      "step": 21956
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.95512875969663e-06,
+      "loss": 0.5069,
+      "step": 21957
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.954552008535917e-06,
+      "loss": 0.4434,
+      "step": 21958
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.953975257375207e-06,
+      "loss": 0.3888,
+      "step": 21959
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.953398506214494e-06,
+      "loss": 0.5175,
+      "step": 21960
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.952821755053782e-06,
+      "loss": 0.5419,
+      "step": 21961
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.952245003893071e-06,
+      "loss": 0.5095,
+      "step": 21962
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.951668252732359e-06,
+      "loss": 0.4285,
+      "step": 21963
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.951091501571647e-06,
+      "loss": 0.4773,
+      "step": 21964
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.950514750410936e-06,
+      "loss": 0.4252,
+      "step": 21965
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.949937999250224e-06,
+      "loss": 0.4359,
+      "step": 21966
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.949361248089512e-06,
+      "loss": 0.3593,
+      "step": 21967
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.948784496928801e-06,
+      "loss": 0.5446,
+      "step": 21968
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.94820774576809e-06,
+      "loss": 0.502,
+      "step": 21969
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.947630994607376e-06,
+      "loss": 0.4663,
+      "step": 21970
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.947054243446666e-06,
+      "loss": 0.4605,
+      "step": 21971
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.946477492285953e-06,
+      "loss": 0.4645,
+      "step": 21972
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.945900741125243e-06,
+      "loss": 0.4404,
+      "step": 21973
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.94532398996453e-06,
+      "loss": 0.3807,
+      "step": 21974
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.944747238803818e-06,
+      "loss": 0.4937,
+      "step": 21975
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.944170487643108e-06,
+      "loss": 0.3892,
+      "step": 21976
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.943593736482395e-06,
+      "loss": 0.5315,
+      "step": 21977
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.943016985321683e-06,
+      "loss": 0.3926,
+      "step": 21978
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.942440234160972e-06,
+      "loss": 0.3575,
+      "step": 21979
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.94186348300026e-06,
+      "loss": 0.5104,
+      "step": 21980
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.941286731839548e-06,
+      "loss": 0.5034,
+      "step": 21981
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.940709980678837e-06,
+      "loss": 0.3663,
+      "step": 21982
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.940133229518125e-06,
+      "loss": 0.3423,
+      "step": 21983
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.939556478357412e-06,
+      "loss": 0.4133,
+      "step": 21984
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.938979727196702e-06,
+      "loss": 0.4227,
+      "step": 21985
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.938402976035991e-06,
+      "loss": 0.3933,
+      "step": 21986
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.937826224875277e-06,
+      "loss": 0.2842,
+      "step": 21987
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.937249473714567e-06,
+      "loss": 0.4078,
+      "step": 21988
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.936672722553854e-06,
+      "loss": 0.482,
+      "step": 21989
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.936095971393144e-06,
+      "loss": 0.5022,
+      "step": 21990
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.935519220232431e-06,
+      "loss": 0.2536,
+      "step": 21991
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.934942469071719e-06,
+      "loss": 0.4882,
+      "step": 21992
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.934365717911008e-06,
+      "loss": 0.3653,
+      "step": 21993
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.933788966750296e-06,
+      "loss": 0.4589,
+      "step": 21994
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.933212215589584e-06,
+      "loss": 0.3438,
+      "step": 21995
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.932635464428873e-06,
+      "loss": 0.4364,
+      "step": 21996
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.93205871326816e-06,
+      "loss": 0.4203,
+      "step": 21997
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.931481962107448e-06,
+      "loss": 0.5181,
+      "step": 21998
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.930905210946738e-06,
+      "loss": 0.5193,
+      "step": 21999
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.930328459786027e-06,
+      "loss": 0.4763,
+      "step": 22000
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.929751708625313e-06,
+      "loss": 0.3309,
+      "step": 22001
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.929174957464603e-06,
+      "loss": 0.3773,
+      "step": 22002
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.928598206303892e-06,
+      "loss": 0.53,
+      "step": 22003
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.92802145514318e-06,
+      "loss": 0.52,
+      "step": 22004
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.927444703982467e-06,
+      "loss": 0.4272,
+      "step": 22005
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.926867952821757e-06,
+      "loss": 0.4397,
+      "step": 22006
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.926291201661044e-06,
+      "loss": 0.4999,
+      "step": 22007
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.925714450500332e-06,
+      "loss": 0.3524,
+      "step": 22008
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.92513769933962e-06,
+      "loss": 0.4449,
+      "step": 22009
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.924560948178909e-06,
+      "loss": 0.4273,
+      "step": 22010
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.923984197018197e-06,
+      "loss": 0.4209,
+      "step": 22011
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.923407445857485e-06,
+      "loss": 0.548,
+      "step": 22012
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.922830694696774e-06,
+      "loss": 0.37,
+      "step": 22013
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.922253943536062e-06,
+      "loss": 0.4312,
+      "step": 22014
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.92167719237535e-06,
+      "loss": 0.4496,
+      "step": 22015
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.921100441214639e-06,
+      "loss": 0.3389,
+      "step": 22016
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.920523690053928e-06,
+      "loss": 0.4402,
+      "step": 22017
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.919946938893216e-06,
+      "loss": 0.3822,
+      "step": 22018
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.919370187732503e-06,
+      "loss": 0.418,
+      "step": 22019
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.918793436571793e-06,
+      "loss": 0.391,
+      "step": 22020
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.91821668541108e-06,
+      "loss": 0.4242,
+      "step": 22021
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.917639934250368e-06,
+      "loss": 0.3723,
+      "step": 22022
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.917063183089658e-06,
+      "loss": 0.5359,
+      "step": 22023
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.916486431928945e-06,
+      "loss": 0.3961,
+      "step": 22024
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.915909680768233e-06,
+      "loss": 0.4138,
+      "step": 22025
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.91533292960752e-06,
+      "loss": 0.3456,
+      "step": 22026
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.91475617844681e-06,
+      "loss": 0.5035,
+      "step": 22027
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.914179427286098e-06,
+      "loss": 0.3802,
+      "step": 22028
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.913602676125385e-06,
+      "loss": 0.3699,
+      "step": 22029
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.913025924964675e-06,
+      "loss": 0.5344,
+      "step": 22030
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.912449173803964e-06,
+      "loss": 0.4779,
+      "step": 22031
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.91187242264325e-06,
+      "loss": 0.3901,
+      "step": 22032
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.91129567148254e-06,
+      "loss": 0.3916,
+      "step": 22033
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.910718920321829e-06,
+      "loss": 0.4484,
+      "step": 22034
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.910142169161117e-06,
+      "loss": 0.4597,
+      "step": 22035
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.909565418000404e-06,
+      "loss": 0.4742,
+      "step": 22036
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.908988666839694e-06,
+      "loss": 0.456,
+      "step": 22037
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.908411915678981e-06,
+      "loss": 0.3555,
+      "step": 22038
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.907835164518269e-06,
+      "loss": 0.4827,
+      "step": 22039
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.907258413357558e-06,
+      "loss": 0.4871,
+      "step": 22040
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.906681662196846e-06,
+      "loss": 0.2898,
+      "step": 22041
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.906104911036134e-06,
+      "loss": 0.3843,
+      "step": 22042
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.905528159875421e-06,
+      "loss": 0.3867,
+      "step": 22043
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.90495140871471e-06,
+      "loss": 0.4554,
+      "step": 22044
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.904374657553998e-06,
+      "loss": 0.4604,
+      "step": 22045
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.903797906393286e-06,
+      "loss": 0.3559,
+      "step": 22046
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.903221155232576e-06,
+      "loss": 0.5962,
+      "step": 22047
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.902644404071865e-06,
+      "loss": 0.4291,
+      "step": 22048
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.902067652911153e-06,
+      "loss": 0.5765,
+      "step": 22049
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.90149090175044e-06,
+      "loss": 0.4019,
+      "step": 22050
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.90091415058973e-06,
+      "loss": 0.4264,
+      "step": 22051
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.900337399429017e-06,
+      "loss": 0.4377,
+      "step": 22052
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.899760648268305e-06,
+      "loss": 0.4861,
+      "step": 22053
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.899183897107594e-06,
+      "loss": 0.4949,
+      "step": 22054
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.898607145946882e-06,
+      "loss": 0.4546,
+      "step": 22055
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.89803039478617e-06,
+      "loss": 0.4727,
+      "step": 22056
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.897453643625459e-06,
+      "loss": 0.3891,
+      "step": 22057
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.896876892464747e-06,
+      "loss": 0.5137,
+      "step": 22058
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.896300141304035e-06,
+      "loss": 0.2997,
+      "step": 22059
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.895723390143322e-06,
+      "loss": 0.3853,
+      "step": 22060
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.895146638982612e-06,
+      "loss": 0.4125,
+      "step": 22061
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.894569887821901e-06,
+      "loss": 0.3445,
+      "step": 22062
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.893993136661187e-06,
+      "loss": 0.4559,
+      "step": 22063
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.893416385500476e-06,
+      "loss": 0.4286,
+      "step": 22064
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.892839634339766e-06,
+      "loss": 0.3355,
+      "step": 22065
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.892262883179053e-06,
+      "loss": 0.4021,
+      "step": 22066
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.891686132018341e-06,
+      "loss": 0.3299,
+      "step": 22067
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.89110938085763e-06,
+      "loss": 0.463,
+      "step": 22068
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.890532629696918e-06,
+      "loss": 0.4532,
+      "step": 22069
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.889955878536206e-06,
+      "loss": 0.4264,
+      "step": 22070
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.889379127375495e-06,
+      "loss": 0.451,
+      "step": 22071
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.888802376214783e-06,
+      "loss": 0.5081,
+      "step": 22072
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.88822562505407e-06,
+      "loss": 0.3819,
+      "step": 22073
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.88764887389336e-06,
+      "loss": 0.4109,
+      "step": 22074
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 7.887072122732648e-06,
+      "loss": 0.4727,
+      "step": 22075
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.886495371571935e-06,
+      "loss": 0.4525,
+      "step": 22076
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.885918620411223e-06,
+      "loss": 0.5019,
+      "step": 22077
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.885341869250512e-06,
+      "loss": 0.3714,
+      "step": 22078
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.884765118089802e-06,
+      "loss": 0.4151,
+      "step": 22079
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.88418836692909e-06,
+      "loss": 0.4315,
+      "step": 22080
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.883611615768377e-06,
+      "loss": 0.3392,
+      "step": 22081
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.883034864607667e-06,
+      "loss": 0.4182,
+      "step": 22082
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.882458113446954e-06,
+      "loss": 0.4748,
+      "step": 22083
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.881881362286242e-06,
+      "loss": 0.3881,
+      "step": 22084
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.881304611125531e-06,
+      "loss": 0.3965,
+      "step": 22085
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880727859964819e-06,
+      "loss": 0.6116,
+      "step": 22086
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.880151108804107e-06,
+      "loss": 0.6204,
+      "step": 22087
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.879574357643396e-06,
+      "loss": 0.3029,
+      "step": 22088
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.878997606482684e-06,
+      "loss": 0.5279,
+      "step": 22089
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.878420855321971e-06,
+      "loss": 0.4127,
+      "step": 22090
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.87784410416126e-06,
+      "loss": 0.4215,
+      "step": 22091
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.877267353000548e-06,
+      "loss": 0.3586,
+      "step": 22092
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.876690601839838e-06,
+      "loss": 0.3591,
+      "step": 22093
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.876113850679124e-06,
+      "loss": 0.3505,
+      "step": 22094
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.875537099518413e-06,
+      "loss": 0.4688,
+      "step": 22095
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.874960348357703e-06,
+      "loss": 0.4782,
+      "step": 22096
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.87438359719699e-06,
+      "loss": 0.3219,
+      "step": 22097
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.873806846036278e-06,
+      "loss": 0.3966,
+      "step": 22098
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.873230094875567e-06,
+      "loss": 0.4208,
+      "step": 22099
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.872653343714855e-06,
+      "loss": 0.4484,
+      "step": 22100
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.872076592554143e-06,
+      "loss": 0.4584,
+      "step": 22101
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.871499841393432e-06,
+      "loss": 0.5262,
+      "step": 22102
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.87092309023272e-06,
+      "loss": 0.4106,
+      "step": 22103
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.870346339072007e-06,
+      "loss": 0.4606,
+      "step": 22104
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.869769587911297e-06,
+      "loss": 0.4176,
+      "step": 22105
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.869192836750585e-06,
+      "loss": 0.4755,
+      "step": 22106
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.868616085589872e-06,
+      "loss": 0.4489,
+      "step": 22107
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.868039334429162e-06,
+      "loss": 0.4648,
+      "step": 22108
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.86746258326845e-06,
+      "loss": 0.4002,
+      "step": 22109
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.866885832107739e-06,
+      "loss": 0.3869,
+      "step": 22110
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.866309080947026e-06,
+      "loss": 0.4194,
+      "step": 22111
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.865732329786314e-06,
+      "loss": 0.5447,
+      "step": 22112
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.865155578625603e-06,
+      "loss": 0.4012,
+      "step": 22113
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.864578827464891e-06,
+      "loss": 0.4631,
+      "step": 22114
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.864002076304179e-06,
+      "loss": 0.3498,
+      "step": 22115
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.863425325143468e-06,
+      "loss": 0.5029,
+      "step": 22116
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.862848573982756e-06,
+      "loss": 0.6253,
+      "step": 22117
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.862271822822044e-06,
+      "loss": 0.2405,
+      "step": 22118
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.861695071661333e-06,
+      "loss": 0.4865,
+      "step": 22119
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.86111832050062e-06,
+      "loss": 0.4491,
+      "step": 22120
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.860541569339908e-06,
+      "loss": 0.3798,
+      "step": 22121
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.859964818179198e-06,
+      "loss": 0.3477,
+      "step": 22122
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.859388067018485e-06,
+      "loss": 0.5098,
+      "step": 22123
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.858811315857775e-06,
+      "loss": 0.4986,
+      "step": 22124
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.858234564697062e-06,
+      "loss": 0.4495,
+      "step": 22125
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.85765781353635e-06,
+      "loss": 0.309,
+      "step": 22126
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.85708106237564e-06,
+      "loss": 0.5831,
+      "step": 22127
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.856504311214927e-06,
+      "loss": 0.4395,
+      "step": 22128
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.855927560054215e-06,
+      "loss": 0.3276,
+      "step": 22129
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.855350808893504e-06,
+      "loss": 0.4068,
+      "step": 22130
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.854774057732792e-06,
+      "loss": 0.497,
+      "step": 22131
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.85419730657208e-06,
+      "loss": 0.5021,
+      "step": 22132
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.853620555411369e-06,
+      "loss": 0.3717,
+      "step": 22133
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.853043804250657e-06,
+      "loss": 0.52,
+      "step": 22134
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.852467053089944e-06,
+      "loss": 0.4124,
+      "step": 22135
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.851890301929234e-06,
+      "loss": 0.4665,
+      "step": 22136
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.851313550768521e-06,
+      "loss": 0.3259,
+      "step": 22137
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.850736799607809e-06,
+      "loss": 0.4547,
+      "step": 22138
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.850160048447098e-06,
+      "loss": 0.4395,
+      "step": 22139
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.849583297286386e-06,
+      "loss": 0.3664,
+      "step": 22140
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.849006546125676e-06,
+      "loss": 0.5031,
+      "step": 22141
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.848429794964963e-06,
+      "loss": 0.5928,
+      "step": 22142
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.847853043804251e-06,
+      "loss": 0.4208,
+      "step": 22143
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.84727629264354e-06,
+      "loss": 0.4913,
+      "step": 22144
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.846699541482828e-06,
+      "loss": 0.4068,
+      "step": 22145
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.846122790322116e-06,
+      "loss": 0.406,
+      "step": 22146
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.845546039161405e-06,
+      "loss": 0.5123,
+      "step": 22147
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.844969288000693e-06,
+      "loss": 0.5095,
+      "step": 22148
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.84439253683998e-06,
+      "loss": 0.4212,
+      "step": 22149
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.84381578567927e-06,
+      "loss": 0.501,
+      "step": 22150
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.843239034518557e-06,
+      "loss": 0.4666,
+      "step": 22151
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.842662283357845e-06,
+      "loss": 0.3656,
+      "step": 22152
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.842085532197135e-06,
+      "loss": 0.3024,
+      "step": 22153
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.841508781036422e-06,
+      "loss": 0.3998,
+      "step": 22154
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.840932029875712e-06,
+      "loss": 0.505,
+      "step": 22155
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.840355278715e-06,
+      "loss": 0.3863,
+      "step": 22156
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839778527554287e-06,
+      "loss": 0.4206,
+      "step": 22157
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.839201776393576e-06,
+      "loss": 0.3694,
+      "step": 22158
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.838625025232864e-06,
+      "loss": 0.4731,
+      "step": 22159
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.838048274072152e-06,
+      "loss": 0.5523,
+      "step": 22160
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.837471522911441e-06,
+      "loss": 0.3843,
+      "step": 22161
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.836894771750729e-06,
+      "loss": 0.4416,
+      "step": 22162
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.836318020590016e-06,
+      "loss": 0.5426,
+      "step": 22163
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.835741269429306e-06,
+      "loss": 0.4359,
+      "step": 22164
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.835164518268594e-06,
+      "loss": 0.3375,
+      "step": 22165
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.834587767107881e-06,
+      "loss": 0.5576,
+      "step": 22166
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.83401101594717e-06,
+      "loss": 0.3554,
+      "step": 22167
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.833434264786458e-06,
+      "loss": 0.3886,
+      "step": 22168
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.832857513625746e-06,
+      "loss": 0.5112,
+      "step": 22169
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.832280762465035e-06,
+      "loss": 0.3842,
+      "step": 22170
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.831704011304323e-06,
+      "loss": 0.5257,
+      "step": 22171
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.831127260143612e-06,
+      "loss": 0.369,
+      "step": 22172
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.8305505089829e-06,
+      "loss": 0.6023,
+      "step": 22173
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.829973757822188e-06,
+      "loss": 0.4417,
+      "step": 22174
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.829397006661477e-06,
+      "loss": 0.4287,
+      "step": 22175
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.828820255500765e-06,
+      "loss": 0.4098,
+      "step": 22176
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.828243504340053e-06,
+      "loss": 0.425,
+      "step": 22177
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.827666753179342e-06,
+      "loss": 0.4745,
+      "step": 22178
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.82709000201863e-06,
+      "loss": 0.3971,
+      "step": 22179
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.826513250857917e-06,
+      "loss": 0.3224,
+      "step": 22180
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.825936499697207e-06,
+      "loss": 0.4436,
+      "step": 22181
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.825359748536494e-06,
+      "loss": 0.4988,
+      "step": 22182
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.824782997375782e-06,
+      "loss": 0.3726,
+      "step": 22183
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.824206246215071e-06,
+      "loss": 0.634,
+      "step": 22184
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.823629495054359e-06,
+      "loss": 0.4689,
+      "step": 22185
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.823052743893648e-06,
+      "loss": 0.4019,
+      "step": 22186
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.822475992732936e-06,
+      "loss": 0.2727,
+      "step": 22187
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.821899241572224e-06,
+      "loss": 0.3497,
+      "step": 22188
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.821322490411513e-06,
+      "loss": 0.5299,
+      "step": 22189
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.820745739250801e-06,
+      "loss": 0.4218,
+      "step": 22190
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.820168988090089e-06,
+      "loss": 0.4337,
+      "step": 22191
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.819592236929378e-06,
+      "loss": 0.4143,
+      "step": 22192
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.819015485768666e-06,
+      "loss": 0.3996,
+      "step": 22193
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.818438734607953e-06,
+      "loss": 0.4896,
+      "step": 22194
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.817861983447243e-06,
+      "loss": 0.3536,
+      "step": 22195
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.81728523228653e-06,
+      "loss": 0.3642,
+      "step": 22196
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.816708481125818e-06,
+      "loss": 0.4152,
+      "step": 22197
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.816131729965107e-06,
+      "loss": 0.3694,
+      "step": 22198
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.815554978804395e-06,
+      "loss": 0.4636,
+      "step": 22199
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.814978227643683e-06,
+      "loss": 0.3276,
+      "step": 22200
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.814401476482972e-06,
+      "loss": 0.4282,
+      "step": 22201
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.81382472532226e-06,
+      "loss": 0.5012,
+      "step": 22202
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.81324797416155e-06,
+      "loss": 0.4357,
+      "step": 22203
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.812671223000837e-06,
+      "loss": 0.4558,
+      "step": 22204
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.812094471840125e-06,
+      "loss": 0.3766,
+      "step": 22205
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.811517720679414e-06,
+      "loss": 0.4281,
+      "step": 22206
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.810940969518702e-06,
+      "loss": 0.5551,
+      "step": 22207
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.81036421835799e-06,
+      "loss": 0.4278,
+      "step": 22208
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.809787467197279e-06,
+      "loss": 0.3117,
+      "step": 22209
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.809210716036566e-06,
+      "loss": 0.4345,
+      "step": 22210
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.808633964875854e-06,
+      "loss": 0.3795,
+      "step": 22211
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.808057213715144e-06,
+      "loss": 0.4145,
+      "step": 22212
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.807480462554431e-06,
+      "loss": 0.2894,
+      "step": 22213
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.806903711393719e-06,
+      "loss": 0.4972,
+      "step": 22214
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.806326960233008e-06,
+      "loss": 0.3551,
+      "step": 22215
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.805750209072296e-06,
+      "loss": 0.4459,
+      "step": 22216
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.805173457911585e-06,
+      "loss": 0.4085,
+      "step": 22217
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.804596706750873e-06,
+      "loss": 0.4508,
+      "step": 22218
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.80401995559016e-06,
+      "loss": 0.3652,
+      "step": 22219
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.80344320442945e-06,
+      "loss": 0.4261,
+      "step": 22220
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.802866453268738e-06,
+      "loss": 0.4731,
+      "step": 22221
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.802289702108025e-06,
+      "loss": 0.4298,
+      "step": 22222
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.801712950947315e-06,
+      "loss": 0.4443,
+      "step": 22223
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.801136199786603e-06,
+      "loss": 0.5195,
+      "step": 22224
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.80055944862589e-06,
+      "loss": 0.4719,
+      "step": 22225
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79998269746518e-06,
+      "loss": 0.4625,
+      "step": 22226
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.799405946304467e-06,
+      "loss": 0.3108,
+      "step": 22227
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.798829195143755e-06,
+      "loss": 0.3547,
+      "step": 22228
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.798252443983044e-06,
+      "loss": 0.542,
+      "step": 22229
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.797675692822334e-06,
+      "loss": 0.5415,
+      "step": 22230
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79709894166162e-06,
+      "loss": 0.4731,
+      "step": 22231
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.796522190500909e-06,
+      "loss": 0.426,
+      "step": 22232
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.795945439340197e-06,
+      "loss": 0.4014,
+      "step": 22233
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.795368688179486e-06,
+      "loss": 0.3714,
+      "step": 22234
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.794791937018774e-06,
+      "loss": 0.4607,
+      "step": 22235
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.794215185858062e-06,
+      "loss": 0.4606,
+      "step": 22236
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.793638434697351e-06,
+      "loss": 0.3683,
+      "step": 22237
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.793061683536639e-06,
+      "loss": 0.3193,
+      "step": 22238
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.792484932375926e-06,
+      "loss": 0.362,
+      "step": 22239
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.791908181215216e-06,
+      "loss": 0.4141,
+      "step": 22240
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.791331430054503e-06,
+      "loss": 0.5243,
+      "step": 22241
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.790754678893791e-06,
+      "loss": 0.3936,
+      "step": 22242
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.79017792773308e-06,
+      "loss": 0.5197,
+      "step": 22243
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.789601176572368e-06,
+      "loss": 0.3967,
+      "step": 22244
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.789024425411656e-06,
+      "loss": 0.422,
+      "step": 22245
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.788447674250945e-06,
+      "loss": 0.4443,
+      "step": 22246
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.787870923090235e-06,
+      "loss": 0.4973,
+      "step": 22247
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.787294171929522e-06,
+      "loss": 0.4431,
+      "step": 22248
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.78671742076881e-06,
+      "loss": 0.3462,
+      "step": 22249
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.786140669608098e-06,
+      "loss": 0.5729,
+      "step": 22250
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.785563918447387e-06,
+      "loss": 0.5147,
+      "step": 22251
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.784987167286675e-06,
+      "loss": 0.5539,
+      "step": 22252
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.784410416125962e-06,
+      "loss": 0.4529,
+      "step": 22253
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 7.783833664965252e-06,
+      "loss": 0.5646,
+      "step": 22254
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.78325691380454e-06,
+      "loss": 0.4003,
+      "step": 22255
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.782680162643827e-06,
+      "loss": 0.3783,
+      "step": 22256
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.782103411483116e-06,
+      "loss": 0.6729,
+      "step": 22257
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.781526660322404e-06,
+      "loss": 0.3922,
+      "step": 22258
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.780949909161692e-06,
+      "loss": 0.5688,
+      "step": 22259
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.780373158000981e-06,
+      "loss": 0.4595,
+      "step": 22260
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.77979640684027e-06,
+      "loss": 0.4414,
+      "step": 22261
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.779219655679557e-06,
+      "loss": 0.4376,
+      "step": 22262
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.778642904518846e-06,
+      "loss": 0.4025,
+      "step": 22263
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.778066153358135e-06,
+      "loss": 0.4507,
+      "step": 22264
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.777489402197423e-06,
+      "loss": 0.3996,
+      "step": 22265
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.77691265103671e-06,
+      "loss": 0.489,
+      "step": 22266
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.776335899876e-06,
+      "loss": 0.5272,
+      "step": 22267
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.775759148715288e-06,
+      "loss": 0.3611,
+      "step": 22268
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.775182397554575e-06,
+      "loss": 0.4766,
+      "step": 22269
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.774605646393863e-06,
+      "loss": 0.4227,
+      "step": 22270
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.774028895233153e-06,
+      "loss": 0.474,
+      "step": 22271
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.77345214407244e-06,
+      "loss": 0.4606,
+      "step": 22272
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.772875392911728e-06,
+      "loss": 0.4312,
+      "step": 22273
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.772298641751017e-06,
+      "loss": 0.4836,
+      "step": 22274
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.771721890590307e-06,
+      "loss": 0.3755,
+      "step": 22275
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.771145139429593e-06,
+      "loss": 0.4553,
+      "step": 22276
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.770568388268882e-06,
+      "loss": 0.5343,
+      "step": 22277
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.769991637108171e-06,
+      "loss": 0.3799,
+      "step": 22278
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.769414885947459e-06,
+      "loss": 0.481,
+      "step": 22279
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.768838134786747e-06,
+      "loss": 0.3949,
+      "step": 22280
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.768261383626036e-06,
+      "loss": 0.5056,
+      "step": 22281
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.767684632465324e-06,
+      "loss": 0.3441,
+      "step": 22282
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.767107881304612e-06,
+      "loss": 0.4974,
+      "step": 22283
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.766531130143901e-06,
+      "loss": 0.5221,
+      "step": 22284
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.765954378983189e-06,
+      "loss": 0.4121,
+      "step": 22285
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.765377627822476e-06,
+      "loss": 0.4239,
+      "step": 22286
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.764800876661764e-06,
+      "loss": 0.4942,
+      "step": 22287
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.764224125501053e-06,
+      "loss": 0.4844,
+      "step": 22288
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.763647374340341e-06,
+      "loss": 0.428,
+      "step": 22289
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.763070623179629e-06,
+      "loss": 0.3424,
+      "step": 22290
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.762493872018918e-06,
+      "loss": 0.4419,
+      "step": 22291
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.761917120858207e-06,
+      "loss": 0.4504,
+      "step": 22292
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.761340369697493e-06,
+      "loss": 0.4448,
+      "step": 22293
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.760763618536783e-06,
+      "loss": 0.5804,
+      "step": 22294
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.760186867376072e-06,
+      "loss": 0.4216,
+      "step": 22295
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.75961011621536e-06,
+      "loss": 0.5285,
+      "step": 22296
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.759033365054648e-06,
+      "loss": 0.4177,
+      "step": 22297
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.758456613893937e-06,
+      "loss": 0.4667,
+      "step": 22298
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757879862733225e-06,
+      "loss": 0.3965,
+      "step": 22299
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.757303111572512e-06,
+      "loss": 0.43,
+      "step": 22300
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.756726360411802e-06,
+      "loss": 0.3602,
+      "step": 22301
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.75614960925109e-06,
+      "loss": 0.5882,
+      "step": 22302
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.755572858090377e-06,
+      "loss": 0.4863,
+      "step": 22303
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.754996106929665e-06,
+      "loss": 0.3888,
+      "step": 22304
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.754419355768954e-06,
+      "loss": 0.4365,
+      "step": 22305
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.753842604608244e-06,
+      "loss": 0.4126,
+      "step": 22306
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.75326585344753e-06,
+      "loss": 0.4722,
+      "step": 22307
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.752689102286819e-06,
+      "loss": 0.441,
+      "step": 22308
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.752112351126108e-06,
+      "loss": 0.4601,
+      "step": 22309
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.751535599965396e-06,
+      "loss": 0.4554,
+      "step": 22310
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.750958848804684e-06,
+      "loss": 0.4982,
+      "step": 22311
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.750382097643973e-06,
+      "loss": 0.3705,
+      "step": 22312
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.74980534648326e-06,
+      "loss": 0.3827,
+      "step": 22313
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.749228595322548e-06,
+      "loss": 0.4486,
+      "step": 22314
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.748651844161838e-06,
+      "loss": 0.4481,
+      "step": 22315
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.748075093001125e-06,
+      "loss": 0.4909,
+      "step": 22316
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.747498341840413e-06,
+      "loss": 0.3264,
+      "step": 22317
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.746921590679703e-06,
+      "loss": 0.4693,
+      "step": 22318
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.74634483951899e-06,
+      "loss": 0.5075,
+      "step": 22319
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.745768088358278e-06,
+      "loss": 0.4738,
+      "step": 22320
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.745191337197566e-06,
+      "loss": 0.5793,
+      "step": 22321
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.744614586036855e-06,
+      "loss": 0.3532,
+      "step": 22322
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.744037834876144e-06,
+      "loss": 0.3312,
+      "step": 22323
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.74346108371543e-06,
+      "loss": 0.3796,
+      "step": 22324
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.74288433255472e-06,
+      "loss": 0.3429,
+      "step": 22325
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.742307581394009e-06,
+      "loss": 0.4228,
+      "step": 22326
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.741730830233297e-06,
+      "loss": 0.5594,
+      "step": 22327
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.741154079072584e-06,
+      "loss": 0.5008,
+      "step": 22328
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.740577327911874e-06,
+      "loss": 0.4491,
+      "step": 22329
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.740000576751162e-06,
+      "loss": 0.4231,
+      "step": 22330
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.73942382559045e-06,
+      "loss": 0.4902,
+      "step": 22331
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.738847074429739e-06,
+      "loss": 0.3893,
+      "step": 22332
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.738270323269026e-06,
+      "loss": 0.2956,
+      "step": 22333
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.737693572108314e-06,
+      "loss": 0.4244,
+      "step": 22334
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.737116820947603e-06,
+      "loss": 0.4178,
+      "step": 22335
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.736540069786891e-06,
+      "loss": 0.3864,
+      "step": 22336
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.73596331862618e-06,
+      "loss": 0.4132,
+      "step": 22337
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.735386567465466e-06,
+      "loss": 0.3488,
+      "step": 22338
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.734809816304756e-06,
+      "loss": 0.4993,
+      "step": 22339
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.734233065144045e-06,
+      "loss": 0.4629,
+      "step": 22340
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.733656313983333e-06,
+      "loss": 0.4951,
+      "step": 22341
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.73307956282262e-06,
+      "loss": 0.4621,
+      "step": 22342
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.73250281166191e-06,
+      "loss": 0.4514,
+      "step": 22343
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.731926060501198e-06,
+      "loss": 0.512,
+      "step": 22344
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.731349309340485e-06,
+      "loss": 0.4194,
+      "step": 22345
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.730772558179775e-06,
+      "loss": 0.4249,
+      "step": 22346
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.730195807019062e-06,
+      "loss": 0.2897,
+      "step": 22347
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.72961905585835e-06,
+      "loss": 0.56,
+      "step": 22348
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.72904230469764e-06,
+      "loss": 0.4719,
+      "step": 22349
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.728465553536927e-06,
+      "loss": 0.2362,
+      "step": 22350
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.727888802376215e-06,
+      "loss": 0.3471,
+      "step": 22351
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.727312051215504e-06,
+      "loss": 0.341,
+      "step": 22352
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.726735300054792e-06,
+      "loss": 0.4442,
+      "step": 22353
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.726158548894081e-06,
+      "loss": 0.3951,
+      "step": 22354
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.725581797733367e-06,
+      "loss": 0.3943,
+      "step": 22355
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.725005046572657e-06,
+      "loss": 0.5165,
+      "step": 22356
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.724428295411946e-06,
+      "loss": 0.5582,
+      "step": 22357
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.723851544251234e-06,
+      "loss": 0.4296,
+      "step": 22358
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.723274793090521e-06,
+      "loss": 0.5673,
+      "step": 22359
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.72269804192981e-06,
+      "loss": 0.4666,
+      "step": 22360
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.722121290769098e-06,
+      "loss": 0.3834,
+      "step": 22361
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.721544539608386e-06,
+      "loss": 0.4489,
+      "step": 22362
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.720967788447675e-06,
+      "loss": 0.5419,
+      "step": 22363
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.720391037286963e-06,
+      "loss": 0.5648,
+      "step": 22364
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.71981428612625e-06,
+      "loss": 0.4267,
+      "step": 22365
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.71923753496554e-06,
+      "loss": 0.4745,
+      "step": 22366
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.718660783804828e-06,
+      "loss": 0.5874,
+      "step": 22367
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.718084032644117e-06,
+      "loss": 0.4886,
+      "step": 22368
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.717507281483405e-06,
+      "loss": 0.4562,
+      "step": 22369
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716930530322693e-06,
+      "loss": 0.3633,
+      "step": 22370
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.716353779161982e-06,
+      "loss": 0.3718,
+      "step": 22371
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.71577702800127e-06,
+      "loss": 0.3789,
+      "step": 22372
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.715200276840557e-06,
+      "loss": 0.3753,
+      "step": 22373
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.714623525679847e-06,
+      "loss": 0.5396,
+      "step": 22374
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.714046774519134e-06,
+      "loss": 0.3784,
+      "step": 22375
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.713470023358422e-06,
+      "loss": 0.3911,
+      "step": 22376
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.712893272197712e-06,
+      "loss": 0.4462,
+      "step": 22377
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.712316521037e-06,
+      "loss": 0.4045,
+      "step": 22378
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.711739769876287e-06,
+      "loss": 0.3471,
+      "step": 22379
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.711163018715576e-06,
+      "loss": 0.3598,
+      "step": 22380
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.710586267554864e-06,
+      "loss": 0.431,
+      "step": 22381
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.710009516394152e-06,
+      "loss": 0.4561,
+      "step": 22382
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.709432765233441e-06,
+      "loss": 0.4453,
+      "step": 22383
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.708856014072729e-06,
+      "loss": 0.3995,
+      "step": 22384
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.708279262912018e-06,
+      "loss": 0.4654,
+      "step": 22385
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.707702511751306e-06,
+      "loss": 0.4026,
+      "step": 22386
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.707125760590593e-06,
+      "loss": 0.3946,
+      "step": 22387
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.706549009429883e-06,
+      "loss": 0.4588,
+      "step": 22388
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.70597225826917e-06,
+      "loss": 0.4608,
+      "step": 22389
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.705395507108458e-06,
+      "loss": 0.4425,
+      "step": 22390
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.704818755947748e-06,
+      "loss": 0.4285,
+      "step": 22391
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.704242004787035e-06,
+      "loss": 0.3851,
+      "step": 22392
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.703665253626323e-06,
+      "loss": 0.5388,
+      "step": 22393
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.703088502465612e-06,
+      "loss": 0.329,
+      "step": 22394
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.7025117513049e-06,
+      "loss": 0.3828,
+      "step": 22395
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.701935000144188e-06,
+      "loss": 0.3901,
+      "step": 22396
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.701358248983477e-06,
+      "loss": 0.3234,
+      "step": 22397
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.700781497822765e-06,
+      "loss": 0.4799,
+      "step": 22398
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.700204746662054e-06,
+      "loss": 0.3728,
+      "step": 22399
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.699627995501342e-06,
+      "loss": 0.4565,
+      "step": 22400
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.69905124434063e-06,
+      "loss": 0.4601,
+      "step": 22401
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.698474493179919e-06,
+      "loss": 0.4248,
+      "step": 22402
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.697897742019207e-06,
+      "loss": 0.4266,
+      "step": 22403
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.697320990858494e-06,
+      "loss": 0.5094,
+      "step": 22404
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.696744239697784e-06,
+      "loss": 0.3821,
+      "step": 22405
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.696167488537071e-06,
+      "loss": 0.4861,
+      "step": 22406
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695590737376359e-06,
+      "loss": 0.4084,
+      "step": 22407
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.695013986215648e-06,
+      "loss": 0.4029,
+      "step": 22408
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.694437235054936e-06,
+      "loss": 0.4153,
+      "step": 22409
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.693860483894224e-06,
+      "loss": 0.3015,
+      "step": 22410
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.693283732733513e-06,
+      "loss": 0.4564,
+      "step": 22411
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.6927069815728e-06,
+      "loss": 0.3529,
+      "step": 22412
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.692130230412088e-06,
+      "loss": 0.4402,
+      "step": 22413
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.691553479251378e-06,
+      "loss": 0.5304,
+      "step": 22414
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.690976728090666e-06,
+      "loss": 0.3317,
+      "step": 22415
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.690399976929955e-06,
+      "loss": 0.3837,
+      "step": 22416
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.689823225769243e-06,
+      "loss": 0.4102,
+      "step": 22417
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.68924647460853e-06,
+      "loss": 0.4118,
+      "step": 22418
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.68866972344782e-06,
+      "loss": 0.3619,
+      "step": 22419
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.688092972287107e-06,
+      "loss": 0.4123,
+      "step": 22420
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.687516221126395e-06,
+      "loss": 0.494,
+      "step": 22421
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.686939469965684e-06,
+      "loss": 0.4044,
+      "step": 22422
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.686362718804972e-06,
+      "loss": 0.5362,
+      "step": 22423
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.68578596764426e-06,
+      "loss": 0.4338,
+      "step": 22424
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.68520921648355e-06,
+      "loss": 0.5407,
+      "step": 22425
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.684632465322837e-06,
+      "loss": 0.3893,
+      "step": 22426
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.684055714162125e-06,
+      "loss": 0.4617,
+      "step": 22427
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.683478963001414e-06,
+      "loss": 0.3793,
+      "step": 22428
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.682902211840702e-06,
+      "loss": 0.4145,
+      "step": 22429
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.682325460679991e-06,
+      "loss": 0.5225,
+      "step": 22430
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.681748709519279e-06,
+      "loss": 0.3609,
+      "step": 22431
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.681171958358566e-06,
+      "loss": 0.4191,
+      "step": 22432
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 7.680595207197856e-06,
+      "loss": 0.5022,
+      "step": 22433
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.680018456037143e-06,
+      "loss": 0.4433,
+      "step": 22434
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.679441704876431e-06,
+      "loss": 0.5457,
+      "step": 22435
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.67886495371572e-06,
+      "loss": 0.4074,
+      "step": 22436
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.678288202555008e-06,
+      "loss": 0.4305,
+      "step": 22437
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.677711451394296e-06,
+      "loss": 0.399,
+      "step": 22438
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.677134700233585e-06,
+      "loss": 0.438,
+      "step": 22439
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.676557949072873e-06,
+      "loss": 0.3663,
+      "step": 22440
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.67598119791216e-06,
+      "loss": 0.3702,
+      "step": 22441
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.67540444675145e-06,
+      "loss": 0.3757,
+      "step": 22442
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674827695590738e-06,
+      "loss": 0.4869,
+      "step": 22443
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.674250944430025e-06,
+      "loss": 0.4447,
+      "step": 22444
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.673674193269315e-06,
+      "loss": 0.357,
+      "step": 22445
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.673097442108602e-06,
+      "loss": 0.3833,
+      "step": 22446
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.672520690947892e-06,
+      "loss": 0.4384,
+      "step": 22447
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.67194393978718e-06,
+      "loss": 0.4623,
+      "step": 22448
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.671367188626467e-06,
+      "loss": 0.4036,
+      "step": 22449
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.670790437465757e-06,
+      "loss": 0.4284,
+      "step": 22450
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.670213686305044e-06,
+      "loss": 0.4934,
+      "step": 22451
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.669636935144332e-06,
+      "loss": 0.4662,
+      "step": 22452
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.669060183983621e-06,
+      "loss": 0.5691,
+      "step": 22453
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.668483432822909e-06,
+      "loss": 0.4146,
+      "step": 22454
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.667906681662197e-06,
+      "loss": 0.3982,
+      "step": 22455
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.667329930501486e-06,
+      "loss": 0.4161,
+      "step": 22456
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.666753179340774e-06,
+      "loss": 0.4083,
+      "step": 22457
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.666176428180061e-06,
+      "loss": 0.5989,
+      "step": 22458
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.66559967701935e-06,
+      "loss": 0.452,
+      "step": 22459
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.665022925858638e-06,
+      "loss": 0.3468,
+      "step": 22460
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.664446174697928e-06,
+      "loss": 0.5578,
+      "step": 22461
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.663869423537216e-06,
+      "loss": 0.4985,
+      "step": 22462
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.663292672376503e-06,
+      "loss": 0.3934,
+      "step": 22463
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.662715921215793e-06,
+      "loss": 0.4843,
+      "step": 22464
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.66213917005508e-06,
+      "loss": 0.3792,
+      "step": 22465
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.661562418894368e-06,
+      "loss": 0.5563,
+      "step": 22466
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.660985667733657e-06,
+      "loss": 0.3514,
+      "step": 22467
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.660408916572945e-06,
+      "loss": 0.5052,
+      "step": 22468
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.659832165412233e-06,
+      "loss": 0.235,
+      "step": 22469
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.659255414251522e-06,
+      "loss": 0.4982,
+      "step": 22470
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65867866309081e-06,
+      "loss": 0.3566,
+      "step": 22471
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.658101911930097e-06,
+      "loss": 0.3428,
+      "step": 22472
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.657525160769387e-06,
+      "loss": 0.4428,
+      "step": 22473
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.656948409608676e-06,
+      "loss": 0.4439,
+      "step": 22474
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.656371658447962e-06,
+      "loss": 0.4269,
+      "step": 22475
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.655794907287252e-06,
+      "loss": 0.4605,
+      "step": 22476
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.65521815612654e-06,
+      "loss": 0.3877,
+      "step": 22477
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.654641404965829e-06,
+      "loss": 0.3793,
+      "step": 22478
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.654064653805116e-06,
+      "loss": 0.3886,
+      "step": 22479
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.653487902644404e-06,
+      "loss": 0.5376,
+      "step": 22480
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.652911151483693e-06,
+      "loss": 0.5512,
+      "step": 22481
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.652334400322981e-06,
+      "loss": 0.4487,
+      "step": 22482
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.651757649162269e-06,
+      "loss": 0.4523,
+      "step": 22483
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.651180898001558e-06,
+      "loss": 0.4705,
+      "step": 22484
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.650604146840846e-06,
+      "loss": 0.5371,
+      "step": 22485
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.650027395680134e-06,
+      "loss": 0.3088,
+      "step": 22486
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.649450644519423e-06,
+      "loss": 0.3474,
+      "step": 22487
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.64887389335871e-06,
+      "loss": 0.3219,
+      "step": 22488
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.648297142197998e-06,
+      "loss": 0.4199,
+      "step": 22489
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.647720391037288e-06,
+      "loss": 0.309,
+      "step": 22490
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.647143639876577e-06,
+      "loss": 0.3612,
+      "step": 22491
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.646566888715865e-06,
+      "loss": 0.4505,
+      "step": 22492
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.645990137555152e-06,
+      "loss": 0.4563,
+      "step": 22493
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.64541338639444e-06,
+      "loss": 0.4812,
+      "step": 22494
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.64483663523373e-06,
+      "loss": 0.4779,
+      "step": 22495
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.644259884073017e-06,
+      "loss": 0.4981,
+      "step": 22496
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.643683132912305e-06,
+      "loss": 0.5757,
+      "step": 22497
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.643106381751594e-06,
+      "loss": 0.5685,
+      "step": 22498
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.642529630590882e-06,
+      "loss": 0.4176,
+      "step": 22499
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.64195287943017e-06,
+      "loss": 0.497,
+      "step": 22500
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.641376128269459e-06,
+      "loss": 0.3904,
+      "step": 22501
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.640799377108747e-06,
+      "loss": 0.5903,
+      "step": 22502
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.640222625948034e-06,
+      "loss": 0.4114,
+      "step": 22503
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.639645874787324e-06,
+      "loss": 0.4725,
+      "step": 22504
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.639069123626613e-06,
+      "loss": 0.516,
+      "step": 22505
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.638492372465899e-06,
+      "loss": 0.2754,
+      "step": 22506
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.637915621305188e-06,
+      "loss": 0.4741,
+      "step": 22507
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.637338870144478e-06,
+      "loss": 0.4181,
+      "step": 22508
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.636762118983766e-06,
+      "loss": 0.314,
+      "step": 22509
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.636185367823053e-06,
+      "loss": 0.44,
+      "step": 22510
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.635608616662343e-06,
+      "loss": 0.4156,
+      "step": 22511
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.63503186550163e-06,
+      "loss": 0.5176,
+      "step": 22512
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.634455114340918e-06,
+      "loss": 0.4551,
+      "step": 22513
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633878363180206e-06,
+      "loss": 0.3634,
+      "step": 22514
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.633301612019495e-06,
+      "loss": 0.3825,
+      "step": 22515
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.632724860858783e-06,
+      "loss": 0.4704,
+      "step": 22516
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.63214810969807e-06,
+      "loss": 0.4196,
+      "step": 22517
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.63157135853736e-06,
+      "loss": 0.3122,
+      "step": 22518
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.630994607376647e-06,
+      "loss": 0.5449,
+      "step": 22519
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.630417856215935e-06,
+      "loss": 0.4715,
+      "step": 22520
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.629841105055225e-06,
+      "loss": 0.3401,
+      "step": 22521
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.629264353894513e-06,
+      "loss": 0.5171,
+      "step": 22522
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.628687602733801e-06,
+      "loss": 0.399,
+      "step": 22523
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.628110851573089e-06,
+      "loss": 0.4463,
+      "step": 22524
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.627534100412378e-06,
+      "loss": 0.4492,
+      "step": 22525
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6269573492516655e-06,
+      "loss": 0.4766,
+      "step": 22526
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.626380598090954e-06,
+      "loss": 0.3992,
+      "step": 22527
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6258038469302434e-06,
+      "loss": 0.3446,
+      "step": 22528
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.62522709576953e-06,
+      "loss": 0.4968,
+      "step": 22529
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.62465034460882e-06,
+      "loss": 0.5653,
+      "step": 22530
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.624073593448107e-06,
+      "loss": 0.4766,
+      "step": 22531
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.623496842287396e-06,
+      "loss": 0.4879,
+      "step": 22532
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.622920091126684e-06,
+      "loss": 0.3467,
+      "step": 22533
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.622343339965972e-06,
+      "loss": 0.4738,
+      "step": 22534
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.621766588805261e-06,
+      "loss": 0.3362,
+      "step": 22535
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.621189837644549e-06,
+      "loss": 0.4446,
+      "step": 22536
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.620613086483837e-06,
+      "loss": 0.3739,
+      "step": 22537
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.620036335323125e-06,
+      "loss": 0.4318,
+      "step": 22538
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.619459584162414e-06,
+      "loss": 0.3813,
+      "step": 22539
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.618882833001702e-06,
+      "loss": 0.4862,
+      "step": 22540
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.61830608184099e-06,
+      "loss": 0.3576,
+      "step": 22541
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6177293306802795e-06,
+      "loss": 0.3942,
+      "step": 22542
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.617152579519566e-06,
+      "loss": 0.3951,
+      "step": 22543
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.616575828358856e-06,
+      "loss": 0.3941,
+      "step": 22544
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.615999077198144e-06,
+      "loss": 0.2704,
+      "step": 22545
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.615422326037432e-06,
+      "loss": 0.4863,
+      "step": 22546
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6148455748767205e-06,
+      "loss": 0.4336,
+      "step": 22547
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.614268823716008e-06,
+      "loss": 0.4653,
+      "step": 22548
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613692072555297e-06,
+      "loss": 0.5465,
+      "step": 22549
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.613115321394585e-06,
+      "loss": 0.4578,
+      "step": 22550
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.612538570233873e-06,
+      "loss": 0.4468,
+      "step": 22551
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6119618190731614e-06,
+      "loss": 0.3279,
+      "step": 22552
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.61138506791245e-06,
+      "loss": 0.3739,
+      "step": 22553
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.610808316751738e-06,
+      "loss": 0.3099,
+      "step": 22554
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.610231565591026e-06,
+      "loss": 0.543,
+      "step": 22555
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.609654814430315e-06,
+      "loss": 0.4018,
+      "step": 22556
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.609078063269602e-06,
+      "loss": 0.4277,
+      "step": 22557
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.608501312108891e-06,
+      "loss": 0.3568,
+      "step": 22558
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.60792456094818e-06,
+      "loss": 0.2902,
+      "step": 22559
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.607347809787467e-06,
+      "loss": 0.4332,
+      "step": 22560
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6067710586267565e-06,
+      "loss": 0.4127,
+      "step": 22561
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.606194307466045e-06,
+      "loss": 0.443,
+      "step": 22562
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.605617556305333e-06,
+      "loss": 0.4318,
+      "step": 22563
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.605040805144621e-06,
+      "loss": 0.343,
+      "step": 22564
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.604464053983909e-06,
+      "loss": 0.393,
+      "step": 22565
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6038873028231975e-06,
+      "loss": 0.5021,
+      "step": 22566
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.603310551662486e-06,
+      "loss": 0.479,
+      "step": 22567
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.602733800501774e-06,
+      "loss": 0.4033,
+      "step": 22568
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.602157049341062e-06,
+      "loss": 0.5572,
+      "step": 22569
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.601580298180351e-06,
+      "loss": 0.4104,
+      "step": 22570
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.6010035470196385e-06,
+      "loss": 0.4441,
+      "step": 22571
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.600426795858927e-06,
+      "loss": 0.3968,
+      "step": 22572
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.599850044698216e-06,
+      "loss": 0.4372,
+      "step": 22573
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.599273293537503e-06,
+      "loss": 0.3684,
+      "step": 22574
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.598696542376793e-06,
+      "loss": 0.4263,
+      "step": 22575
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.598119791216081e-06,
+      "loss": 0.5809,
+      "step": 22576
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.597543040055369e-06,
+      "loss": 0.4251,
+      "step": 22577
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.596966288894657e-06,
+      "loss": 0.4293,
+      "step": 22578
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.596389537733946e-06,
+      "loss": 0.4061,
+      "step": 22579
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.5958127865732336e-06,
+      "loss": 0.5122,
+      "step": 22580
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.595236035412522e-06,
+      "loss": 0.4679,
+      "step": 22581
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.59465928425181e-06,
+      "loss": 0.4555,
+      "step": 22582
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.594082533091098e-06,
+      "loss": 0.3918,
+      "step": 22583
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.593505781930387e-06,
+      "loss": 0.3801,
+      "step": 22584
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.5929290307696745e-06,
+      "loss": 0.3835,
+      "step": 22585
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.592352279608963e-06,
+      "loss": 0.516,
+      "step": 22586
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.591775528448252e-06,
+      "loss": 0.4056,
+      "step": 22587
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.591198777287539e-06,
+      "loss": 0.4952,
+      "step": 22588
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.590622026126828e-06,
+      "loss": 0.4253,
+      "step": 22589
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.590045274966117e-06,
+      "loss": 0.3875,
+      "step": 22590
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.589468523805405e-06,
+      "loss": 0.3981,
+      "step": 22591
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.588891772644693e-06,
+      "loss": 0.5257,
+      "step": 22592
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.588315021483982e-06,
+      "loss": 0.458,
+      "step": 22593
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.58773827032327e-06,
+      "loss": 0.5417,
+      "step": 22594
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.587161519162558e-06,
+      "loss": 0.3063,
+      "step": 22595
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.586584768001847e-06,
+      "loss": 0.3232,
+      "step": 22596
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.586008016841134e-06,
+      "loss": 0.4508,
+      "step": 22597
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.585431265680423e-06,
+      "loss": 0.4549,
+      "step": 22598
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.584854514519711e-06,
+      "loss": 0.4225,
+      "step": 22599
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.584277763358999e-06,
+      "loss": 0.3567,
+      "step": 22600
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.583701012198288e-06,
+      "loss": 0.4178,
+      "step": 22601
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.583124261037575e-06,
+      "loss": 0.4756,
+      "step": 22602
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.582547509876864e-06,
+      "loss": 0.5098,
+      "step": 22603
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.581970758716153e-06,
+      "loss": 0.5372,
+      "step": 22604
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.58139400755544e-06,
+      "loss": 0.4065,
+      "step": 22605
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.5808172563947295e-06,
+      "loss": 0.3825,
+      "step": 22606
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.580240505234018e-06,
+      "loss": 0.3719,
+      "step": 22607
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.579663754073306e-06,
+      "loss": 0.4552,
+      "step": 22608
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.579087002912594e-06,
+      "loss": 0.4353,
+      "step": 22609
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.578510251751883e-06,
+      "loss": 0.3829,
+      "step": 22610
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 7.5779335005911704e-06,
+      "loss": 0.4691,
+      "step": 22611
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.577356749430459e-06,
+      "loss": 0.5124,
+      "step": 22612
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5767799982697475e-06,
+      "loss": 0.536,
+      "step": 22613
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.576203247109035e-06,
+      "loss": 0.4467,
+      "step": 22614
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.575626495948324e-06,
+      "loss": 0.3311,
+      "step": 22615
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.575049744787611e-06,
+      "loss": 0.5012,
+      "step": 22616
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5744729936269e-06,
+      "loss": 0.4167,
+      "step": 22617
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5738962424661885e-06,
+      "loss": 0.3826,
+      "step": 22618
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.573319491305476e-06,
+      "loss": 0.398,
+      "step": 22619
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.572742740144765e-06,
+      "loss": 0.5695,
+      "step": 22620
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.572165988984054e-06,
+      "loss": 0.4157,
+      "step": 22621
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.571589237823342e-06,
+      "loss": 0.4476,
+      "step": 22622
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.57101248666263e-06,
+      "loss": 0.4171,
+      "step": 22623
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.570435735501919e-06,
+      "loss": 0.3613,
+      "step": 22624
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5698589843412065e-06,
+      "loss": 0.3973,
+      "step": 22625
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.569282233180495e-06,
+      "loss": 0.438,
+      "step": 22626
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5687054820197836e-06,
+      "loss": 0.3364,
+      "step": 22627
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.568128730859071e-06,
+      "loss": 0.4014,
+      "step": 22628
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.56755197969836e-06,
+      "loss": 0.4502,
+      "step": 22629
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.566975228537648e-06,
+      "loss": 0.5316,
+      "step": 22630
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.566398477376936e-06,
+      "loss": 0.45,
+      "step": 22631
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5658217262162245e-06,
+      "loss": 0.4489,
+      "step": 22632
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.565244975055512e-06,
+      "loss": 0.5035,
+      "step": 22633
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.564668223894801e-06,
+      "loss": 0.5519,
+      "step": 22634
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.56409147273409e-06,
+      "loss": 0.3899,
+      "step": 22635
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.563514721573377e-06,
+      "loss": 0.4626,
+      "step": 22636
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.562937970412666e-06,
+      "loss": 0.3361,
+      "step": 22637
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.562361219251955e-06,
+      "loss": 0.5567,
+      "step": 22638
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5617844680912425e-06,
+      "loss": 0.4387,
+      "step": 22639
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.561207716930531e-06,
+      "loss": 0.4807,
+      "step": 22640
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.56063096576982e-06,
+      "loss": 0.4311,
+      "step": 22641
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.560054214609107e-06,
+      "loss": 0.4144,
+      "step": 22642
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.559477463448396e-06,
+      "loss": 0.408,
+      "step": 22643
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.558900712287684e-06,
+      "loss": 0.4657,
+      "step": 22644
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.558323961126972e-06,
+      "loss": 0.5748,
+      "step": 22645
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.557747209966261e-06,
+      "loss": 0.5323,
+      "step": 22646
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.557170458805549e-06,
+      "loss": 0.4046,
+      "step": 22647
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.556593707644837e-06,
+      "loss": 0.487,
+      "step": 22648
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.556016956484125e-06,
+      "loss": 0.4316,
+      "step": 22649
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.555440205323415e-06,
+      "loss": 0.4116,
+      "step": 22650
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5548634541627015e-06,
+      "loss": 0.4007,
+      "step": 22651
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.554286703001991e-06,
+      "loss": 0.3853,
+      "step": 22652
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.553709951841279e-06,
+      "loss": 0.3702,
+      "step": 22653
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.553133200680567e-06,
+      "loss": 0.4428,
+      "step": 22654
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.552556449519856e-06,
+      "loss": 0.4694,
+      "step": 22655
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551979698359143e-06,
+      "loss": 0.4597,
+      "step": 22656
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.551402947198432e-06,
+      "loss": 0.4889,
+      "step": 22657
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5508261960377204e-06,
+      "loss": 0.4077,
+      "step": 22658
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.550249444877008e-06,
+      "loss": 0.4618,
+      "step": 22659
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.549672693716297e-06,
+      "loss": 0.4234,
+      "step": 22660
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.549095942555585e-06,
+      "loss": 0.3467,
+      "step": 22661
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.548519191394873e-06,
+      "loss": 0.4103,
+      "step": 22662
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.547942440234161e-06,
+      "loss": 0.3909,
+      "step": 22663
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.547365689073451e-06,
+      "loss": 0.404,
+      "step": 22664
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.546788937912738e-06,
+      "loss": 0.462,
+      "step": 22665
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.546212186752027e-06,
+      "loss": 0.4365,
+      "step": 22666
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5456354355913155e-06,
+      "loss": 0.4652,
+      "step": 22667
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.545058684430603e-06,
+      "loss": 0.4897,
+      "step": 22668
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.544481933269892e-06,
+      "loss": 0.4834,
+      "step": 22669
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.543905182109179e-06,
+      "loss": 0.3943,
+      "step": 22670
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.543328430948468e-06,
+      "loss": 0.5051,
+      "step": 22671
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5427516797877565e-06,
+      "loss": 0.339,
+      "step": 22672
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.542174928627044e-06,
+      "loss": 0.3584,
+      "step": 22673
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.541598177466333e-06,
+      "loss": 0.3145,
+      "step": 22674
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.541021426305621e-06,
+      "loss": 0.4539,
+      "step": 22675
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.540444675144909e-06,
+      "loss": 0.3811,
+      "step": 22676
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5398679239841975e-06,
+      "loss": 0.336,
+      "step": 22677
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.539291172823486e-06,
+      "loss": 0.4023,
+      "step": 22678
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.538714421662774e-06,
+      "loss": 0.5496,
+      "step": 22679
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.538137670502062e-06,
+      "loss": 0.4913,
+      "step": 22680
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.537560919341352e-06,
+      "loss": 0.4463,
+      "step": 22681
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.536984168180638e-06,
+      "loss": 0.4248,
+      "step": 22682
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.536407417019928e-06,
+      "loss": 0.3449,
+      "step": 22683
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.535830665859216e-06,
+      "loss": 0.5454,
+      "step": 22684
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.535253914698504e-06,
+      "loss": 0.4069,
+      "step": 22685
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5346771635377925e-06,
+      "loss": 0.3854,
+      "step": 22686
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.53410041237708e-06,
+      "loss": 0.4566,
+      "step": 22687
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.533523661216369e-06,
+      "loss": 0.3998,
+      "step": 22688
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.532946910055657e-06,
+      "loss": 0.4158,
+      "step": 22689
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.532370158894945e-06,
+      "loss": 0.4696,
+      "step": 22690
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5317934077342335e-06,
+      "loss": 0.3832,
+      "step": 22691
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.531216656573522e-06,
+      "loss": 0.4032,
+      "step": 22692
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.53063990541281e-06,
+      "loss": 0.4257,
+      "step": 22693
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.530063154252098e-06,
+      "loss": 0.4171,
+      "step": 22694
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.529486403091388e-06,
+      "loss": 0.4504,
+      "step": 22695
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5289096519306745e-06,
+      "loss": 0.5682,
+      "step": 22696
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.528332900769964e-06,
+      "loss": 0.445,
+      "step": 22697
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.527756149609252e-06,
+      "loss": 0.3414,
+      "step": 22698
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.52717939844854e-06,
+      "loss": 0.3218,
+      "step": 22699
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.526602647287829e-06,
+      "loss": 0.5149,
+      "step": 22700
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.526025896127117e-06,
+      "loss": 0.428,
+      "step": 22701
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.525449144966405e-06,
+      "loss": 0.4079,
+      "step": 22702
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.524872393805693e-06,
+      "loss": 0.3252,
+      "step": 22703
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.524295642644981e-06,
+      "loss": 0.4232,
+      "step": 22704
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5237188914842696e-06,
+      "loss": 0.442,
+      "step": 22705
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.523142140323558e-06,
+      "loss": 0.3262,
+      "step": 22706
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.522565389162846e-06,
+      "loss": 0.4818,
+      "step": 22707
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.521988638002134e-06,
+      "loss": 0.3986,
+      "step": 22708
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.521411886841423e-06,
+      "loss": 0.4453,
+      "step": 22709
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5208351356807105e-06,
+      "loss": 0.5184,
+      "step": 22710
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.520258384519999e-06,
+      "loss": 0.3748,
+      "step": 22711
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5196816333592885e-06,
+      "loss": 0.394,
+      "step": 22712
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.519104882198575e-06,
+      "loss": 0.4285,
+      "step": 22713
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.518528131037865e-06,
+      "loss": 0.4311,
+      "step": 22714
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.517951379877153e-06,
+      "loss": 0.4503,
+      "step": 22715
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.517374628716441e-06,
+      "loss": 0.4915,
+      "step": 22716
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.516797877555729e-06,
+      "loss": 0.346,
+      "step": 22717
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.516221126395018e-06,
+      "loss": 0.3529,
+      "step": 22718
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.515644375234306e-06,
+      "loss": 0.4034,
+      "step": 22719
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.515067624073594e-06,
+      "loss": 0.425,
+      "step": 22720
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.514490872912882e-06,
+      "loss": 0.386,
+      "step": 22721
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.51391412175217e-06,
+      "loss": 0.5029,
+      "step": 22722
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.513337370591459e-06,
+      "loss": 0.5082,
+      "step": 22723
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.512760619430747e-06,
+      "loss": 0.4788,
+      "step": 22724
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.512183868270035e-06,
+      "loss": 0.4408,
+      "step": 22725
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5116071171093245e-06,
+      "loss": 0.4199,
+      "step": 22726
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.511030365948611e-06,
+      "loss": 0.3488,
+      "step": 22727
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.510453614787901e-06,
+      "loss": 0.4157,
+      "step": 22728
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.509876863627189e-06,
+      "loss": 0.4772,
+      "step": 22729
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.509300112466477e-06,
+      "loss": 0.3325,
+      "step": 22730
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5087233613057655e-06,
+      "loss": 0.5031,
+      "step": 22731
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.508146610145054e-06,
+      "loss": 0.3545,
+      "step": 22732
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.507569858984342e-06,
+      "loss": 0.4052,
+      "step": 22733
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.50699310782363e-06,
+      "loss": 0.3194,
+      "step": 22734
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.506416356662919e-06,
+      "loss": 0.6491,
+      "step": 22735
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5058396055022064e-06,
+      "loss": 0.4813,
+      "step": 22736
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.505262854341495e-06,
+      "loss": 0.523,
+      "step": 22737
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.504686103180783e-06,
+      "loss": 0.4201,
+      "step": 22738
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.504109352020071e-06,
+      "loss": 0.4467,
+      "step": 22739
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.50353260085936e-06,
+      "loss": 0.3618,
+      "step": 22740
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.502955849698647e-06,
+      "loss": 0.3402,
+      "step": 22741
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.502379098537936e-06,
+      "loss": 0.4539,
+      "step": 22742
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.501802347377225e-06,
+      "loss": 0.4972,
+      "step": 22743
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.501225596216512e-06,
+      "loss": 0.409,
+      "step": 22744
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.5006488450558015e-06,
+      "loss": 0.5505,
+      "step": 22745
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.50007209389509e-06,
+      "loss": 0.5004,
+      "step": 22746
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.499495342734378e-06,
+      "loss": 0.4494,
+      "step": 22747
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.498918591573666e-06,
+      "loss": 0.4857,
+      "step": 22748
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.498341840412955e-06,
+      "loss": 0.3947,
+      "step": 22749
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4977650892522425e-06,
+      "loss": 0.393,
+      "step": 22750
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.497188338091531e-06,
+      "loss": 0.4408,
+      "step": 22751
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4966115869308196e-06,
+      "loss": 0.518,
+      "step": 22752
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.496034835770107e-06,
+      "loss": 0.3296,
+      "step": 22753
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.495458084609396e-06,
+      "loss": 0.3914,
+      "step": 22754
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4948813334486835e-06,
+      "loss": 0.4152,
+      "step": 22755
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.494304582287972e-06,
+      "loss": 0.453,
+      "step": 22756
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.493727831127261e-06,
+      "loss": 0.4523,
+      "step": 22757
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.493151079966548e-06,
+      "loss": 0.4874,
+      "step": 22758
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.492574328805838e-06,
+      "loss": 0.4463,
+      "step": 22759
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.491997577645126e-06,
+      "loss": 0.4325,
+      "step": 22760
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.491420826484414e-06,
+      "loss": 0.4222,
+      "step": 22761
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.490844075323702e-06,
+      "loss": 0.2996,
+      "step": 22762
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.490267324162991e-06,
+      "loss": 0.5772,
+      "step": 22763
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4896905730022786e-06,
+      "loss": 0.5511,
+      "step": 22764
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.489113821841567e-06,
+      "loss": 0.4597,
+      "step": 22765
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.488537070680856e-06,
+      "loss": 0.3674,
+      "step": 22766
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.487960319520143e-06,
+      "loss": 0.2944,
+      "step": 22767
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.487383568359432e-06,
+      "loss": 0.342,
+      "step": 22768
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48680681719872e-06,
+      "loss": 0.4803,
+      "step": 22769
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.486230066038008e-06,
+      "loss": 0.4676,
+      "step": 22770
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.485653314877297e-06,
+      "loss": 0.3399,
+      "step": 22771
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.485076563716586e-06,
+      "loss": 0.5435,
+      "step": 22772
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.484499812555873e-06,
+      "loss": 0.4692,
+      "step": 22773
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.483923061395162e-06,
+      "loss": 0.5212,
+      "step": 22774
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.48334631023445e-06,
+      "loss": 0.3808,
+      "step": 22775
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.482769559073738e-06,
+      "loss": 0.381,
+      "step": 22776
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.482192807913027e-06,
+      "loss": 0.3376,
+      "step": 22777
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.481616056752315e-06,
+      "loss": 0.5977,
+      "step": 22778
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.481039305591603e-06,
+      "loss": 0.3599,
+      "step": 22779
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.480462554430892e-06,
+      "loss": 0.3273,
+      "step": 22780
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.479885803270179e-06,
+      "loss": 0.4984,
+      "step": 22781
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.479309052109468e-06,
+      "loss": 0.535,
+      "step": 22782
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4787323009487564e-06,
+      "loss": 0.3709,
+      "step": 22783
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.478155549788044e-06,
+      "loss": 0.4323,
+      "step": 22784
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.477578798627333e-06,
+      "loss": 0.3467,
+      "step": 22785
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.477002047466621e-06,
+      "loss": 0.4,
+      "step": 22786
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.476425296305909e-06,
+      "loss": 0.5414,
+      "step": 22787
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.475848545145198e-06,
+      "loss": 0.4716,
+      "step": 22788
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.475271793984487e-06,
+      "loss": 0.4796,
+      "step": 22789
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 7.4746950428237745e-06,
+      "loss": 0.4598,
+      "step": 22790
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.474118291663063e-06,
+      "loss": 0.5283,
+      "step": 22791
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.473541540502351e-06,
+      "loss": 0.4174,
+      "step": 22792
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.472964789341639e-06,
+      "loss": 0.3772,
+      "step": 22793
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.472388038180928e-06,
+      "loss": 0.3875,
+      "step": 22794
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4718112870202154e-06,
+      "loss": 0.4637,
+      "step": 22795
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.471234535859504e-06,
+      "loss": 0.3822,
+      "step": 22796
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4706577846987925e-06,
+      "loss": 0.4906,
+      "step": 22797
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.47008103353808e-06,
+      "loss": 0.421,
+      "step": 22798
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.469504282377369e-06,
+      "loss": 0.3812,
+      "step": 22799
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.468927531216657e-06,
+      "loss": 0.5108,
+      "step": 22800
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.468350780055945e-06,
+      "loss": 0.4316,
+      "step": 22801
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4677740288952335e-06,
+      "loss": 0.4577,
+      "step": 22802
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.467197277734523e-06,
+      "loss": 0.4335,
+      "step": 22803
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.46662052657381e-06,
+      "loss": 0.4361,
+      "step": 22804
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.466043775413099e-06,
+      "loss": 0.4074,
+      "step": 22805
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.465467024252388e-06,
+      "loss": 0.4511,
+      "step": 22806
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.464890273091675e-06,
+      "loss": 0.4541,
+      "step": 22807
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.464313521930964e-06,
+      "loss": 0.4381,
+      "step": 22808
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4637367707702515e-06,
+      "loss": 0.3043,
+      "step": 22809
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.46316001960954e-06,
+      "loss": 0.4078,
+      "step": 22810
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4625832684488286e-06,
+      "loss": 0.4414,
+      "step": 22811
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.462006517288116e-06,
+      "loss": 0.4814,
+      "step": 22812
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.461429766127405e-06,
+      "loss": 0.5462,
+      "step": 22813
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.460853014966693e-06,
+      "loss": 0.5346,
+      "step": 22814
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.460276263805981e-06,
+      "loss": 0.4117,
+      "step": 22815
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4596995126452695e-06,
+      "loss": 0.4277,
+      "step": 22816
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.459122761484559e-06,
+      "loss": 0.5028,
+      "step": 22817
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.458546010323846e-06,
+      "loss": 0.589,
+      "step": 22818
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.457969259163135e-06,
+      "loss": 0.4079,
+      "step": 22819
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.457392508002424e-06,
+      "loss": 0.3803,
+      "step": 22820
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.456815756841711e-06,
+      "loss": 0.3321,
+      "step": 22821
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.456239005681e-06,
+      "loss": 0.4291,
+      "step": 22822
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.455662254520288e-06,
+      "loss": 0.3773,
+      "step": 22823
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.455085503359576e-06,
+      "loss": 0.3992,
+      "step": 22824
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.454508752198865e-06,
+      "loss": 0.4207,
+      "step": 22825
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.453932001038152e-06,
+      "loss": 0.4982,
+      "step": 22826
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.453355249877441e-06,
+      "loss": 0.3895,
+      "step": 22827
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.452778498716729e-06,
+      "loss": 0.5144,
+      "step": 22828
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.452201747556017e-06,
+      "loss": 0.4183,
+      "step": 22829
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.451624996395306e-06,
+      "loss": 0.543,
+      "step": 22830
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.451048245234594e-06,
+      "loss": 0.4409,
+      "step": 22831
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.450471494073882e-06,
+      "loss": 0.3524,
+      "step": 22832
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.44989474291317e-06,
+      "loss": 0.4248,
+      "step": 22833
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.44931799175246e-06,
+      "loss": 0.3905,
+      "step": 22834
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4487412405917466e-06,
+      "loss": 0.4235,
+      "step": 22835
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.448164489431036e-06,
+      "loss": 0.3997,
+      "step": 22836
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4475877382703245e-06,
+      "loss": 0.4614,
+      "step": 22837
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.447010987109612e-06,
+      "loss": 0.3162,
+      "step": 22838
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.446434235948901e-06,
+      "loss": 0.5153,
+      "step": 22839
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.445857484788189e-06,
+      "loss": 0.4515,
+      "step": 22840
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.445280733627477e-06,
+      "loss": 0.3942,
+      "step": 22841
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4447039824667654e-06,
+      "loss": 0.5392,
+      "step": 22842
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.444127231306053e-06,
+      "loss": 0.3348,
+      "step": 22843
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.443550480145342e-06,
+      "loss": 0.4066,
+      "step": 22844
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.44297372898463e-06,
+      "loss": 0.4928,
+      "step": 22845
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.442396977823918e-06,
+      "loss": 0.4677,
+      "step": 22846
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.441820226663206e-06,
+      "loss": 0.5053,
+      "step": 22847
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.441243475502496e-06,
+      "loss": 0.4846,
+      "step": 22848
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.440666724341783e-06,
+      "loss": 0.4811,
+      "step": 22849
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.440089973181072e-06,
+      "loss": 0.4509,
+      "step": 22850
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4395132220203605e-06,
+      "loss": 0.4062,
+      "step": 22851
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.438936470859648e-06,
+      "loss": 0.5068,
+      "step": 22852
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.438359719698937e-06,
+      "loss": 0.437,
+      "step": 22853
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.437782968538225e-06,
+      "loss": 0.3836,
+      "step": 22854
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.437206217377513e-06,
+      "loss": 0.4452,
+      "step": 22855
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4366294662168015e-06,
+      "loss": 0.4529,
+      "step": 22856
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.43605271505609e-06,
+      "loss": 0.4369,
+      "step": 22857
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.435475963895378e-06,
+      "loss": 0.3726,
+      "step": 22858
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.434899212734666e-06,
+      "loss": 0.3527,
+      "step": 22859
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.434322461573954e-06,
+      "loss": 0.4533,
+      "step": 22860
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4337457104132425e-06,
+      "loss": 0.3663,
+      "step": 22861
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.433168959252531e-06,
+      "loss": 0.6043,
+      "step": 22862
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.432592208091819e-06,
+      "loss": 0.5059,
+      "step": 22863
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.432015456931107e-06,
+      "loss": 0.4027,
+      "step": 22864
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.431438705770397e-06,
+      "loss": 0.4433,
+      "step": 22865
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4308619546096834e-06,
+      "loss": 0.4279,
+      "step": 22866
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.430285203448973e-06,
+      "loss": 0.4006,
+      "step": 22867
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.429708452288261e-06,
+      "loss": 0.4795,
+      "step": 22868
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.429131701127549e-06,
+      "loss": 0.3488,
+      "step": 22869
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4285549499668376e-06,
+      "loss": 0.5149,
+      "step": 22870
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427978198806126e-06,
+      "loss": 0.3653,
+      "step": 22871
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.427401447645414e-06,
+      "loss": 0.519,
+      "step": 22872
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.426824696484702e-06,
+      "loss": 0.3903,
+      "step": 22873
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.426247945323991e-06,
+      "loss": 0.4911,
+      "step": 22874
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4256711941632785e-06,
+      "loss": 0.4467,
+      "step": 22875
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.425094443002567e-06,
+      "loss": 0.5525,
+      "step": 22876
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.424517691841855e-06,
+      "loss": 0.3963,
+      "step": 22877
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.423940940681143e-06,
+      "loss": 0.3904,
+      "step": 22878
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.423364189520433e-06,
+      "loss": 0.3905,
+      "step": 22879
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4227874383597195e-06,
+      "loss": 0.3469,
+      "step": 22880
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.422210687199009e-06,
+      "loss": 0.5065,
+      "step": 22881
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.421633936038297e-06,
+      "loss": 0.4167,
+      "step": 22882
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.421057184877585e-06,
+      "loss": 0.4434,
+      "step": 22883
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.420480433716874e-06,
+      "loss": 0.4425,
+      "step": 22884
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.419903682556162e-06,
+      "loss": 0.4025,
+      "step": 22885
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.41932693139545e-06,
+      "loss": 0.3481,
+      "step": 22886
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.418750180234738e-06,
+      "loss": 0.4654,
+      "step": 22887
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.418173429074027e-06,
+      "loss": 0.3482,
+      "step": 22888
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.417596677913315e-06,
+      "loss": 0.4475,
+      "step": 22889
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.417019926752603e-06,
+      "loss": 0.4747,
+      "step": 22890
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.416443175591892e-06,
+      "loss": 0.4729,
+      "step": 22891
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.415866424431179e-06,
+      "loss": 0.4402,
+      "step": 22892
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.415289673270468e-06,
+      "loss": 0.6373,
+      "step": 22893
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.414712922109757e-06,
+      "loss": 0.2979,
+      "step": 22894
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.414136170949044e-06,
+      "loss": 0.5189,
+      "step": 22895
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4135594197883335e-06,
+      "loss": 0.5777,
+      "step": 22896
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.41298266862762e-06,
+      "loss": 0.4434,
+      "step": 22897
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.41240591746691e-06,
+      "loss": 0.4074,
+      "step": 22898
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.411829166306198e-06,
+      "loss": 0.373,
+      "step": 22899
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.411252415145486e-06,
+      "loss": 0.4782,
+      "step": 22900
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4106756639847744e-06,
+      "loss": 0.5435,
+      "step": 22901
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.410098912824063e-06,
+      "loss": 0.4616,
+      "step": 22902
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.409522161663351e-06,
+      "loss": 0.4206,
+      "step": 22903
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.408945410502639e-06,
+      "loss": 0.3263,
+      "step": 22904
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.408368659341928e-06,
+      "loss": 0.446,
+      "step": 22905
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407791908181215e-06,
+      "loss": 0.482,
+      "step": 22906
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.407215157020504e-06,
+      "loss": 0.4757,
+      "step": 22907
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4066384058597925e-06,
+      "loss": 0.4187,
+      "step": 22908
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.40606165469908e-06,
+      "loss": 0.4913,
+      "step": 22909
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4054849035383695e-06,
+      "loss": 0.5614,
+      "step": 22910
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.404908152377658e-06,
+      "loss": 0.4448,
+      "step": 22911
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.404331401216946e-06,
+      "loss": 0.2881,
+      "step": 22912
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.403754650056234e-06,
+      "loss": 0.5052,
+      "step": 22913
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.403177898895522e-06,
+      "loss": 0.4282,
+      "step": 22914
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.4026011477348105e-06,
+      "loss": 0.4064,
+      "step": 22915
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.402024396574099e-06,
+      "loss": 0.427,
+      "step": 22916
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.401447645413387e-06,
+      "loss": 0.3678,
+      "step": 22917
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.400870894252675e-06,
+      "loss": 0.4566,
+      "step": 22918
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.400294143091964e-06,
+      "loss": 0.3809,
+      "step": 22919
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3997173919312515e-06,
+      "loss": 0.4023,
+      "step": 22920
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.39914064077054e-06,
+      "loss": 0.4743,
+      "step": 22921
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3985638896098285e-06,
+      "loss": 0.4284,
+      "step": 22922
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.397987138449116e-06,
+      "loss": 0.3956,
+      "step": 22923
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.397410387288405e-06,
+      "loss": 0.3982,
+      "step": 22924
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.396833636127694e-06,
+      "loss": 0.5021,
+      "step": 22925
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.396256884966981e-06,
+      "loss": 0.4414,
+      "step": 22926
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.39568013380627e-06,
+      "loss": 0.5111,
+      "step": 22927
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.395103382645559e-06,
+      "loss": 0.3548,
+      "step": 22928
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3945266314848466e-06,
+      "loss": 0.4026,
+      "step": 22929
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.393949880324135e-06,
+      "loss": 0.5511,
+      "step": 22930
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.393373129163423e-06,
+      "loss": 0.5206,
+      "step": 22931
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.392796378002711e-06,
+      "loss": 0.4875,
+      "step": 22932
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.392219626842e-06,
+      "loss": 0.4607,
+      "step": 22933
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3916428756812875e-06,
+      "loss": 0.3693,
+      "step": 22934
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.391066124520576e-06,
+      "loss": 0.372,
+      "step": 22935
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.390489373359865e-06,
+      "loss": 0.3928,
+      "step": 22936
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.389912622199152e-06,
+      "loss": 0.5182,
+      "step": 22937
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.389335871038441e-06,
+      "loss": 0.4701,
+      "step": 22938
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.388759119877729e-06,
+      "loss": 0.4287,
+      "step": 22939
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.388182368717017e-06,
+      "loss": 0.3326,
+      "step": 22940
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.387605617556306e-06,
+      "loss": 0.4055,
+      "step": 22941
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.387028866395595e-06,
+      "loss": 0.6133,
+      "step": 22942
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.386452115234883e-06,
+      "loss": 0.4001,
+      "step": 22943
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.385875364074171e-06,
+      "loss": 0.4335,
+      "step": 22944
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.38529861291346e-06,
+      "loss": 0.5128,
+      "step": 22945
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.384721861752747e-06,
+      "loss": 0.3694,
+      "step": 22946
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.384145110592036e-06,
+      "loss": 0.4418,
+      "step": 22947
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.383568359431324e-06,
+      "loss": 0.3955,
+      "step": 22948
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.382991608270612e-06,
+      "loss": 0.4723,
+      "step": 22949
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.382414857109901e-06,
+      "loss": 0.4724,
+      "step": 22950
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.381838105949188e-06,
+      "loss": 0.405,
+      "step": 22951
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.381261354788477e-06,
+      "loss": 0.4,
+      "step": 22952
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.380684603627765e-06,
+      "loss": 0.4789,
+      "step": 22953
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.380107852467053e-06,
+      "loss": 0.3739,
+      "step": 22954
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.379531101306342e-06,
+      "loss": 0.3943,
+      "step": 22955
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.378954350145631e-06,
+      "loss": 0.4968,
+      "step": 22956
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.378377598984918e-06,
+      "loss": 0.5011,
+      "step": 22957
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.377800847824207e-06,
+      "loss": 0.3185,
+      "step": 22958
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.377224096663496e-06,
+      "loss": 0.4256,
+      "step": 22959
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3766473455027834e-06,
+      "loss": 0.337,
+      "step": 22960
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.376070594342072e-06,
+      "loss": 0.4152,
+      "step": 22961
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3754938431813605e-06,
+      "loss": 0.2474,
+      "step": 22962
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.374917092020648e-06,
+      "loss": 0.4178,
+      "step": 22963
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.374340340859937e-06,
+      "loss": 0.3438,
+      "step": 22964
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.373763589699224e-06,
+      "loss": 0.3965,
+      "step": 22965
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.373186838538513e-06,
+      "loss": 0.4785,
+      "step": 22966
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.3726100873778015e-06,
+      "loss": 0.2947,
+      "step": 22967
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.372033336217089e-06,
+      "loss": 0.3652,
+      "step": 22968
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 7.371456585056378e-06,
+      "loss": 0.4395,
+      "step": 22969
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.370879833895666e-06,
+      "loss": 0.4886,
+      "step": 22970
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.370303082734954e-06,
+      "loss": 0.4281,
+      "step": 22971
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.369726331574243e-06,
+      "loss": 0.3719,
+      "step": 22972
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.369149580413532e-06,
+      "loss": 0.455,
+      "step": 22973
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3685728292528195e-06,
+      "loss": 0.4585,
+      "step": 22974
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.367996078092108e-06,
+      "loss": 0.3299,
+      "step": 22975
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3674193269313966e-06,
+      "loss": 0.3273,
+      "step": 22976
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366842575770684e-06,
+      "loss": 0.4839,
+      "step": 22977
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.366265824609973e-06,
+      "loss": 0.4822,
+      "step": 22978
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.365689073449261e-06,
+      "loss": 0.5091,
+      "step": 22979
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.365112322288549e-06,
+      "loss": 0.5014,
+      "step": 22980
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3645355711278375e-06,
+      "loss": 0.3449,
+      "step": 22981
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.363958819967125e-06,
+      "loss": 0.3344,
+      "step": 22982
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.363382068806414e-06,
+      "loss": 0.4225,
+      "step": 22983
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.362805317645702e-06,
+      "loss": 0.5021,
+      "step": 22984
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.36222856648499e-06,
+      "loss": 0.3789,
+      "step": 22985
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3616518153242785e-06,
+      "loss": 0.3928,
+      "step": 22986
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.361075064163568e-06,
+      "loss": 0.3835,
+      "step": 22987
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.360498313002855e-06,
+      "loss": 0.4858,
+      "step": 22988
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.359921561842144e-06,
+      "loss": 0.4327,
+      "step": 22989
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.359344810681433e-06,
+      "loss": 0.4755,
+      "step": 22990
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.35876805952072e-06,
+      "loss": 0.5857,
+      "step": 22991
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.358191308360009e-06,
+      "loss": 0.4057,
+      "step": 22992
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.357614557199297e-06,
+      "loss": 0.4027,
+      "step": 22993
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.357037806038585e-06,
+      "loss": 0.4781,
+      "step": 22994
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.356461054877874e-06,
+      "loss": 0.5573,
+      "step": 22995
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.355884303717162e-06,
+      "loss": 0.4973,
+      "step": 22996
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.35530755255645e-06,
+      "loss": 0.4635,
+      "step": 22997
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.354730801395738e-06,
+      "loss": 0.4204,
+      "step": 22998
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.354154050235026e-06,
+      "loss": 0.4864,
+      "step": 22999
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3535772990743146e-06,
+      "loss": 0.417,
+      "step": 23000
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.353000547913604e-06,
+      "loss": 0.4617,
+      "step": 23001
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.352423796752891e-06,
+      "loss": 0.4204,
+      "step": 23002
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.35184704559218e-06,
+      "loss": 0.364,
+      "step": 23003
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.351270294431469e-06,
+      "loss": 0.3818,
+      "step": 23004
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.350693543270756e-06,
+      "loss": 0.4321,
+      "step": 23005
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.350116792110045e-06,
+      "loss": 0.3426,
+      "step": 23006
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3495400409493334e-06,
+      "loss": 0.486,
+      "step": 23007
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.348963289788621e-06,
+      "loss": 0.4127,
+      "step": 23008
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.34838653862791e-06,
+      "loss": 0.4367,
+      "step": 23009
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.347809787467198e-06,
+      "loss": 0.4374,
+      "step": 23010
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.347233036306486e-06,
+      "loss": 0.5297,
+      "step": 23011
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.346656285145774e-06,
+      "loss": 0.5271,
+      "step": 23012
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.346079533985063e-06,
+      "loss": 0.4648,
+      "step": 23013
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.345502782824351e-06,
+      "loss": 0.4547,
+      "step": 23014
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.344926031663639e-06,
+      "loss": 0.5133,
+      "step": 23015
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.344349280502927e-06,
+      "loss": 0.4482,
+      "step": 23016
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.343772529342215e-06,
+      "loss": 0.5285,
+      "step": 23017
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.343195778181505e-06,
+      "loss": 0.3879,
+      "step": 23018
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.342619027020792e-06,
+      "loss": 0.4345,
+      "step": 23019
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.342042275860081e-06,
+      "loss": 0.4246,
+      "step": 23020
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3414655246993695e-06,
+      "loss": 0.351,
+      "step": 23021
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.340888773538657e-06,
+      "loss": 0.4814,
+      "step": 23022
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.340312022377946e-06,
+      "loss": 0.3664,
+      "step": 23023
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.339735271217234e-06,
+      "loss": 0.4408,
+      "step": 23024
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.339158520056522e-06,
+      "loss": 0.5147,
+      "step": 23025
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3385817688958105e-06,
+      "loss": 0.3869,
+      "step": 23026
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.338005017735099e-06,
+      "loss": 0.4867,
+      "step": 23027
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.337428266574387e-06,
+      "loss": 0.4453,
+      "step": 23028
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.336851515413675e-06,
+      "loss": 0.5122,
+      "step": 23029
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.336274764252964e-06,
+      "loss": 0.4887,
+      "step": 23030
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3356980130922514e-06,
+      "loss": 0.5317,
+      "step": 23031
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.335121261931541e-06,
+      "loss": 0.4067,
+      "step": 23032
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.334544510770829e-06,
+      "loss": 0.3999,
+      "step": 23033
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.333967759610117e-06,
+      "loss": 0.4662,
+      "step": 23034
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3333910084494056e-06,
+      "loss": 0.3617,
+      "step": 23035
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.332814257288693e-06,
+      "loss": 0.4666,
+      "step": 23036
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.332237506127982e-06,
+      "loss": 0.4553,
+      "step": 23037
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.33166075496727e-06,
+      "loss": 0.3956,
+      "step": 23038
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.331084003806558e-06,
+      "loss": 0.5108,
+      "step": 23039
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3305072526458465e-06,
+      "loss": 0.4536,
+      "step": 23040
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.329930501485135e-06,
+      "loss": 0.3455,
+      "step": 23041
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.329353750324423e-06,
+      "loss": 0.4088,
+      "step": 23042
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.328776999163711e-06,
+      "loss": 0.4997,
+      "step": 23043
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.328200248003e-06,
+      "loss": 0.4374,
+      "step": 23044
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3276234968422875e-06,
+      "loss": 0.3591,
+      "step": 23045
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.327046745681576e-06,
+      "loss": 0.5479,
+      "step": 23046
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.326469994520865e-06,
+      "loss": 0.4333,
+      "step": 23047
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325893243360152e-06,
+      "loss": 0.4114,
+      "step": 23048
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.325316492199442e-06,
+      "loss": 0.4469,
+      "step": 23049
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.32473974103873e-06,
+      "loss": 0.5143,
+      "step": 23050
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.324162989878018e-06,
+      "loss": 0.3449,
+      "step": 23051
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.323586238717306e-06,
+      "loss": 0.4788,
+      "step": 23052
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.323009487556594e-06,
+      "loss": 0.3197,
+      "step": 23053
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.322432736395883e-06,
+      "loss": 0.4589,
+      "step": 23054
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.321855985235171e-06,
+      "loss": 0.4818,
+      "step": 23055
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.321279234074459e-06,
+      "loss": 0.3258,
+      "step": 23056
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.320702482913747e-06,
+      "loss": 0.4002,
+      "step": 23057
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.320125731753036e-06,
+      "loss": 0.3834,
+      "step": 23058
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3195489805923236e-06,
+      "loss": 0.3485,
+      "step": 23059
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.318972229431612e-06,
+      "loss": 0.5172,
+      "step": 23060
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.318395478270901e-06,
+      "loss": 0.4601,
+      "step": 23061
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.317818727110188e-06,
+      "loss": 0.3858,
+      "step": 23062
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.317241975949478e-06,
+      "loss": 0.4099,
+      "step": 23063
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.316665224788766e-06,
+      "loss": 0.4577,
+      "step": 23064
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.316088473628054e-06,
+      "loss": 0.4264,
+      "step": 23065
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3155117224673424e-06,
+      "loss": 0.4709,
+      "step": 23066
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.314934971306631e-06,
+      "loss": 0.373,
+      "step": 23067
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.314358220145919e-06,
+      "loss": 0.5072,
+      "step": 23068
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.313781468985207e-06,
+      "loss": 0.3861,
+      "step": 23069
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.313204717824495e-06,
+      "loss": 0.4801,
+      "step": 23070
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.312627966663783e-06,
+      "loss": 0.4418,
+      "step": 23071
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.312051215503072e-06,
+      "loss": 0.4273,
+      "step": 23072
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.31147446434236e-06,
+      "loss": 0.4655,
+      "step": 23073
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.310897713181648e-06,
+      "loss": 0.3293,
+      "step": 23074
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.310320962020937e-06,
+      "loss": 0.4095,
+      "step": 23075
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.309744210860224e-06,
+      "loss": 0.4352,
+      "step": 23076
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.309167459699513e-06,
+      "loss": 0.3417,
+      "step": 23077
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.308590708538802e-06,
+      "loss": 0.506,
+      "step": 23078
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.308013957378089e-06,
+      "loss": 0.4612,
+      "step": 23079
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3074372062173785e-06,
+      "loss": 0.4262,
+      "step": 23080
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.306860455056667e-06,
+      "loss": 0.4933,
+      "step": 23081
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.306283703895955e-06,
+      "loss": 0.4437,
+      "step": 23082
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.305706952735243e-06,
+      "loss": 0.4056,
+      "step": 23083
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.305130201574532e-06,
+      "loss": 0.3926,
+      "step": 23084
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3045534504138195e-06,
+      "loss": 0.4973,
+      "step": 23085
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.303976699253108e-06,
+      "loss": 0.4115,
+      "step": 23086
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.303399948092396e-06,
+      "loss": 0.4687,
+      "step": 23087
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.302823196931684e-06,
+      "loss": 0.4196,
+      "step": 23088
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.302246445770973e-06,
+      "loss": 0.4343,
+      "step": 23089
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3016696946102604e-06,
+      "loss": 0.4345,
+      "step": 23090
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.301092943449549e-06,
+      "loss": 0.4539,
+      "step": 23091
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.3005161922888375e-06,
+      "loss": 0.4449,
+      "step": 23092
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.299939441128125e-06,
+      "loss": 0.5608,
+      "step": 23093
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2993626899674146e-06,
+      "loss": 0.35,
+      "step": 23094
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.298785938806703e-06,
+      "loss": 0.4734,
+      "step": 23095
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.298209187645991e-06,
+      "loss": 0.5299,
+      "step": 23096
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.297632436485279e-06,
+      "loss": 0.334,
+      "step": 23097
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.297055685324568e-06,
+      "loss": 0.4026,
+      "step": 23098
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2964789341638555e-06,
+      "loss": 0.5178,
+      "step": 23099
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.295902183003144e-06,
+      "loss": 0.3981,
+      "step": 23100
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.295325431842433e-06,
+      "loss": 0.3701,
+      "step": 23101
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.29474868068172e-06,
+      "loss": 0.5361,
+      "step": 23102
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.294171929521009e-06,
+      "loss": 0.5052,
+      "step": 23103
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2935951783602965e-06,
+      "loss": 0.3576,
+      "step": 23104
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.293018427199585e-06,
+      "loss": 0.4542,
+      "step": 23105
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2924416760388736e-06,
+      "loss": 0.3993,
+      "step": 23106
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.291864924878161e-06,
+      "loss": 0.458,
+      "step": 23107
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.29128817371745e-06,
+      "loss": 0.5673,
+      "step": 23108
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.290711422556739e-06,
+      "loss": 0.5264,
+      "step": 23109
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.290134671396026e-06,
+      "loss": 0.4898,
+      "step": 23110
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.289557920235315e-06,
+      "loss": 0.4367,
+      "step": 23111
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.288981169074604e-06,
+      "loss": 0.4226,
+      "step": 23112
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.288404417913892e-06,
+      "loss": 0.5171,
+      "step": 23113
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.28782766675318e-06,
+      "loss": 0.3844,
+      "step": 23114
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.287250915592469e-06,
+      "loss": 0.4471,
+      "step": 23115
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.286674164431756e-06,
+      "loss": 0.5173,
+      "step": 23116
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.286097413271045e-06,
+      "loss": 0.3288,
+      "step": 23117
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.285520662110333e-06,
+      "loss": 0.4361,
+      "step": 23118
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.284943910949621e-06,
+      "loss": 0.4573,
+      "step": 23119
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.28436715978891e-06,
+      "loss": 0.4313,
+      "step": 23120
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283790408628197e-06,
+      "loss": 0.4466,
+      "step": 23121
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.283213657467486e-06,
+      "loss": 0.2785,
+      "step": 23122
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.282636906306774e-06,
+      "loss": 0.3603,
+      "step": 23123
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.282060155146062e-06,
+      "loss": 0.3643,
+      "step": 23124
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2814834039853514e-06,
+      "loss": 0.4408,
+      "step": 23125
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.28090665282464e-06,
+      "loss": 0.4683,
+      "step": 23126
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.280329901663928e-06,
+      "loss": 0.4035,
+      "step": 23127
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.279753150503216e-06,
+      "loss": 0.4628,
+      "step": 23128
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.279176399342505e-06,
+      "loss": 0.4052,
+      "step": 23129
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.278599648181792e-06,
+      "loss": 0.4282,
+      "step": 23130
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.278022897021081e-06,
+      "loss": 0.3579,
+      "step": 23131
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2774461458603695e-06,
+      "loss": 0.3916,
+      "step": 23132
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.276869394699657e-06,
+      "loss": 0.3926,
+      "step": 23133
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.276292643538946e-06,
+      "loss": 0.404,
+      "step": 23134
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.275715892378234e-06,
+      "loss": 0.5169,
+      "step": 23135
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.275139141217522e-06,
+      "loss": 0.424,
+      "step": 23136
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2745623900568104e-06,
+      "loss": 0.4083,
+      "step": 23137
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.273985638896098e-06,
+      "loss": 0.3744,
+      "step": 23138
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.273408887735387e-06,
+      "loss": 0.3856,
+      "step": 23139
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.272832136574676e-06,
+      "loss": 0.554,
+      "step": 23140
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.272255385413963e-06,
+      "loss": 0.3681,
+      "step": 23141
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.271678634253252e-06,
+      "loss": 0.4141,
+      "step": 23142
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.271101883092541e-06,
+      "loss": 0.4971,
+      "step": 23143
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2705251319318285e-06,
+      "loss": 0.4558,
+      "step": 23144
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.269948380771117e-06,
+      "loss": 0.5307,
+      "step": 23145
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.2693716296104055e-06,
+      "loss": 0.4109,
+      "step": 23146
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.268794878449693e-06,
+      "loss": 0.5201,
+      "step": 23147
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 7.268218127288982e-06,
+      "loss": 0.4035,
+      "step": 23148
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.26764137612827e-06,
+      "loss": 0.5505,
+      "step": 23149
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.267064624967558e-06,
+      "loss": 0.4704,
+      "step": 23150
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2664878738068465e-06,
+      "loss": 0.4327,
+      "step": 23151
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.265911122646135e-06,
+      "loss": 0.3229,
+      "step": 23152
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.265334371485423e-06,
+      "loss": 0.4552,
+      "step": 23153
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.264757620324711e-06,
+      "loss": 0.3943,
+      "step": 23154
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.264180869164001e-06,
+      "loss": 0.4529,
+      "step": 23155
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263604118003288e-06,
+      "loss": 0.3656,
+      "step": 23156
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.263027366842577e-06,
+      "loss": 0.4136,
+      "step": 23157
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2624506156818645e-06,
+      "loss": 0.4162,
+      "step": 23158
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.261873864521153e-06,
+      "loss": 0.4782,
+      "step": 23159
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.261297113360442e-06,
+      "loss": 0.3377,
+      "step": 23160
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.260720362199729e-06,
+      "loss": 0.4859,
+      "step": 23161
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.260143611039018e-06,
+      "loss": 0.4603,
+      "step": 23162
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.259566859878306e-06,
+      "loss": 0.4074,
+      "step": 23163
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.258990108717594e-06,
+      "loss": 0.4271,
+      "step": 23164
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2584133575568826e-06,
+      "loss": 0.4762,
+      "step": 23165
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.257836606396171e-06,
+      "loss": 0.4077,
+      "step": 23166
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.257259855235459e-06,
+      "loss": 0.4368,
+      "step": 23167
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.256683104074747e-06,
+      "loss": 0.5107,
+      "step": 23168
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.256106352914037e-06,
+      "loss": 0.5611,
+      "step": 23169
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2555296017533235e-06,
+      "loss": 0.3667,
+      "step": 23170
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.254952850592613e-06,
+      "loss": 0.3681,
+      "step": 23171
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2543760994319014e-06,
+      "loss": 0.4269,
+      "step": 23172
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.253799348271189e-06,
+      "loss": 0.3887,
+      "step": 23173
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.253222597110478e-06,
+      "loss": 0.5102,
+      "step": 23174
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.252645845949765e-06,
+      "loss": 0.3418,
+      "step": 23175
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.252069094789054e-06,
+      "loss": 0.4309,
+      "step": 23176
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.251492343628342e-06,
+      "loss": 0.3392,
+      "step": 23177
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.25091559246763e-06,
+      "loss": 0.4111,
+      "step": 23178
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.250338841306919e-06,
+      "loss": 0.4203,
+      "step": 23179
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.249762090146207e-06,
+      "loss": 0.4032,
+      "step": 23180
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.249185338985495e-06,
+      "loss": 0.4676,
+      "step": 23181
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.248608587824783e-06,
+      "loss": 0.4796,
+      "step": 23182
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.248031836664072e-06,
+      "loss": 0.5007,
+      "step": 23183
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.24745508550336e-06,
+      "loss": 0.4018,
+      "step": 23184
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.246878334342649e-06,
+      "loss": 0.3926,
+      "step": 23185
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2463015831819375e-06,
+      "loss": 0.4633,
+      "step": 23186
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.245724832021225e-06,
+      "loss": 0.4499,
+      "step": 23187
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.245148080860514e-06,
+      "loss": 0.4392,
+      "step": 23188
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.244571329699802e-06,
+      "loss": 0.4409,
+      "step": 23189
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.24399457853909e-06,
+      "loss": 0.4559,
+      "step": 23190
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2434178273783785e-06,
+      "loss": 0.4123,
+      "step": 23191
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242841076217666e-06,
+      "loss": 0.4861,
+      "step": 23192
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.242264325056955e-06,
+      "loss": 0.4374,
+      "step": 23193
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.241687573896243e-06,
+      "loss": 0.3902,
+      "step": 23194
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.241110822735531e-06,
+      "loss": 0.3195,
+      "step": 23195
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.240534071574819e-06,
+      "loss": 0.3991,
+      "step": 23196
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.239957320414108e-06,
+      "loss": 0.4255,
+      "step": 23197
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.239380569253396e-06,
+      "loss": 0.5313,
+      "step": 23198
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.238803818092684e-06,
+      "loss": 0.4219,
+      "step": 23199
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2382270669319736e-06,
+      "loss": 0.4636,
+      "step": 23200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.23765031577126e-06,
+      "loss": 0.3781,
+      "step": 23201
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.23707356461055e-06,
+      "loss": 0.4664,
+      "step": 23202
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.236496813449838e-06,
+      "loss": 0.3835,
+      "step": 23203
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.235920062289126e-06,
+      "loss": 0.4923,
+      "step": 23204
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2353433111284145e-06,
+      "loss": 0.4578,
+      "step": 23205
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.234766559967703e-06,
+      "loss": 0.3895,
+      "step": 23206
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.234189808806991e-06,
+      "loss": 0.377,
+      "step": 23207
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.233613057646279e-06,
+      "loss": 0.3619,
+      "step": 23208
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.233036306485567e-06,
+      "loss": 0.3741,
+      "step": 23209
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2324595553248555e-06,
+      "loss": 0.4997,
+      "step": 23210
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.231882804164144e-06,
+      "loss": 0.4496,
+      "step": 23211
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.231306053003432e-06,
+      "loss": 0.3508,
+      "step": 23212
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.23072930184272e-06,
+      "loss": 0.4872,
+      "step": 23213
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.230152550682009e-06,
+      "loss": 0.3696,
+      "step": 23214
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2295757995212965e-06,
+      "loss": 0.5925,
+      "step": 23215
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.228999048360586e-06,
+      "loss": 0.5248,
+      "step": 23216
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.228422297199874e-06,
+      "loss": 0.4902,
+      "step": 23217
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.227845546039162e-06,
+      "loss": 0.4102,
+      "step": 23218
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.227268794878451e-06,
+      "loss": 0.4314,
+      "step": 23219
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.226692043717739e-06,
+      "loss": 0.5349,
+      "step": 23220
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.226115292557027e-06,
+      "loss": 0.3855,
+      "step": 23221
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.225538541396315e-06,
+      "loss": 0.3425,
+      "step": 23222
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.224961790235604e-06,
+      "loss": 0.4596,
+      "step": 23223
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2243850390748915e-06,
+      "loss": 0.3829,
+      "step": 23224
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.22380828791418e-06,
+      "loss": 0.443,
+      "step": 23225
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.223231536753468e-06,
+      "loss": 0.3375,
+      "step": 23226
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222654785592756e-06,
+      "loss": 0.441,
+      "step": 23227
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.222078034432045e-06,
+      "loss": 0.4484,
+      "step": 23228
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2215012832713325e-06,
+      "loss": 0.5172,
+      "step": 23229
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.220924532110621e-06,
+      "loss": 0.4742,
+      "step": 23230
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2203477809499104e-06,
+      "loss": 0.3314,
+      "step": 23231
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.219771029789197e-06,
+      "loss": 0.4444,
+      "step": 23232
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.219194278628487e-06,
+      "loss": 0.4884,
+      "step": 23233
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.218617527467775e-06,
+      "loss": 0.2936,
+      "step": 23234
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.218040776307063e-06,
+      "loss": 0.389,
+      "step": 23235
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.217464025146351e-06,
+      "loss": 0.3817,
+      "step": 23236
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.21688727398564e-06,
+      "loss": 0.4058,
+      "step": 23237
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.216310522824928e-06,
+      "loss": 0.4704,
+      "step": 23238
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.215733771664216e-06,
+      "loss": 0.511,
+      "step": 23239
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.215157020503505e-06,
+      "loss": 0.4387,
+      "step": 23240
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.214580269342792e-06,
+      "loss": 0.3828,
+      "step": 23241
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.214003518182081e-06,
+      "loss": 0.3525,
+      "step": 23242
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2134267670213686e-06,
+      "loss": 0.47,
+      "step": 23243
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.212850015860657e-06,
+      "loss": 0.4955,
+      "step": 23244
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.212273264699946e-06,
+      "loss": 0.4897,
+      "step": 23245
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.211696513539233e-06,
+      "loss": 0.4941,
+      "step": 23246
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.211119762378523e-06,
+      "loss": 0.5039,
+      "step": 23247
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.210543011217811e-06,
+      "loss": 0.4068,
+      "step": 23248
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.209966260057099e-06,
+      "loss": 0.3927,
+      "step": 23249
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2093895088963875e-06,
+      "loss": 0.4317,
+      "step": 23250
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.208812757735676e-06,
+      "loss": 0.3678,
+      "step": 23251
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.208236006574964e-06,
+      "loss": 0.4472,
+      "step": 23252
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.207659255414252e-06,
+      "loss": 0.3385,
+      "step": 23253
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.207082504253541e-06,
+      "loss": 0.5374,
+      "step": 23254
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.206505753092828e-06,
+      "loss": 0.4329,
+      "step": 23255
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.205929001932117e-06,
+      "loss": 0.2675,
+      "step": 23256
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2053522507714055e-06,
+      "loss": 0.4074,
+      "step": 23257
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.204775499610693e-06,
+      "loss": 0.557,
+      "step": 23258
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.204198748449982e-06,
+      "loss": 0.4209,
+      "step": 23259
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.203621997289269e-06,
+      "loss": 0.4835,
+      "step": 23260
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.203045246128558e-06,
+      "loss": 0.4277,
+      "step": 23261
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.202468494967847e-06,
+      "loss": 0.3493,
+      "step": 23262
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.201891743807134e-06,
+      "loss": 0.4309,
+      "step": 23263
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.2013149926464235e-06,
+      "loss": 0.3959,
+      "step": 23264
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.200738241485712e-06,
+      "loss": 0.4657,
+      "step": 23265
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.200161490325e-06,
+      "loss": 0.4371,
+      "step": 23266
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.199584739164288e-06,
+      "loss": 0.346,
+      "step": 23267
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.199007988003577e-06,
+      "loss": 0.358,
+      "step": 23268
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1984312368428645e-06,
+      "loss": 0.4267,
+      "step": 23269
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.197854485682153e-06,
+      "loss": 0.4462,
+      "step": 23270
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1972777345214415e-06,
+      "loss": 0.4961,
+      "step": 23271
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.196700983360729e-06,
+      "loss": 0.4166,
+      "step": 23272
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.196124232200018e-06,
+      "loss": 0.3844,
+      "step": 23273
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.195547481039306e-06,
+      "loss": 0.4407,
+      "step": 23274
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.194970729878594e-06,
+      "loss": 0.3756,
+      "step": 23275
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1943939787178825e-06,
+      "loss": 0.4769,
+      "step": 23276
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.193817227557172e-06,
+      "loss": 0.4486,
+      "step": 23277
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.19324047639646e-06,
+      "loss": 0.4262,
+      "step": 23278
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.192663725235748e-06,
+      "loss": 0.4744,
+      "step": 23279
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.192086974075036e-06,
+      "loss": 0.4489,
+      "step": 23280
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.191510222914324e-06,
+      "loss": 0.5878,
+      "step": 23281
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.190933471753613e-06,
+      "loss": 0.3709,
+      "step": 23282
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1903567205929005e-06,
+      "loss": 0.3614,
+      "step": 23283
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.189779969432189e-06,
+      "loss": 0.5005,
+      "step": 23284
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.189203218271478e-06,
+      "loss": 0.3765,
+      "step": 23285
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.188626467110765e-06,
+      "loss": 0.426,
+      "step": 23286
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.188049715950054e-06,
+      "loss": 0.486,
+      "step": 23287
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.187472964789342e-06,
+      "loss": 0.2733,
+      "step": 23288
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.18689621362863e-06,
+      "loss": 0.4651,
+      "step": 23289
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1863194624679186e-06,
+      "loss": 0.482,
+      "step": 23290
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.185742711307208e-06,
+      "loss": 0.5029,
+      "step": 23291
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.185165960146495e-06,
+      "loss": 0.3257,
+      "step": 23292
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.184589208985784e-06,
+      "loss": 0.3933,
+      "step": 23293
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.184012457825073e-06,
+      "loss": 0.4276,
+      "step": 23294
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.18343570666436e-06,
+      "loss": 0.3905,
+      "step": 23295
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.182858955503649e-06,
+      "loss": 0.5143,
+      "step": 23296
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.182282204342937e-06,
+      "loss": 0.297,
+      "step": 23297
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181705453182225e-06,
+      "loss": 0.4899,
+      "step": 23298
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.181128702021514e-06,
+      "loss": 0.5065,
+      "step": 23299
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.180551950860801e-06,
+      "loss": 0.4682,
+      "step": 23300
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.17997519970009e-06,
+      "loss": 0.3385,
+      "step": 23301
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.179398448539378e-06,
+      "loss": 0.4539,
+      "step": 23302
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.178821697378666e-06,
+      "loss": 0.4369,
+      "step": 23303
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.178244946217955e-06,
+      "loss": 0.4388,
+      "step": 23304
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.177668195057243e-06,
+      "loss": 0.4829,
+      "step": 23305
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.177091443896531e-06,
+      "loss": 0.4216,
+      "step": 23306
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.176514692735819e-06,
+      "loss": 0.5036,
+      "step": 23307
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.175937941575109e-06,
+      "loss": 0.4351,
+      "step": 23308
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1753611904143964e-06,
+      "loss": 0.5296,
+      "step": 23309
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.174784439253685e-06,
+      "loss": 0.4718,
+      "step": 23310
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1742076880929735e-06,
+      "loss": 0.2806,
+      "step": 23311
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.173630936932261e-06,
+      "loss": 0.5014,
+      "step": 23312
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.17305418577155e-06,
+      "loss": 0.3549,
+      "step": 23313
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.172477434610837e-06,
+      "loss": 0.3409,
+      "step": 23314
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.171900683450126e-06,
+      "loss": 0.3622,
+      "step": 23315
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1713239322894145e-06,
+      "loss": 0.3016,
+      "step": 23316
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.170747181128702e-06,
+      "loss": 0.439,
+      "step": 23317
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.170170429967991e-06,
+      "loss": 0.45,
+      "step": 23318
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.169593678807279e-06,
+      "loss": 0.4666,
+      "step": 23319
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.169016927646567e-06,
+      "loss": 0.4414,
+      "step": 23320
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1684401764858554e-06,
+      "loss": 0.4031,
+      "step": 23321
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.167863425325145e-06,
+      "loss": 0.5635,
+      "step": 23322
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.167286674164432e-06,
+      "loss": 0.447,
+      "step": 23323
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.166709923003721e-06,
+      "loss": 0.3226,
+      "step": 23324
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.1661331718430096e-06,
+      "loss": 0.4536,
+      "step": 23325
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 7.165556420682297e-06,
+      "loss": 0.4962,
+      "step": 23326
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.164979669521586e-06,
+      "loss": 0.3769,
+      "step": 23327
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.164402918360874e-06,
+      "loss": 0.4645,
+      "step": 23328
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.163826167200162e-06,
+      "loss": 0.5303,
+      "step": 23329
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1632494160394505e-06,
+      "loss": 0.4302,
+      "step": 23330
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.162672664878738e-06,
+      "loss": 0.4466,
+      "step": 23331
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.162095913718027e-06,
+      "loss": 0.4197,
+      "step": 23332
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.161519162557315e-06,
+      "loss": 0.4563,
+      "step": 23333
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.160942411396603e-06,
+      "loss": 0.4326,
+      "step": 23334
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1603656602358915e-06,
+      "loss": 0.3632,
+      "step": 23335
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.15978890907518e-06,
+      "loss": 0.4016,
+      "step": 23336
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.159212157914468e-06,
+      "loss": 0.5534,
+      "step": 23337
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.158635406753756e-06,
+      "loss": 0.4736,
+      "step": 23338
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.158058655593046e-06,
+      "loss": 0.5243,
+      "step": 23339
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.157481904432333e-06,
+      "loss": 0.4834,
+      "step": 23340
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.156905153271622e-06,
+      "loss": 0.4593,
+      "step": 23341
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.15632840211091e-06,
+      "loss": 0.3833,
+      "step": 23342
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.155751650950198e-06,
+      "loss": 0.354,
+      "step": 23343
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.155174899789487e-06,
+      "loss": 0.3769,
+      "step": 23344
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.154598148628775e-06,
+      "loss": 0.4008,
+      "step": 23345
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.154021397468063e-06,
+      "loss": 0.4449,
+      "step": 23346
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.153444646307351e-06,
+      "loss": 0.3797,
+      "step": 23347
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.152867895146639e-06,
+      "loss": 0.3844,
+      "step": 23348
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1522911439859276e-06,
+      "loss": 0.443,
+      "step": 23349
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.151714392825216e-06,
+      "loss": 0.565,
+      "step": 23350
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.151137641664504e-06,
+      "loss": 0.3959,
+      "step": 23351
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.150560890503792e-06,
+      "loss": 0.4335,
+      "step": 23352
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.149984139343082e-06,
+      "loss": 0.4437,
+      "step": 23353
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1494073881823685e-06,
+      "loss": 0.4103,
+      "step": 23354
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.148830637021658e-06,
+      "loss": 0.5029,
+      "step": 23355
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1482538858609464e-06,
+      "loss": 0.3384,
+      "step": 23356
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.147677134700234e-06,
+      "loss": 0.3932,
+      "step": 23357
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.147100383539523e-06,
+      "loss": 0.4263,
+      "step": 23358
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.146523632378811e-06,
+      "loss": 0.5427,
+      "step": 23359
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.145946881218099e-06,
+      "loss": 0.3405,
+      "step": 23360
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.145370130057387e-06,
+      "loss": 0.464,
+      "step": 23361
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.144793378896676e-06,
+      "loss": 0.494,
+      "step": 23362
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.144216627735964e-06,
+      "loss": 0.482,
+      "step": 23363
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.143639876575252e-06,
+      "loss": 0.3715,
+      "step": 23364
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.14306312541454e-06,
+      "loss": 0.2991,
+      "step": 23365
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.142486374253828e-06,
+      "loss": 0.5719,
+      "step": 23366
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.141909623093117e-06,
+      "loss": 0.5098,
+      "step": 23367
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.141332871932405e-06,
+      "loss": 0.4907,
+      "step": 23368
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.140756120771694e-06,
+      "loss": 0.5218,
+      "step": 23369
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1401793696109825e-06,
+      "loss": 0.4127,
+      "step": 23370
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.13960261845027e-06,
+      "loss": 0.3653,
+      "step": 23371
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.139025867289559e-06,
+      "loss": 0.3691,
+      "step": 23372
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.138449116128847e-06,
+      "loss": 0.3992,
+      "step": 23373
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.137872364968135e-06,
+      "loss": 0.4035,
+      "step": 23374
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1372956138074235e-06,
+      "loss": 0.4509,
+      "step": 23375
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.136718862646712e-06,
+      "loss": 0.3858,
+      "step": 23376
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.136142111486e-06,
+      "loss": 0.3779,
+      "step": 23377
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.135565360325288e-06,
+      "loss": 0.4996,
+      "step": 23378
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.134988609164577e-06,
+      "loss": 0.5896,
+      "step": 23379
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1344118580038644e-06,
+      "loss": 0.4653,
+      "step": 23380
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.133835106843153e-06,
+      "loss": 0.479,
+      "step": 23381
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.133258355682441e-06,
+      "loss": 0.3993,
+      "step": 23382
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.132681604521729e-06,
+      "loss": 0.3879,
+      "step": 23383
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1321048533610186e-06,
+      "loss": 0.3646,
+      "step": 23384
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.131528102200305e-06,
+      "loss": 0.4436,
+      "step": 23385
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.130951351039595e-06,
+      "loss": 0.4549,
+      "step": 23386
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.130374599878883e-06,
+      "loss": 0.3962,
+      "step": 23387
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.129797848718171e-06,
+      "loss": 0.2975,
+      "step": 23388
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1292210975574595e-06,
+      "loss": 0.3057,
+      "step": 23389
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.128644346396748e-06,
+      "loss": 0.4157,
+      "step": 23390
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.128067595236036e-06,
+      "loss": 0.4002,
+      "step": 23391
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.127490844075324e-06,
+      "loss": 0.4433,
+      "step": 23392
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.126914092914613e-06,
+      "loss": 0.4428,
+      "step": 23393
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1263373417539005e-06,
+      "loss": 0.3322,
+      "step": 23394
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.125760590593189e-06,
+      "loss": 0.411,
+      "step": 23395
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1251838394324776e-06,
+      "loss": 0.4188,
+      "step": 23396
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.124607088271765e-06,
+      "loss": 0.4737,
+      "step": 23397
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.124030337111054e-06,
+      "loss": 0.3582,
+      "step": 23398
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1234535859503415e-06,
+      "loss": 0.337,
+      "step": 23399
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.122876834789631e-06,
+      "loss": 0.3576,
+      "step": 23400
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.122300083628919e-06,
+      "loss": 0.4603,
+      "step": 23401
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.121723332468207e-06,
+      "loss": 0.3932,
+      "step": 23402
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.121146581307496e-06,
+      "loss": 0.4117,
+      "step": 23403
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.120569830146784e-06,
+      "loss": 0.4845,
+      "step": 23404
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.119993078986072e-06,
+      "loss": 0.4053,
+      "step": 23405
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.11941632782536e-06,
+      "loss": 0.3271,
+      "step": 23406
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.118839576664649e-06,
+      "loss": 0.4857,
+      "step": 23407
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1182628255039366e-06,
+      "loss": 0.4432,
+      "step": 23408
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.117686074343225e-06,
+      "loss": 0.2926,
+      "step": 23409
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.117109323182514e-06,
+      "loss": 0.3578,
+      "step": 23410
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.116532572021801e-06,
+      "loss": 0.4463,
+      "step": 23411
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.11595582086109e-06,
+      "loss": 0.5425,
+      "step": 23412
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.115379069700379e-06,
+      "loss": 0.4893,
+      "step": 23413
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.114802318539666e-06,
+      "loss": 0.3351,
+      "step": 23414
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1142255673789554e-06,
+      "loss": 0.4828,
+      "step": 23415
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.113648816218244e-06,
+      "loss": 0.4602,
+      "step": 23416
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.113072065057532e-06,
+      "loss": 0.3949,
+      "step": 23417
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.11249531389682e-06,
+      "loss": 0.3751,
+      "step": 23418
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.111918562736108e-06,
+      "loss": 0.4633,
+      "step": 23419
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.111341811575396e-06,
+      "loss": 0.4055,
+      "step": 23420
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.110765060414685e-06,
+      "loss": 0.6202,
+      "step": 23421
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.110188309253973e-06,
+      "loss": 0.4085,
+      "step": 23422
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.109611558093261e-06,
+      "loss": 0.5423,
+      "step": 23423
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.10903480693255e-06,
+      "loss": 0.3848,
+      "step": 23424
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.108458055771837e-06,
+      "loss": 0.4226,
+      "step": 23425
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.107881304611126e-06,
+      "loss": 0.4621,
+      "step": 23426
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1073045534504144e-06,
+      "loss": 0.2952,
+      "step": 23427
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.106727802289702e-06,
+      "loss": 0.3902,
+      "step": 23428
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.106151051128991e-06,
+      "loss": 0.3987,
+      "step": 23429
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.10557429996828e-06,
+      "loss": 0.4728,
+      "step": 23430
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.104997548807568e-06,
+      "loss": 0.4372,
+      "step": 23431
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.104420797646856e-06,
+      "loss": 0.4193,
+      "step": 23432
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.103844046486145e-06,
+      "loss": 0.4255,
+      "step": 23433
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1032672953254325e-06,
+      "loss": 0.3771,
+      "step": 23434
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.102690544164721e-06,
+      "loss": 0.4711,
+      "step": 23435
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.102113793004009e-06,
+      "loss": 0.4947,
+      "step": 23436
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.101537041843297e-06,
+      "loss": 0.4399,
+      "step": 23437
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.100960290682586e-06,
+      "loss": 0.4873,
+      "step": 23438
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.1003835395218734e-06,
+      "loss": 0.5011,
+      "step": 23439
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.099806788361162e-06,
+      "loss": 0.4391,
+      "step": 23440
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0992300372004505e-06,
+      "loss": 0.372,
+      "step": 23441
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098653286039738e-06,
+      "loss": 0.38,
+      "step": 23442
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.098076534879027e-06,
+      "loss": 0.5331,
+      "step": 23443
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.097499783718316e-06,
+      "loss": 0.3332,
+      "step": 23444
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.096923032557603e-06,
+      "loss": 0.3758,
+      "step": 23445
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.096346281396892e-06,
+      "loss": 0.386,
+      "step": 23446
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.095769530236181e-06,
+      "loss": 0.4164,
+      "step": 23447
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0951927790754685e-06,
+      "loss": 0.3667,
+      "step": 23448
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.094616027914757e-06,
+      "loss": 0.3607,
+      "step": 23449
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.094039276754046e-06,
+      "loss": 0.3286,
+      "step": 23450
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.093462525593333e-06,
+      "loss": 0.4113,
+      "step": 23451
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.092885774432622e-06,
+      "loss": 0.3911,
+      "step": 23452
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0923090232719095e-06,
+      "loss": 0.5028,
+      "step": 23453
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.091732272111198e-06,
+      "loss": 0.3578,
+      "step": 23454
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0911555209504866e-06,
+      "loss": 0.5437,
+      "step": 23455
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.090578769789774e-06,
+      "loss": 0.4644,
+      "step": 23456
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.090002018629063e-06,
+      "loss": 0.4502,
+      "step": 23457
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.089425267468351e-06,
+      "loss": 0.4478,
+      "step": 23458
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.088848516307639e-06,
+      "loss": 0.4978,
+      "step": 23459
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0882717651469275e-06,
+      "loss": 0.2655,
+      "step": 23460
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.087695013986217e-06,
+      "loss": 0.5347,
+      "step": 23461
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.087118262825505e-06,
+      "loss": 0.5332,
+      "step": 23462
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.086541511664793e-06,
+      "loss": 0.5024,
+      "step": 23463
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.085964760504082e-06,
+      "loss": 0.6142,
+      "step": 23464
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.085388009343369e-06,
+      "loss": 0.4226,
+      "step": 23465
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.084811258182658e-06,
+      "loss": 0.3749,
+      "step": 23466
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.084234507021946e-06,
+      "loss": 0.3618,
+      "step": 23467
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.083657755861234e-06,
+      "loss": 0.5092,
+      "step": 23468
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.083081004700523e-06,
+      "loss": 0.3846,
+      "step": 23469
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.08250425353981e-06,
+      "loss": 0.4677,
+      "step": 23470
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.081927502379099e-06,
+      "loss": 0.4637,
+      "step": 23471
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.081350751218387e-06,
+      "loss": 0.3361,
+      "step": 23472
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.080774000057675e-06,
+      "loss": 0.3872,
+      "step": 23473
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.080197248896964e-06,
+      "loss": 0.3917,
+      "step": 23474
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.079620497736253e-06,
+      "loss": 0.4824,
+      "step": 23475
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.07904374657554e-06,
+      "loss": 0.4147,
+      "step": 23476
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.078466995414829e-06,
+      "loss": 0.5526,
+      "step": 23477
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.077890244254118e-06,
+      "loss": 0.352,
+      "step": 23478
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.077313493093405e-06,
+      "loss": 0.3569,
+      "step": 23479
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.076736741932694e-06,
+      "loss": 0.4065,
+      "step": 23480
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0761599907719825e-06,
+      "loss": 0.4935,
+      "step": 23481
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.07558323961127e-06,
+      "loss": 0.4034,
+      "step": 23482
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.075006488450559e-06,
+      "loss": 0.4264,
+      "step": 23483
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.074429737289847e-06,
+      "loss": 0.3973,
+      "step": 23484
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.073852986129135e-06,
+      "loss": 0.4794,
+      "step": 23485
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0732762349684234e-06,
+      "loss": 0.3441,
+      "step": 23486
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.072699483807711e-06,
+      "loss": 0.3534,
+      "step": 23487
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.072122732647e-06,
+      "loss": 0.3534,
+      "step": 23488
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.071545981486288e-06,
+      "loss": 0.3832,
+      "step": 23489
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.070969230325576e-06,
+      "loss": 0.4055,
+      "step": 23490
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.070392479164864e-06,
+      "loss": 0.454,
+      "step": 23491
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.069815728004154e-06,
+      "loss": 0.5581,
+      "step": 23492
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0692389768434415e-06,
+      "loss": 0.4635,
+      "step": 23493
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.06866222568273e-06,
+      "loss": 0.4299,
+      "step": 23494
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0680854745220185e-06,
+      "loss": 0.3767,
+      "step": 23495
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.067508723361306e-06,
+      "loss": 0.5065,
+      "step": 23496
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.066931972200595e-06,
+      "loss": 0.4977,
+      "step": 23497
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.066355221039883e-06,
+      "loss": 0.4964,
+      "step": 23498
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.065778469879171e-06,
+      "loss": 0.4087,
+      "step": 23499
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0652017187184595e-06,
+      "loss": 0.3984,
+      "step": 23500
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.064624967557748e-06,
+      "loss": 0.529,
+      "step": 23501
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.064048216397036e-06,
+      "loss": 0.4478,
+      "step": 23502
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.063471465236324e-06,
+      "loss": 0.3973,
+      "step": 23503
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.062894714075612e-06,
+      "loss": 0.3907,
+      "step": 23504
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 7.0623179629149005e-06,
+      "loss": 0.4396,
+      "step": 23505
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.06174121175419e-06,
+      "loss": 0.5497,
+      "step": 23506
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.061164460593477e-06,
+      "loss": 0.3912,
+      "step": 23507
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.060587709432766e-06,
+      "loss": 0.3522,
+      "step": 23508
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.060010958272055e-06,
+      "loss": 0.488,
+      "step": 23509
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.059434207111342e-06,
+      "loss": 0.4388,
+      "step": 23510
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.058857455950631e-06,
+      "loss": 0.5056,
+      "step": 23511
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.058280704789919e-06,
+      "loss": 0.4778,
+      "step": 23512
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.057703953629207e-06,
+      "loss": 0.5447,
+      "step": 23513
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0571272024684956e-06,
+      "loss": 0.445,
+      "step": 23514
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.056550451307784e-06,
+      "loss": 0.4695,
+      "step": 23515
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.055973700147072e-06,
+      "loss": 0.4183,
+      "step": 23516
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.05539694898636e-06,
+      "loss": 0.4371,
+      "step": 23517
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.054820197825649e-06,
+      "loss": 0.4435,
+      "step": 23518
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0542434466649365e-06,
+      "loss": 0.5069,
+      "step": 23519
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.053666695504225e-06,
+      "loss": 0.461,
+      "step": 23520
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.053089944343513e-06,
+      "loss": 0.5259,
+      "step": 23521
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.052513193182801e-06,
+      "loss": 0.4741,
+      "step": 23522
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.051936442022091e-06,
+      "loss": 0.4985,
+      "step": 23523
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.051359690861378e-06,
+      "loss": 0.4984,
+      "step": 23524
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.050782939700667e-06,
+      "loss": 0.4363,
+      "step": 23525
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.050206188539955e-06,
+      "loss": 0.3518,
+      "step": 23526
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.049629437379243e-06,
+      "loss": 0.4756,
+      "step": 23527
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.049052686218532e-06,
+      "loss": 0.3617,
+      "step": 23528
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.04847593505782e-06,
+      "loss": 0.4061,
+      "step": 23529
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.047899183897108e-06,
+      "loss": 0.3592,
+      "step": 23530
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.047322432736396e-06,
+      "loss": 0.3874,
+      "step": 23531
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.046745681575685e-06,
+      "loss": 0.345,
+      "step": 23532
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.046168930414973e-06,
+      "loss": 0.4956,
+      "step": 23533
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.045592179254261e-06,
+      "loss": 0.4132,
+      "step": 23534
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0450154280935505e-06,
+      "loss": 0.4974,
+      "step": 23535
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.044438676932837e-06,
+      "loss": 0.4391,
+      "step": 23536
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.043861925772127e-06,
+      "loss": 0.399,
+      "step": 23537
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.043285174611415e-06,
+      "loss": 0.4236,
+      "step": 23538
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.042708423450703e-06,
+      "loss": 0.364,
+      "step": 23539
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0421316722899915e-06,
+      "loss": 0.2939,
+      "step": 23540
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.041554921129279e-06,
+      "loss": 0.4742,
+      "step": 23541
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.040978169968568e-06,
+      "loss": 0.3977,
+      "step": 23542
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.040401418807856e-06,
+      "loss": 0.4177,
+      "step": 23543
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.039824667647144e-06,
+      "loss": 0.4771,
+      "step": 23544
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0392479164864324e-06,
+      "loss": 0.3799,
+      "step": 23545
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.038671165325721e-06,
+      "loss": 0.3431,
+      "step": 23546
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.038094414165009e-06,
+      "loss": 0.414,
+      "step": 23547
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.037517663004297e-06,
+      "loss": 0.3535,
+      "step": 23548
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.036940911843586e-06,
+      "loss": 0.442,
+      "step": 23549
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.036364160682873e-06,
+      "loss": 0.3764,
+      "step": 23550
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.035787409522162e-06,
+      "loss": 0.3285,
+      "step": 23551
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.035210658361451e-06,
+      "loss": 0.4167,
+      "step": 23552
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.034633907200739e-06,
+      "loss": 0.4034,
+      "step": 23553
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0340571560400275e-06,
+      "loss": 0.3652,
+      "step": 23554
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.033480404879316e-06,
+      "loss": 0.4567,
+      "step": 23555
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.032903653718604e-06,
+      "loss": 0.3718,
+      "step": 23556
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.032326902557892e-06,
+      "loss": 0.4943,
+      "step": 23557
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.03175015139718e-06,
+      "loss": 0.4593,
+      "step": 23558
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0311734002364685e-06,
+      "loss": 0.3961,
+      "step": 23559
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.030596649075757e-06,
+      "loss": 0.4311,
+      "step": 23560
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.030019897915045e-06,
+      "loss": 0.3564,
+      "step": 23561
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.029443146754333e-06,
+      "loss": 0.4102,
+      "step": 23562
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.028866395593622e-06,
+      "loss": 0.4354,
+      "step": 23563
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0282896444329095e-06,
+      "loss": 0.3447,
+      "step": 23564
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.027712893272198e-06,
+      "loss": 0.5101,
+      "step": 23565
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.027136142111487e-06,
+      "loss": 0.4889,
+      "step": 23566
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.026559390950774e-06,
+      "loss": 0.4482,
+      "step": 23567
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.025982639790064e-06,
+      "loss": 0.3687,
+      "step": 23568
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.025405888629352e-06,
+      "loss": 0.3742,
+      "step": 23569
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.02482913746864e-06,
+      "loss": 0.4166,
+      "step": 23570
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.024252386307928e-06,
+      "loss": 0.3809,
+      "step": 23571
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.023675635147217e-06,
+      "loss": 0.3076,
+      "step": 23572
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0230988839865046e-06,
+      "loss": 0.3505,
+      "step": 23573
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.022522132825793e-06,
+      "loss": 0.5376,
+      "step": 23574
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.021945381665081e-06,
+      "loss": 0.397,
+      "step": 23575
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.021368630504369e-06,
+      "loss": 0.5286,
+      "step": 23576
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.020791879343658e-06,
+      "loss": 0.3345,
+      "step": 23577
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0202151281829455e-06,
+      "loss": 0.4512,
+      "step": 23578
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.019638377022234e-06,
+      "loss": 0.3856,
+      "step": 23579
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.019061625861523e-06,
+      "loss": 0.427,
+      "step": 23580
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.01848487470081e-06,
+      "loss": 0.4415,
+      "step": 23581
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.017908123540099e-06,
+      "loss": 0.3573,
+      "step": 23582
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.017331372379388e-06,
+      "loss": 0.3004,
+      "step": 23583
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016754621218676e-06,
+      "loss": 0.4619,
+      "step": 23584
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.016177870057964e-06,
+      "loss": 0.4748,
+      "step": 23585
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.015601118897253e-06,
+      "loss": 0.3512,
+      "step": 23586
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.015024367736541e-06,
+      "loss": 0.411,
+      "step": 23587
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.014447616575829e-06,
+      "loss": 0.4603,
+      "step": 23588
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.013870865415118e-06,
+      "loss": 0.3798,
+      "step": 23589
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.013294114254405e-06,
+      "loss": 0.3597,
+      "step": 23590
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.012717363093694e-06,
+      "loss": 0.5065,
+      "step": 23591
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.012140611932982e-06,
+      "loss": 0.4738,
+      "step": 23592
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.01156386077227e-06,
+      "loss": 0.3427,
+      "step": 23593
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.010987109611559e-06,
+      "loss": 0.4559,
+      "step": 23594
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.010410358450846e-06,
+      "loss": 0.4359,
+      "step": 23595
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.009833607290135e-06,
+      "loss": 0.4766,
+      "step": 23596
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.009256856129424e-06,
+      "loss": 0.4457,
+      "step": 23597
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.008680104968711e-06,
+      "loss": 0.4413,
+      "step": 23598
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0081033538080005e-06,
+      "loss": 0.2683,
+      "step": 23599
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.007526602647289e-06,
+      "loss": 0.5509,
+      "step": 23600
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.006949851486577e-06,
+      "loss": 0.5195,
+      "step": 23601
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.006373100325865e-06,
+      "loss": 0.4764,
+      "step": 23602
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.005796349165154e-06,
+      "loss": 0.2949,
+      "step": 23603
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0052195980044414e-06,
+      "loss": 0.3563,
+      "step": 23604
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.00464284684373e-06,
+      "loss": 0.3877,
+      "step": 23605
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0040660956830185e-06,
+      "loss": 0.4544,
+      "step": 23606
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.003489344522306e-06,
+      "loss": 0.4397,
+      "step": 23607
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.002912593361595e-06,
+      "loss": 0.4388,
+      "step": 23608
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.002335842200882e-06,
+      "loss": 0.5036,
+      "step": 23609
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.001759091040171e-06,
+      "loss": 0.4033,
+      "step": 23610
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.0011823398794595e-06,
+      "loss": 0.4711,
+      "step": 23611
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.000605588718747e-06,
+      "loss": 0.3941,
+      "step": 23612
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 7.000028837558036e-06,
+      "loss": 0.4397,
+      "step": 23613
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.999452086397325e-06,
+      "loss": 0.4516,
+      "step": 23614
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.998875335236613e-06,
+      "loss": 0.4438,
+      "step": 23615
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.998298584075901e-06,
+      "loss": 0.3845,
+      "step": 23616
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.99772183291519e-06,
+      "loss": 0.3908,
+      "step": 23617
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9971450817544775e-06,
+      "loss": 0.493,
+      "step": 23618
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.996568330593766e-06,
+      "loss": 0.4021,
+      "step": 23619
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9959915794330546e-06,
+      "loss": 0.4435,
+      "step": 23620
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.995414828272342e-06,
+      "loss": 0.3735,
+      "step": 23621
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.994838077111631e-06,
+      "loss": 0.5365,
+      "step": 23622
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.994261325950919e-06,
+      "loss": 0.4256,
+      "step": 23623
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.993684574790207e-06,
+      "loss": 0.5059,
+      "step": 23624
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9931078236294955e-06,
+      "loss": 0.3371,
+      "step": 23625
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.992531072468783e-06,
+      "loss": 0.3852,
+      "step": 23626
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.991954321308072e-06,
+      "loss": 0.4054,
+      "step": 23627
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.991377570147361e-06,
+      "loss": 0.3919,
+      "step": 23628
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.990800818986648e-06,
+      "loss": 0.4081,
+      "step": 23629
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.990224067825937e-06,
+      "loss": 0.4275,
+      "step": 23630
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.989647316665226e-06,
+      "loss": 0.4006,
+      "step": 23631
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9890705655045136e-06,
+      "loss": 0.4644,
+      "step": 23632
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.988493814343802e-06,
+      "loss": 0.4172,
+      "step": 23633
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.987917063183091e-06,
+      "loss": 0.3205,
+      "step": 23634
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.987340312022378e-06,
+      "loss": 0.4857,
+      "step": 23635
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.986763560861667e-06,
+      "loss": 0.4451,
+      "step": 23636
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.986186809700955e-06,
+      "loss": 0.4061,
+      "step": 23637
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.985610058540243e-06,
+      "loss": 0.3655,
+      "step": 23638
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.985033307379532e-06,
+      "loss": 0.3754,
+      "step": 23639
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.98445655621882e-06,
+      "loss": 0.4481,
+      "step": 23640
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.983879805058108e-06,
+      "loss": 0.412,
+      "step": 23641
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.983303053897396e-06,
+      "loss": 0.4646,
+      "step": 23642
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.982726302736684e-06,
+      "loss": 0.4712,
+      "step": 23643
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9821495515759726e-06,
+      "loss": 0.471,
+      "step": 23644
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.981572800415262e-06,
+      "loss": 0.3888,
+      "step": 23645
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.98099604925455e-06,
+      "loss": 0.3243,
+      "step": 23646
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.980419298093838e-06,
+      "loss": 0.4367,
+      "step": 23647
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.979842546933127e-06,
+      "loss": 0.5231,
+      "step": 23648
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.979265795772414e-06,
+      "loss": 0.5485,
+      "step": 23649
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.978689044611703e-06,
+      "loss": 0.51,
+      "step": 23650
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9781122934509914e-06,
+      "loss": 0.4431,
+      "step": 23651
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.977535542290279e-06,
+      "loss": 0.4159,
+      "step": 23652
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.976958791129568e-06,
+      "loss": 0.3823,
+      "step": 23653
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.976382039968856e-06,
+      "loss": 0.3872,
+      "step": 23654
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975805288808144e-06,
+      "loss": 0.3322,
+      "step": 23655
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.975228537647432e-06,
+      "loss": 0.421,
+      "step": 23656
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.974651786486722e-06,
+      "loss": 0.4481,
+      "step": 23657
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.974075035326009e-06,
+      "loss": 0.3907,
+      "step": 23658
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.973498284165298e-06,
+      "loss": 0.3913,
+      "step": 23659
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9729215330045865e-06,
+      "loss": 0.4498,
+      "step": 23660
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.972344781843874e-06,
+      "loss": 0.4133,
+      "step": 23661
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.971768030683163e-06,
+      "loss": 0.3633,
+      "step": 23662
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9711912795224504e-06,
+      "loss": 0.4071,
+      "step": 23663
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.970614528361739e-06,
+      "loss": 0.4676,
+      "step": 23664
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9700377772010275e-06,
+      "loss": 0.566,
+      "step": 23665
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.969461026040315e-06,
+      "loss": 0.3237,
+      "step": 23666
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.968884274879604e-06,
+      "loss": 0.3429,
+      "step": 23667
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.968307523718892e-06,
+      "loss": 0.3984,
+      "step": 23668
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.96773077255818e-06,
+      "loss": 0.4971,
+      "step": 23669
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9671540213974685e-06,
+      "loss": 0.3375,
+      "step": 23670
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.966577270236757e-06,
+      "loss": 0.449,
+      "step": 23671
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.966000519076045e-06,
+      "loss": 0.4094,
+      "step": 23672
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.965423767915333e-06,
+      "loss": 0.5224,
+      "step": 23673
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.964847016754623e-06,
+      "loss": 0.4151,
+      "step": 23674
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9642702655939094e-06,
+      "loss": 0.4231,
+      "step": 23675
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.963693514433199e-06,
+      "loss": 0.4559,
+      "step": 23676
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.963116763272487e-06,
+      "loss": 0.398,
+      "step": 23677
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.962540012111775e-06,
+      "loss": 0.4532,
+      "step": 23678
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9619632609510636e-06,
+      "loss": 0.4794,
+      "step": 23679
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.961386509790351e-06,
+      "loss": 0.3872,
+      "step": 23680
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.96080975862964e-06,
+      "loss": 0.355,
+      "step": 23681
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.960233007468928e-06,
+      "loss": 0.3407,
+      "step": 23682
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.959656256308216e-06,
+      "loss": 0.4564,
+      "step": 23683
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 6.9590795051475045e-06,
+      "loss": 0.3832,
+      "step": 23684
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.958502753986793e-06,
+      "loss": 0.5062,
+      "step": 23685
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.957926002826081e-06,
+      "loss": 0.4981,
+      "step": 23686
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.957349251665369e-06,
+      "loss": 0.4401,
+      "step": 23687
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.956772500504659e-06,
+      "loss": 0.4418,
+      "step": 23688
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9561957493439455e-06,
+      "loss": 0.3851,
+      "step": 23689
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.955618998183235e-06,
+      "loss": 0.4018,
+      "step": 23690
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.955042247022523e-06,
+      "loss": 0.3573,
+      "step": 23691
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.954465495861811e-06,
+      "loss": 0.4106,
+      "step": 23692
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9538887447011e-06,
+      "loss": 0.5354,
+      "step": 23693
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.953311993540388e-06,
+      "loss": 0.3998,
+      "step": 23694
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.952735242379676e-06,
+      "loss": 0.249,
+      "step": 23695
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.952158491218964e-06,
+      "loss": 0.4712,
+      "step": 23696
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.951581740058252e-06,
+      "loss": 0.4325,
+      "step": 23697
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.951004988897541e-06,
+      "loss": 0.3865,
+      "step": 23698
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.950428237736829e-06,
+      "loss": 0.3798,
+      "step": 23699
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.949851486576117e-06,
+      "loss": 0.3842,
+      "step": 23700
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.949274735415405e-06,
+      "loss": 0.4481,
+      "step": 23701
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.948697984254694e-06,
+      "loss": 0.4873,
+      "step": 23702
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9481212330939815e-06,
+      "loss": 0.423,
+      "step": 23703
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.94754448193327e-06,
+      "loss": 0.4585,
+      "step": 23704
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9469677307725595e-06,
+      "loss": 0.4792,
+      "step": 23705
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.946390979611846e-06,
+      "loss": 0.3729,
+      "step": 23706
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.945814228451136e-06,
+      "loss": 0.4737,
+      "step": 23707
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.945237477290424e-06,
+      "loss": 0.3881,
+      "step": 23708
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.944660726129712e-06,
+      "loss": 0.3764,
+      "step": 23709
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9440839749690004e-06,
+      "loss": 0.4667,
+      "step": 23710
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.943507223808289e-06,
+      "loss": 0.4536,
+      "step": 23711
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.942930472647577e-06,
+      "loss": 0.3388,
+      "step": 23712
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.942353721486865e-06,
+      "loss": 0.5144,
+      "step": 23713
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.941776970326153e-06,
+      "loss": 0.4744,
+      "step": 23714
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.941200219165441e-06,
+      "loss": 0.4282,
+      "step": 23715
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.94062346800473e-06,
+      "loss": 0.3706,
+      "step": 23716
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.940046716844018e-06,
+      "loss": 0.4297,
+      "step": 23717
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.939469965683306e-06,
+      "loss": 0.4633,
+      "step": 23718
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9388932145225955e-06,
+      "loss": 0.3593,
+      "step": 23719
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.938316463361882e-06,
+      "loss": 0.3534,
+      "step": 23720
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.937739712201172e-06,
+      "loss": 0.3643,
+      "step": 23721
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.93716296104046e-06,
+      "loss": 0.3957,
+      "step": 23722
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.936586209879748e-06,
+      "loss": 0.333,
+      "step": 23723
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9360094587190365e-06,
+      "loss": 0.4332,
+      "step": 23724
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.935432707558325e-06,
+      "loss": 0.3384,
+      "step": 23725
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934855956397613e-06,
+      "loss": 0.3912,
+      "step": 23726
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.934279205236901e-06,
+      "loss": 0.3262,
+      "step": 23727
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.93370245407619e-06,
+      "loss": 0.354,
+      "step": 23728
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9331257029154775e-06,
+      "loss": 0.475,
+      "step": 23729
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.932548951754766e-06,
+      "loss": 0.3369,
+      "step": 23730
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.931972200594054e-06,
+      "loss": 0.5233,
+      "step": 23731
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.931395449433342e-06,
+      "loss": 0.3818,
+      "step": 23732
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.930818698272631e-06,
+      "loss": 0.5132,
+      "step": 23733
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.930241947111918e-06,
+      "loss": 0.3907,
+      "step": 23734
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.929665195951207e-06,
+      "loss": 0.4535,
+      "step": 23735
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.929088444790496e-06,
+      "loss": 0.52,
+      "step": 23736
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.928511693629784e-06,
+      "loss": 0.3558,
+      "step": 23737
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9279349424690726e-06,
+      "loss": 0.5052,
+      "step": 23738
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.927358191308361e-06,
+      "loss": 0.3863,
+      "step": 23739
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.926781440147649e-06,
+      "loss": 0.4686,
+      "step": 23740
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.926204688986937e-06,
+      "loss": 0.5348,
+      "step": 23741
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.925627937826226e-06,
+      "loss": 0.4202,
+      "step": 23742
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9250511866655135e-06,
+      "loss": 0.3433,
+      "step": 23743
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.924474435504802e-06,
+      "loss": 0.4606,
+      "step": 23744
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.923897684344091e-06,
+      "loss": 0.4885,
+      "step": 23745
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.923320933183378e-06,
+      "loss": 0.5048,
+      "step": 23746
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.922744182022667e-06,
+      "loss": 0.384,
+      "step": 23747
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9221674308619545e-06,
+      "loss": 0.5491,
+      "step": 23748
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.921590679701243e-06,
+      "loss": 0.414,
+      "step": 23749
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.921013928540532e-06,
+      "loss": 0.4074,
+      "step": 23750
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.920437177379819e-06,
+      "loss": 0.4681,
+      "step": 23751
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.919860426219109e-06,
+      "loss": 0.3377,
+      "step": 23752
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.919283675058397e-06,
+      "loss": 0.3242,
+      "step": 23753
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.918706923897685e-06,
+      "loss": 0.4281,
+      "step": 23754
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.918130172736973e-06,
+      "loss": 0.4825,
+      "step": 23755
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.917553421576262e-06,
+      "loss": 0.4935,
+      "step": 23756
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.91697667041555e-06,
+      "loss": 0.408,
+      "step": 23757
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.916399919254838e-06,
+      "loss": 0.333,
+      "step": 23758
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.915823168094127e-06,
+      "loss": 0.446,
+      "step": 23759
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.915246416933414e-06,
+      "loss": 0.3957,
+      "step": 23760
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.914669665772703e-06,
+      "loss": 0.322,
+      "step": 23761
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.914092914611991e-06,
+      "loss": 0.3846,
+      "step": 23762
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.913516163451279e-06,
+      "loss": 0.5211,
+      "step": 23763
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.912939412290568e-06,
+      "loss": 0.4135,
+      "step": 23764
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.912362661129855e-06,
+      "loss": 0.4579,
+      "step": 23765
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.911785909969144e-06,
+      "loss": 0.4965,
+      "step": 23766
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.911209158808433e-06,
+      "loss": 0.4862,
+      "step": 23767
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.910632407647721e-06,
+      "loss": 0.3327,
+      "step": 23768
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9100556564870094e-06,
+      "loss": 0.4295,
+      "step": 23769
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.909478905326298e-06,
+      "loss": 0.4064,
+      "step": 23770
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.908902154165586e-06,
+      "loss": 0.4988,
+      "step": 23771
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.908325403004874e-06,
+      "loss": 0.3073,
+      "step": 23772
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.907748651844163e-06,
+      "loss": 0.3697,
+      "step": 23773
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.90717190068345e-06,
+      "loss": 0.3887,
+      "step": 23774
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.906595149522739e-06,
+      "loss": 0.3663,
+      "step": 23775
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9060183983620275e-06,
+      "loss": 0.4618,
+      "step": 23776
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.905441647201315e-06,
+      "loss": 0.3305,
+      "step": 23777
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.904864896040604e-06,
+      "loss": 0.4067,
+      "step": 23778
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.904288144879893e-06,
+      "loss": 0.4047,
+      "step": 23779
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.90371139371918e-06,
+      "loss": 0.461,
+      "step": 23780
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.903134642558469e-06,
+      "loss": 0.6047,
+      "step": 23781
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.902557891397756e-06,
+      "loss": 0.4724,
+      "step": 23782
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.9019811402370455e-06,
+      "loss": 0.3763,
+      "step": 23783
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.901404389076334e-06,
+      "loss": 0.4448,
+      "step": 23784
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.900827637915622e-06,
+      "loss": 0.4782,
+      "step": 23785
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.90025088675491e-06,
+      "loss": 0.4236,
+      "step": 23786
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.899674135594199e-06,
+      "loss": 0.4618,
+      "step": 23787
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8990973844334865e-06,
+      "loss": 0.3622,
+      "step": 23788
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.898520633272775e-06,
+      "loss": 0.527,
+      "step": 23789
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8979438821120635e-06,
+      "loss": 0.4904,
+      "step": 23790
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.897367130951351e-06,
+      "loss": 0.4041,
+      "step": 23791
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.89679037979064e-06,
+      "loss": 0.4737,
+      "step": 23792
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.896213628629928e-06,
+      "loss": 0.3832,
+      "step": 23793
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.895636877469216e-06,
+      "loss": 0.4767,
+      "step": 23794
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8950601263085045e-06,
+      "loss": 0.3823,
+      "step": 23795
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.894483375147794e-06,
+      "loss": 0.3256,
+      "step": 23796
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.893906623987081e-06,
+      "loss": 0.414,
+      "step": 23797
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.89332987282637e-06,
+      "loss": 0.3656,
+      "step": 23798
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.892753121665659e-06,
+      "loss": 0.4612,
+      "step": 23799
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.892176370504946e-06,
+      "loss": 0.4654,
+      "step": 23800
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.891599619344235e-06,
+      "loss": 0.409,
+      "step": 23801
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8910228681835225e-06,
+      "loss": 0.4147,
+      "step": 23802
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.890446117022811e-06,
+      "loss": 0.5506,
+      "step": 23803
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8898693658621e-06,
+      "loss": 0.4502,
+      "step": 23804
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.889292614701387e-06,
+      "loss": 0.4653,
+      "step": 23805
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.888715863540676e-06,
+      "loss": 0.4247,
+      "step": 23806
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.888139112379964e-06,
+      "loss": 0.453,
+      "step": 23807
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.887562361219252e-06,
+      "loss": 0.3858,
+      "step": 23808
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8869856100585405e-06,
+      "loss": 0.3967,
+      "step": 23809
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.88640885889783e-06,
+      "loss": 0.4154,
+      "step": 23810
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.885832107737117e-06,
+      "loss": 0.4385,
+      "step": 23811
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.885255356576406e-06,
+      "loss": 0.4108,
+      "step": 23812
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.884678605415695e-06,
+      "loss": 0.4833,
+      "step": 23813
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.884101854254982e-06,
+      "loss": 0.3609,
+      "step": 23814
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.883525103094271e-06,
+      "loss": 0.4639,
+      "step": 23815
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.882948351933559e-06,
+      "loss": 0.3695,
+      "step": 23816
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.882371600772847e-06,
+      "loss": 0.422,
+      "step": 23817
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.881794849612136e-06,
+      "loss": 0.45,
+      "step": 23818
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.881218098451423e-06,
+      "loss": 0.3903,
+      "step": 23819
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.880641347290712e-06,
+      "loss": 0.4828,
+      "step": 23820
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.88006459613e-06,
+      "loss": 0.3963,
+      "step": 23821
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.879487844969288e-06,
+      "loss": 0.3671,
+      "step": 23822
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.878911093808577e-06,
+      "loss": 0.3734,
+      "step": 23823
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.878334342647865e-06,
+      "loss": 0.4745,
+      "step": 23824
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.877757591487153e-06,
+      "loss": 0.5159,
+      "step": 23825
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.877180840326441e-06,
+      "loss": 0.4366,
+      "step": 23826
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.876604089165731e-06,
+      "loss": 0.4376,
+      "step": 23827
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8760273380050176e-06,
+      "loss": 0.3782,
+      "step": 23828
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.875450586844307e-06,
+      "loss": 0.358,
+      "step": 23829
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8748738356835955e-06,
+      "loss": 0.5727,
+      "step": 23830
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.874297084522883e-06,
+      "loss": 0.3907,
+      "step": 23831
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.873720333362172e-06,
+      "loss": 0.3954,
+      "step": 23832
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.87314358220146e-06,
+      "loss": 0.5078,
+      "step": 23833
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.872566831040748e-06,
+      "loss": 0.3997,
+      "step": 23834
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8719900798800365e-06,
+      "loss": 0.3074,
+      "step": 23835
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.871413328719324e-06,
+      "loss": 0.4356,
+      "step": 23836
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.870836577558613e-06,
+      "loss": 0.4253,
+      "step": 23837
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.870259826397901e-06,
+      "loss": 0.4756,
+      "step": 23838
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.869683075237189e-06,
+      "loss": 0.4469,
+      "step": 23839
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.869106324076477e-06,
+      "loss": 0.4227,
+      "step": 23840
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.868529572915767e-06,
+      "loss": 0.429,
+      "step": 23841
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.867952821755054e-06,
+      "loss": 0.5013,
+      "step": 23842
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.867376070594343e-06,
+      "loss": 0.5633,
+      "step": 23843
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8667993194336315e-06,
+      "loss": 0.4787,
+      "step": 23844
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.866222568272919e-06,
+      "loss": 0.4475,
+      "step": 23845
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.865645817112208e-06,
+      "loss": 0.3991,
+      "step": 23846
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.865069065951496e-06,
+      "loss": 0.4654,
+      "step": 23847
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.864492314790784e-06,
+      "loss": 0.3808,
+      "step": 23848
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8639155636300725e-06,
+      "loss": 0.3264,
+      "step": 23849
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.863338812469361e-06,
+      "loss": 0.3655,
+      "step": 23850
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.862762061308649e-06,
+      "loss": 0.3589,
+      "step": 23851
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.862185310147937e-06,
+      "loss": 0.3916,
+      "step": 23852
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.861608558987225e-06,
+      "loss": 0.4152,
+      "step": 23853
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8610318078265135e-06,
+      "loss": 0.3893,
+      "step": 23854
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.860455056665802e-06,
+      "loss": 0.484,
+      "step": 23855
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.85987830550509e-06,
+      "loss": 0.5716,
+      "step": 23856
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.859301554344378e-06,
+      "loss": 0.4532,
+      "step": 23857
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.858724803183668e-06,
+      "loss": 0.4495,
+      "step": 23858
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8581480520229544e-06,
+      "loss": 0.3481,
+      "step": 23859
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.857571300862244e-06,
+      "loss": 0.4752,
+      "step": 23860
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.856994549701532e-06,
+      "loss": 0.3967,
+      "step": 23861
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.85641779854082e-06,
+      "loss": 0.3415,
+      "step": 23862
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 6.8558410473801086e-06,
+      "loss": 0.4324,
+      "step": 23863
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.855264296219397e-06,
+      "loss": 0.5205,
+      "step": 23864
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.854687545058685e-06,
+      "loss": 0.4074,
+      "step": 23865
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.854110793897973e-06,
+      "loss": 0.4816,
+      "step": 23866
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.853534042737262e-06,
+      "loss": 0.4468,
+      "step": 23867
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8529572915765495e-06,
+      "loss": 0.4308,
+      "step": 23868
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.852380540415838e-06,
+      "loss": 0.4664,
+      "step": 23869
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851803789255126e-06,
+      "loss": 0.4232,
+      "step": 23870
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.851227038094414e-06,
+      "loss": 0.3856,
+      "step": 23871
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.850650286933704e-06,
+      "loss": 0.4193,
+      "step": 23872
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8500735357729905e-06,
+      "loss": 0.4934,
+      "step": 23873
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.84949678461228e-06,
+      "loss": 0.4249,
+      "step": 23874
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.848920033451568e-06,
+      "loss": 0.5027,
+      "step": 23875
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.848343282290856e-06,
+      "loss": 0.4857,
+      "step": 23876
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.847766531130145e-06,
+      "loss": 0.5137,
+      "step": 23877
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.847189779969433e-06,
+      "loss": 0.4029,
+      "step": 23878
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.846613028808721e-06,
+      "loss": 0.3892,
+      "step": 23879
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.846036277648009e-06,
+      "loss": 0.4603,
+      "step": 23880
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.845459526487298e-06,
+      "loss": 0.386,
+      "step": 23881
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.844882775326586e-06,
+      "loss": 0.3429,
+      "step": 23882
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.844306024165874e-06,
+      "loss": 0.5636,
+      "step": 23883
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.843729273005163e-06,
+      "loss": 0.4769,
+      "step": 23884
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.84315252184445e-06,
+      "loss": 0.4981,
+      "step": 23885
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.842575770683739e-06,
+      "loss": 0.5308,
+      "step": 23886
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8419990195230266e-06,
+      "loss": 0.4108,
+      "step": 23887
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.841422268362315e-06,
+      "loss": 0.443,
+      "step": 23888
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8408455172016045e-06,
+      "loss": 0.3948,
+      "step": 23889
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.840268766040891e-06,
+      "loss": 0.4097,
+      "step": 23890
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.839692014880181e-06,
+      "loss": 0.3876,
+      "step": 23891
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.839115263719469e-06,
+      "loss": 0.3965,
+      "step": 23892
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.838538512558757e-06,
+      "loss": 0.4269,
+      "step": 23893
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8379617613980454e-06,
+      "loss": 0.3862,
+      "step": 23894
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.837385010237334e-06,
+      "loss": 0.4136,
+      "step": 23895
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.836808259076622e-06,
+      "loss": 0.375,
+      "step": 23896
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.83623150791591e-06,
+      "loss": 0.4642,
+      "step": 23897
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.835654756755199e-06,
+      "loss": 0.4447,
+      "step": 23898
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.835078005594486e-06,
+      "loss": 0.342,
+      "step": 23899
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.834501254433775e-06,
+      "loss": 0.4191,
+      "step": 23900
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8339245032730635e-06,
+      "loss": 0.4374,
+      "step": 23901
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.833347752112351e-06,
+      "loss": 0.5409,
+      "step": 23902
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8327710009516405e-06,
+      "loss": 0.4397,
+      "step": 23903
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.832194249790927e-06,
+      "loss": 0.3819,
+      "step": 23904
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831617498630217e-06,
+      "loss": 0.5073,
+      "step": 23905
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.831040747469505e-06,
+      "loss": 0.3904,
+      "step": 23906
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.830463996308793e-06,
+      "loss": 0.4433,
+      "step": 23907
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8298872451480815e-06,
+      "loss": 0.3361,
+      "step": 23908
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.82931049398737e-06,
+      "loss": 0.441,
+      "step": 23909
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.828733742826658e-06,
+      "loss": 0.3832,
+      "step": 23910
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.828156991665946e-06,
+      "loss": 0.3553,
+      "step": 23911
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.827580240505235e-06,
+      "loss": 0.4882,
+      "step": 23912
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8270034893445225e-06,
+      "loss": 0.4228,
+      "step": 23913
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.826426738183811e-06,
+      "loss": 0.3863,
+      "step": 23914
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8258499870230995e-06,
+      "loss": 0.3484,
+      "step": 23915
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.825273235862387e-06,
+      "loss": 0.4414,
+      "step": 23916
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.824696484701676e-06,
+      "loss": 0.4099,
+      "step": 23917
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.824119733540965e-06,
+      "loss": 0.4984,
+      "step": 23918
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.823542982380252e-06,
+      "loss": 0.4582,
+      "step": 23919
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.822966231219541e-06,
+      "loss": 0.4171,
+      "step": 23920
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.82238948005883e-06,
+      "loss": 0.3281,
+      "step": 23921
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8218127288981176e-06,
+      "loss": 0.5548,
+      "step": 23922
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.821235977737406e-06,
+      "loss": 0.3511,
+      "step": 23923
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.820659226576694e-06,
+      "loss": 0.4156,
+      "step": 23924
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.820082475415982e-06,
+      "loss": 0.4575,
+      "step": 23925
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.819505724255271e-06,
+      "loss": 0.3926,
+      "step": 23926
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8189289730945585e-06,
+      "loss": 0.4495,
+      "step": 23927
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.818352221933847e-06,
+      "loss": 0.4542,
+      "step": 23928
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.817775470773136e-06,
+      "loss": 0.603,
+      "step": 23929
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.817198719612423e-06,
+      "loss": 0.4671,
+      "step": 23930
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.816621968451712e-06,
+      "loss": 0.4291,
+      "step": 23931
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.816045217291e-06,
+      "loss": 0.4187,
+      "step": 23932
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.815468466130288e-06,
+      "loss": 0.4778,
+      "step": 23933
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.814891714969577e-06,
+      "loss": 0.4292,
+      "step": 23934
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.814314963808866e-06,
+      "loss": 0.3246,
+      "step": 23935
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.813738212648154e-06,
+      "loss": 0.4256,
+      "step": 23936
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.813161461487442e-06,
+      "loss": 0.4349,
+      "step": 23937
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.812584710326731e-06,
+      "loss": 0.3543,
+      "step": 23938
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.812007959166018e-06,
+      "loss": 0.4399,
+      "step": 23939
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.811431208005307e-06,
+      "loss": 0.4778,
+      "step": 23940
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810854456844595e-06,
+      "loss": 0.4038,
+      "step": 23941
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.810277705683883e-06,
+      "loss": 0.6355,
+      "step": 23942
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.809700954523172e-06,
+      "loss": 0.3734,
+      "step": 23943
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.809124203362459e-06,
+      "loss": 0.3686,
+      "step": 23944
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.808547452201748e-06,
+      "loss": 0.4512,
+      "step": 23945
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.807970701041036e-06,
+      "loss": 0.4501,
+      "step": 23946
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.807393949880324e-06,
+      "loss": 0.4089,
+      "step": 23947
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.806817198719613e-06,
+      "loss": 0.4757,
+      "step": 23948
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.806240447558902e-06,
+      "loss": 0.4461,
+      "step": 23949
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.805663696398189e-06,
+      "loss": 0.4174,
+      "step": 23950
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.805086945237478e-06,
+      "loss": 0.4703,
+      "step": 23951
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.804510194076767e-06,
+      "loss": 0.5788,
+      "step": 23952
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8039334429160544e-06,
+      "loss": 0.4425,
+      "step": 23953
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.803356691755343e-06,
+      "loss": 0.3911,
+      "step": 23954
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.8027799405946315e-06,
+      "loss": 0.3654,
+      "step": 23955
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.802203189433919e-06,
+      "loss": 0.3975,
+      "step": 23956
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.801626438273208e-06,
+      "loss": 0.47,
+      "step": 23957
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.801049687112495e-06,
+      "loss": 0.6291,
+      "step": 23958
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.800472935951784e-06,
+      "loss": 0.3933,
+      "step": 23959
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7998961847910725e-06,
+      "loss": 0.4362,
+      "step": 23960
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.79931943363036e-06,
+      "loss": 0.4517,
+      "step": 23961
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.798742682469649e-06,
+      "loss": 0.4795,
+      "step": 23962
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.798165931308938e-06,
+      "loss": 0.4187,
+      "step": 23963
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.797589180148225e-06,
+      "loss": 0.4884,
+      "step": 23964
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.797012428987514e-06,
+      "loss": 0.4021,
+      "step": 23965
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.796435677826803e-06,
+      "loss": 0.3941,
+      "step": 23966
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7958589266660905e-06,
+      "loss": 0.3871,
+      "step": 23967
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.795282175505379e-06,
+      "loss": 0.4466,
+      "step": 23968
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7947054243446676e-06,
+      "loss": 0.453,
+      "step": 23969
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.794128673183955e-06,
+      "loss": 0.3763,
+      "step": 23970
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.793551922023244e-06,
+      "loss": 0.3093,
+      "step": 23971
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.792975170862532e-06,
+      "loss": 0.5146,
+      "step": 23972
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.79239841970182e-06,
+      "loss": 0.3324,
+      "step": 23973
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7918216685411085e-06,
+      "loss": 0.3337,
+      "step": 23974
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.791244917380396e-06,
+      "loss": 0.435,
+      "step": 23975
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790668166219685e-06,
+      "loss": 0.4308,
+      "step": 23976
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.790091415058973e-06,
+      "loss": 0.4322,
+      "step": 23977
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.789514663898261e-06,
+      "loss": 0.4879,
+      "step": 23978
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7889379127375495e-06,
+      "loss": 0.4146,
+      "step": 23979
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.788361161576839e-06,
+      "loss": 0.5211,
+      "step": 23980
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.787784410416126e-06,
+      "loss": 0.31,
+      "step": 23981
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.787207659255415e-06,
+      "loss": 0.4648,
+      "step": 23982
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.786630908094704e-06,
+      "loss": 0.4394,
+      "step": 23983
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.786054156933991e-06,
+      "loss": 0.4273,
+      "step": 23984
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.78547740577328e-06,
+      "loss": 0.4144,
+      "step": 23985
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.784900654612568e-06,
+      "loss": 0.4396,
+      "step": 23986
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.784323903451856e-06,
+      "loss": 0.5424,
+      "step": 23987
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.783747152291145e-06,
+      "loss": 0.4345,
+      "step": 23988
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.783170401130433e-06,
+      "loss": 0.4366,
+      "step": 23989
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.782593649969721e-06,
+      "loss": 0.4894,
+      "step": 23990
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.782016898809009e-06,
+      "loss": 0.4003,
+      "step": 23991
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.781440147648297e-06,
+      "loss": 0.3686,
+      "step": 23992
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7808633964875856e-06,
+      "loss": 0.5565,
+      "step": 23993
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.780286645326875e-06,
+      "loss": 0.4051,
+      "step": 23994
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.779709894166162e-06,
+      "loss": 0.4854,
+      "step": 23995
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.779133143005451e-06,
+      "loss": 0.4481,
+      "step": 23996
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.77855639184474e-06,
+      "loss": 0.4128,
+      "step": 23997
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.777979640684027e-06,
+      "loss": 0.5041,
+      "step": 23998
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.777402889523316e-06,
+      "loss": 0.5289,
+      "step": 23999
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7768261383626044e-06,
+      "loss": 0.463,
+      "step": 24000
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.776249387201892e-06,
+      "loss": 0.4512,
+      "step": 24001
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.775672636041181e-06,
+      "loss": 0.2717,
+      "step": 24002
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.775095884880469e-06,
+      "loss": 0.3382,
+      "step": 24003
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.774519133719757e-06,
+      "loss": 0.5581,
+      "step": 24004
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.773942382559045e-06,
+      "loss": 0.4501,
+      "step": 24005
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.773365631398334e-06,
+      "loss": 0.4869,
+      "step": 24006
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.772788880237622e-06,
+      "loss": 0.4133,
+      "step": 24007
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.77221212907691e-06,
+      "loss": 0.4851,
+      "step": 24008
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.771635377916198e-06,
+      "loss": 0.4082,
+      "step": 24009
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.771058626755486e-06,
+      "loss": 0.4414,
+      "step": 24010
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.770481875594776e-06,
+      "loss": 0.3195,
+      "step": 24011
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.769905124434063e-06,
+      "loss": 0.4346,
+      "step": 24012
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.769328373273352e-06,
+      "loss": 0.4063,
+      "step": 24013
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7687516221126405e-06,
+      "loss": 0.4526,
+      "step": 24014
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.768174870951928e-06,
+      "loss": 0.3348,
+      "step": 24015
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.767598119791217e-06,
+      "loss": 0.3432,
+      "step": 24016
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.767021368630505e-06,
+      "loss": 0.3904,
+      "step": 24017
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.766444617469793e-06,
+      "loss": 0.4568,
+      "step": 24018
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7658678663090815e-06,
+      "loss": 0.3942,
+      "step": 24019
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.76529111514837e-06,
+      "loss": 0.4118,
+      "step": 24020
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.764714363987658e-06,
+      "loss": 0.3995,
+      "step": 24021
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.764137612826946e-06,
+      "loss": 0.4285,
+      "step": 24022
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.763560861666235e-06,
+      "loss": 0.4638,
+      "step": 24023
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7629841105055224e-06,
+      "loss": 0.4969,
+      "step": 24024
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.762407359344812e-06,
+      "loss": 0.464,
+      "step": 24025
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.761830608184099e-06,
+      "loss": 0.4142,
+      "step": 24026
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.761253857023388e-06,
+      "loss": 0.4517,
+      "step": 24027
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7606771058626766e-06,
+      "loss": 0.5124,
+      "step": 24028
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.760100354701964e-06,
+      "loss": 0.4389,
+      "step": 24029
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.759523603541253e-06,
+      "loss": 0.4809,
+      "step": 24030
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.758946852380541e-06,
+      "loss": 0.333,
+      "step": 24031
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.758370101219829e-06,
+      "loss": 0.4885,
+      "step": 24032
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7577933500591175e-06,
+      "loss": 0.3941,
+      "step": 24033
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.757216598898406e-06,
+      "loss": 0.3545,
+      "step": 24034
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.756639847737694e-06,
+      "loss": 0.5117,
+      "step": 24035
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.756063096576982e-06,
+      "loss": 0.4096,
+      "step": 24036
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.755486345416271e-06,
+      "loss": 0.4032,
+      "step": 24037
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.7549095942555585e-06,
+      "loss": 0.4107,
+      "step": 24038
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.754332843094847e-06,
+      "loss": 0.3908,
+      "step": 24039
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.753756091934136e-06,
+      "loss": 0.3501,
+      "step": 24040
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 6.753179340773423e-06,
+      "loss": 0.3737,
+      "step": 24041
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.752602589612713e-06,
+      "loss": 0.3955,
+      "step": 24042
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.752025838452001e-06,
+      "loss": 0.4846,
+      "step": 24043
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.751449087291289e-06,
+      "loss": 0.3185,
+      "step": 24044
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.750872336130577e-06,
+      "loss": 0.374,
+      "step": 24045
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.750295584969865e-06,
+      "loss": 0.4763,
+      "step": 24046
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.749718833809154e-06,
+      "loss": 0.3665,
+      "step": 24047
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.749142082648442e-06,
+      "loss": 0.3875,
+      "step": 24048
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.74856533148773e-06,
+      "loss": 0.4387,
+      "step": 24049
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.747988580327018e-06,
+      "loss": 0.3863,
+      "step": 24050
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.747411829166307e-06,
+      "loss": 0.4501,
+      "step": 24051
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7468350780055946e-06,
+      "loss": 0.4629,
+      "step": 24052
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.746258326844883e-06,
+      "loss": 0.4574,
+      "step": 24053
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.745681575684172e-06,
+      "loss": 0.4157,
+      "step": 24054
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.745104824523459e-06,
+      "loss": 0.3894,
+      "step": 24055
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.744528073362749e-06,
+      "loss": 0.4872,
+      "step": 24056
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.743951322202037e-06,
+      "loss": 0.3445,
+      "step": 24057
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.743374571041325e-06,
+      "loss": 0.3604,
+      "step": 24058
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7427978198806134e-06,
+      "loss": 0.4461,
+      "step": 24059
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.742221068719902e-06,
+      "loss": 0.3663,
+      "step": 24060
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.74164431755919e-06,
+      "loss": 0.3713,
+      "step": 24061
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.741067566398478e-06,
+      "loss": 0.3405,
+      "step": 24062
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.740490815237766e-06,
+      "loss": 0.3535,
+      "step": 24063
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.739914064077054e-06,
+      "loss": 0.3796,
+      "step": 24064
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.739337312916343e-06,
+      "loss": 0.4802,
+      "step": 24065
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.738760561755631e-06,
+      "loss": 0.4384,
+      "step": 24066
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.738183810594919e-06,
+      "loss": 0.4278,
+      "step": 24067
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.737607059434208e-06,
+      "loss": 0.4384,
+      "step": 24068
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.737030308273495e-06,
+      "loss": 0.3455,
+      "step": 24069
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.736453557112784e-06,
+      "loss": 0.4089,
+      "step": 24070
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.735876805952073e-06,
+      "loss": 0.4135,
+      "step": 24071
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.73530005479136e-06,
+      "loss": 0.3866,
+      "step": 24072
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7347233036306495e-06,
+      "loss": 0.4965,
+      "step": 24073
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.734146552469938e-06,
+      "loss": 0.4291,
+      "step": 24074
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.733569801309226e-06,
+      "loss": 0.4851,
+      "step": 24075
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.732993050148514e-06,
+      "loss": 0.3841,
+      "step": 24076
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.732416298987803e-06,
+      "loss": 0.4532,
+      "step": 24077
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7318395478270905e-06,
+      "loss": 0.469,
+      "step": 24078
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.731262796666379e-06,
+      "loss": 0.412,
+      "step": 24079
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.730686045505667e-06,
+      "loss": 0.4068,
+      "step": 24080
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.730109294344955e-06,
+      "loss": 0.4012,
+      "step": 24081
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.729532543184244e-06,
+      "loss": 0.4161,
+      "step": 24082
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7289557920235314e-06,
+      "loss": 0.4846,
+      "step": 24083
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.72837904086282e-06,
+      "loss": 0.4114,
+      "step": 24084
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7278022897021085e-06,
+      "loss": 0.4967,
+      "step": 24085
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.727225538541396e-06,
+      "loss": 0.4953,
+      "step": 24086
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7266487873806856e-06,
+      "loss": 0.5246,
+      "step": 24087
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.726072036219974e-06,
+      "loss": 0.4624,
+      "step": 24088
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.725495285059262e-06,
+      "loss": 0.3885,
+      "step": 24089
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.72491853389855e-06,
+      "loss": 0.3864,
+      "step": 24090
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.724341782737839e-06,
+      "loss": 0.4839,
+      "step": 24091
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7237650315771265e-06,
+      "loss": 0.4405,
+      "step": 24092
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.723188280416415e-06,
+      "loss": 0.4749,
+      "step": 24093
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.722611529255704e-06,
+      "loss": 0.5686,
+      "step": 24094
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.722034778094991e-06,
+      "loss": 0.3798,
+      "step": 24095
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.72145802693428e-06,
+      "loss": 0.3956,
+      "step": 24096
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7208812757735675e-06,
+      "loss": 0.419,
+      "step": 24097
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.720304524612856e-06,
+      "loss": 0.4861,
+      "step": 24098
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7197277734521446e-06,
+      "loss": 0.4039,
+      "step": 24099
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.719151022291432e-06,
+      "loss": 0.3692,
+      "step": 24100
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.718574271130721e-06,
+      "loss": 0.4132,
+      "step": 24101
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.71799751997001e-06,
+      "loss": 0.4424,
+      "step": 24102
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.717420768809297e-06,
+      "loss": 0.3256,
+      "step": 24103
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.716844017648586e-06,
+      "loss": 0.3325,
+      "step": 24104
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.716267266487875e-06,
+      "loss": 0.4252,
+      "step": 24105
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.715690515327163e-06,
+      "loss": 0.4462,
+      "step": 24106
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.715113764166451e-06,
+      "loss": 0.4323,
+      "step": 24107
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.71453701300574e-06,
+      "loss": 0.4797,
+      "step": 24108
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.713960261845027e-06,
+      "loss": 0.4182,
+      "step": 24109
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.713383510684316e-06,
+      "loss": 0.4243,
+      "step": 24110
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.712806759523604e-06,
+      "loss": 0.511,
+      "step": 24111
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.712230008362892e-06,
+      "loss": 0.3532,
+      "step": 24112
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.711653257202181e-06,
+      "loss": 0.3184,
+      "step": 24113
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.711076506041468e-06,
+      "loss": 0.4168,
+      "step": 24114
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.710499754880757e-06,
+      "loss": 0.4233,
+      "step": 24115
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.709923003720045e-06,
+      "loss": 0.282,
+      "step": 24116
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.709346252559333e-06,
+      "loss": 0.4717,
+      "step": 24117
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7087695013986224e-06,
+      "loss": 0.3891,
+      "step": 24118
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.708192750237911e-06,
+      "loss": 0.431,
+      "step": 24119
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707615999077199e-06,
+      "loss": 0.3255,
+      "step": 24120
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.707039247916487e-06,
+      "loss": 0.4487,
+      "step": 24121
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.706462496755776e-06,
+      "loss": 0.3297,
+      "step": 24122
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.705885745595063e-06,
+      "loss": 0.3382,
+      "step": 24123
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.705308994434352e-06,
+      "loss": 0.2419,
+      "step": 24124
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7047322432736405e-06,
+      "loss": 0.4428,
+      "step": 24125
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.704155492112928e-06,
+      "loss": 0.4965,
+      "step": 24126
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.703578740952217e-06,
+      "loss": 0.3787,
+      "step": 24127
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.703001989791505e-06,
+      "loss": 0.4241,
+      "step": 24128
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.702425238630793e-06,
+      "loss": 0.3504,
+      "step": 24129
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.7018484874700814e-06,
+      "loss": 0.4518,
+      "step": 24130
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.701271736309369e-06,
+      "loss": 0.4079,
+      "step": 24131
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.700694985148658e-06,
+      "loss": 0.4978,
+      "step": 24132
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.700118233987947e-06,
+      "loss": 0.3995,
+      "step": 24133
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.699541482827234e-06,
+      "loss": 0.4168,
+      "step": 24134
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.698964731666523e-06,
+      "loss": 0.3959,
+      "step": 24135
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.698387980505812e-06,
+      "loss": 0.4646,
+      "step": 24136
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6978112293450995e-06,
+      "loss": 0.3124,
+      "step": 24137
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.697234478184388e-06,
+      "loss": 0.5395,
+      "step": 24138
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6966577270236765e-06,
+      "loss": 0.3878,
+      "step": 24139
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.696080975862964e-06,
+      "loss": 0.4435,
+      "step": 24140
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.695504224702253e-06,
+      "loss": 0.4031,
+      "step": 24141
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.694927473541541e-06,
+      "loss": 0.3779,
+      "step": 24142
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.694350722380829e-06,
+      "loss": 0.4025,
+      "step": 24143
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6937739712201175e-06,
+      "loss": 0.4685,
+      "step": 24144
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.693197220059406e-06,
+      "loss": 0.4463,
+      "step": 24145
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.692620468898694e-06,
+      "loss": 0.4771,
+      "step": 24146
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.692043717737983e-06,
+      "loss": 0.6242,
+      "step": 24147
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.69146696657727e-06,
+      "loss": 0.3638,
+      "step": 24148
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.690890215416559e-06,
+      "loss": 0.4208,
+      "step": 24149
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.690313464255848e-06,
+      "loss": 0.4312,
+      "step": 24150
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6897367130951355e-06,
+      "loss": 0.4933,
+      "step": 24151
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.689159961934424e-06,
+      "loss": 0.5072,
+      "step": 24152
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.688583210773713e-06,
+      "loss": 0.387,
+      "step": 24153
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.688006459613e-06,
+      "loss": 0.4903,
+      "step": 24154
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.687429708452289e-06,
+      "loss": 0.2695,
+      "step": 24155
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.686852957291577e-06,
+      "loss": 0.4509,
+      "step": 24156
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.686276206130865e-06,
+      "loss": 0.2894,
+      "step": 24157
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6856994549701536e-06,
+      "loss": 0.396,
+      "step": 24158
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.685122703809442e-06,
+      "loss": 0.4281,
+      "step": 24159
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68454595264873e-06,
+      "loss": 0.4667,
+      "step": 24160
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.683969201488018e-06,
+      "loss": 0.4154,
+      "step": 24161
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.683392450327308e-06,
+      "loss": 0.3522,
+      "step": 24162
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6828156991665945e-06,
+      "loss": 0.4746,
+      "step": 24163
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.682238948005884e-06,
+      "loss": 0.3624,
+      "step": 24164
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.681662196845171e-06,
+      "loss": 0.4485,
+      "step": 24165
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.68108544568446e-06,
+      "loss": 0.3538,
+      "step": 24166
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.680508694523749e-06,
+      "loss": 0.3431,
+      "step": 24167
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.679931943363036e-06,
+      "loss": 0.5033,
+      "step": 24168
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.679355192202325e-06,
+      "loss": 0.4685,
+      "step": 24169
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.678778441041613e-06,
+      "loss": 0.4353,
+      "step": 24170
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.678201689880901e-06,
+      "loss": 0.478,
+      "step": 24171
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.67762493872019e-06,
+      "loss": 0.3858,
+      "step": 24172
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.677048187559478e-06,
+      "loss": 0.4672,
+      "step": 24173
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.676471436398766e-06,
+      "loss": 0.3983,
+      "step": 24174
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.675894685238054e-06,
+      "loss": 0.3501,
+      "step": 24175
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.675317934077343e-06,
+      "loss": 0.3764,
+      "step": 24176
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.674741182916631e-06,
+      "loss": 0.415,
+      "step": 24177
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.67416443175592e-06,
+      "loss": 0.5434,
+      "step": 24178
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6735876805952085e-06,
+      "loss": 0.409,
+      "step": 24179
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.673010929434496e-06,
+      "loss": 0.3431,
+      "step": 24180
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.672434178273785e-06,
+      "loss": 0.4003,
+      "step": 24181
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.671857427113073e-06,
+      "loss": 0.4284,
+      "step": 24182
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.671280675952361e-06,
+      "loss": 0.3831,
+      "step": 24183
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6707039247916495e-06,
+      "loss": 0.306,
+      "step": 24184
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.670127173630937e-06,
+      "loss": 0.3351,
+      "step": 24185
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.669550422470226e-06,
+      "loss": 0.4342,
+      "step": 24186
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.668973671309514e-06,
+      "loss": 0.4428,
+      "step": 24187
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.668396920148802e-06,
+      "loss": 0.3898,
+      "step": 24188
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6678201689880904e-06,
+      "loss": 0.4107,
+      "step": 24189
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.667243417827379e-06,
+      "loss": 0.4494,
+      "step": 24190
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.5414,
+      "step": 24191
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.666089915505955e-06,
+      "loss": 0.5051,
+      "step": 24192
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6655131643452446e-06,
+      "loss": 0.4081,
+      "step": 24193
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.664936413184531e-06,
+      "loss": 0.5066,
+      "step": 24194
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.664359662023821e-06,
+      "loss": 0.4474,
+      "step": 24195
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.663782910863109e-06,
+      "loss": 0.3694,
+      "step": 24196
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.663206159702397e-06,
+      "loss": 0.3681,
+      "step": 24197
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6626294085416855e-06,
+      "loss": 0.3953,
+      "step": 24198
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.662052657380974e-06,
+      "loss": 0.4421,
+      "step": 24199
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.661475906220262e-06,
+      "loss": 0.4966,
+      "step": 24200
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.66089915505955e-06,
+      "loss": 0.3882,
+      "step": 24201
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.660322403898838e-06,
+      "loss": 0.5262,
+      "step": 24202
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6597456527381265e-06,
+      "loss": 0.5099,
+      "step": 24203
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.659168901577415e-06,
+      "loss": 0.4945,
+      "step": 24204
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.658592150416703e-06,
+      "loss": 0.3964,
+      "step": 24205
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.658015399255991e-06,
+      "loss": 0.3079,
+      "step": 24206
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.65743864809528e-06,
+      "loss": 0.4831,
+      "step": 24207
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6568618969345675e-06,
+      "loss": 0.3752,
+      "step": 24208
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.656285145773857e-06,
+      "loss": 0.4763,
+      "step": 24209
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.655708394613145e-06,
+      "loss": 0.5289,
+      "step": 24210
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.655131643452433e-06,
+      "loss": 0.3768,
+      "step": 24211
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.654554892291722e-06,
+      "loss": 0.4389,
+      "step": 24212
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.65397814113101e-06,
+      "loss": 0.5172,
+      "step": 24213
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.653401389970298e-06,
+      "loss": 0.3936,
+      "step": 24214
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.652824638809586e-06,
+      "loss": 0.495,
+      "step": 24215
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.652247887648875e-06,
+      "loss": 0.3156,
+      "step": 24216
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.6516711364881626e-06,
+      "loss": 0.3839,
+      "step": 24217
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.651094385327451e-06,
+      "loss": 0.4577,
+      "step": 24218
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.650517634166739e-06,
+      "loss": 0.4139,
+      "step": 24219
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 6.649940883006027e-06,
+      "loss": 0.4248,
+      "step": 24220
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.649364131845316e-06,
+      "loss": 0.4117,
+      "step": 24221
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6487873806846035e-06,
+      "loss": 0.4271,
+      "step": 24222
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.648210629523892e-06,
+      "loss": 0.4106,
+      "step": 24223
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6476338783631814e-06,
+      "loss": 0.3338,
+      "step": 24224
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.647057127202468e-06,
+      "loss": 0.4041,
+      "step": 24225
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.646480376041758e-06,
+      "loss": 0.2916,
+      "step": 24226
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.645903624881046e-06,
+      "loss": 0.4052,
+      "step": 24227
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.645326873720334e-06,
+      "loss": 0.4018,
+      "step": 24228
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.644750122559622e-06,
+      "loss": 0.3501,
+      "step": 24229
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.644173371398911e-06,
+      "loss": 0.4079,
+      "step": 24230
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.643596620238199e-06,
+      "loss": 0.435,
+      "step": 24231
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.643019869077487e-06,
+      "loss": 0.3767,
+      "step": 24232
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.642443117916776e-06,
+      "loss": 0.4118,
+      "step": 24233
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.641866366756063e-06,
+      "loss": 0.4086,
+      "step": 24234
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.641289615595352e-06,
+      "loss": 0.4611,
+      "step": 24235
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.64071286443464e-06,
+      "loss": 0.4574,
+      "step": 24236
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.640136113273928e-06,
+      "loss": 0.5069,
+      "step": 24237
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.639559362113217e-06,
+      "loss": 0.4578,
+      "step": 24238
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.638982610952504e-06,
+      "loss": 0.5503,
+      "step": 24239
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.638405859791794e-06,
+      "loss": 0.5251,
+      "step": 24240
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.637829108631082e-06,
+      "loss": 0.5104,
+      "step": 24241
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.63725235747037e-06,
+      "loss": 0.5203,
+      "step": 24242
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6366756063096585e-06,
+      "loss": 0.3516,
+      "step": 24243
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.636098855148947e-06,
+      "loss": 0.3986,
+      "step": 24244
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.635522103988235e-06,
+      "loss": 0.4856,
+      "step": 24245
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.634945352827523e-06,
+      "loss": 0.464,
+      "step": 24246
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.634368601666812e-06,
+      "loss": 0.4336,
+      "step": 24247
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6337918505060994e-06,
+      "loss": 0.3676,
+      "step": 24248
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.633215099345388e-06,
+      "loss": 0.5549,
+      "step": 24249
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6326383481846765e-06,
+      "loss": 0.62,
+      "step": 24250
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.632061597023964e-06,
+      "loss": 0.3513,
+      "step": 24251
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.631484845863253e-06,
+      "loss": 0.4396,
+      "step": 24252
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.63090809470254e-06,
+      "loss": 0.368,
+      "step": 24253
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.630331343541829e-06,
+      "loss": 0.5347,
+      "step": 24254
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.629754592381118e-06,
+      "loss": 0.4758,
+      "step": 24255
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.629177841220405e-06,
+      "loss": 0.5178,
+      "step": 24256
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6286010900596945e-06,
+      "loss": 0.495,
+      "step": 24257
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.628024338898983e-06,
+      "loss": 0.5351,
+      "step": 24258
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.627447587738271e-06,
+      "loss": 0.4672,
+      "step": 24259
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.626870836577559e-06,
+      "loss": 0.4087,
+      "step": 24260
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.626294085416848e-06,
+      "loss": 0.4725,
+      "step": 24261
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6257173342561355e-06,
+      "loss": 0.433,
+      "step": 24262
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.625140583095424e-06,
+      "loss": 0.3821,
+      "step": 24263
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6245638319347126e-06,
+      "loss": 0.3881,
+      "step": 24264
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.623987080774e-06,
+      "loss": 0.3632,
+      "step": 24265
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.623410329613289e-06,
+      "loss": 0.36,
+      "step": 24266
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.622833578452577e-06,
+      "loss": 0.32,
+      "step": 24267
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.622256827291865e-06,
+      "loss": 0.4531,
+      "step": 24268
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6216800761311535e-06,
+      "loss": 0.3009,
+      "step": 24269
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.621103324970441e-06,
+      "loss": 0.4185,
+      "step": 24270
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.620526573809731e-06,
+      "loss": 0.4177,
+      "step": 24271
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.619949822649019e-06,
+      "loss": 0.4624,
+      "step": 24272
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.619373071488307e-06,
+      "loss": 0.4366,
+      "step": 24273
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.618796320327595e-06,
+      "loss": 0.4263,
+      "step": 24274
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.618219569166884e-06,
+      "loss": 0.4219,
+      "step": 24275
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6176428180061716e-06,
+      "loss": 0.4491,
+      "step": 24276
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.61706606684546e-06,
+      "loss": 0.4154,
+      "step": 24277
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.616489315684749e-06,
+      "loss": 0.3942,
+      "step": 24278
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.615912564524036e-06,
+      "loss": 0.3738,
+      "step": 24279
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.615335813363325e-06,
+      "loss": 0.4071,
+      "step": 24280
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.614759062202613e-06,
+      "loss": 0.3109,
+      "step": 24281
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.614182311041901e-06,
+      "loss": 0.3647,
+      "step": 24282
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.61360555988119e-06,
+      "loss": 0.4369,
+      "step": 24283
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.613028808720479e-06,
+      "loss": 0.4537,
+      "step": 24284
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.612452057559766e-06,
+      "loss": 0.4863,
+      "step": 24285
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.611875306399055e-06,
+      "loss": 0.412,
+      "step": 24286
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.611298555238342e-06,
+      "loss": 0.4264,
+      "step": 24287
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.610721804077631e-06,
+      "loss": 0.4224,
+      "step": 24288
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.61014505291692e-06,
+      "loss": 0.4016,
+      "step": 24289
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.609568301756208e-06,
+      "loss": 0.3567,
+      "step": 24290
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.608991550595496e-06,
+      "loss": 0.6862,
+      "step": 24291
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.608414799434785e-06,
+      "loss": 0.4664,
+      "step": 24292
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.607838048274072e-06,
+      "loss": 0.3898,
+      "step": 24293
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.607261297113361e-06,
+      "loss": 0.4216,
+      "step": 24294
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6066845459526494e-06,
+      "loss": 0.4159,
+      "step": 24295
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.606107794791937e-06,
+      "loss": 0.3861,
+      "step": 24296
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.605531043631226e-06,
+      "loss": 0.5326,
+      "step": 24297
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.604954292470514e-06,
+      "loss": 0.4023,
+      "step": 24298
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.604377541309802e-06,
+      "loss": 0.3289,
+      "step": 24299
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60380079014909e-06,
+      "loss": 0.397,
+      "step": 24300
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.60322403898838e-06,
+      "loss": 0.3549,
+      "step": 24301
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6026472878276675e-06,
+      "loss": 0.443,
+      "step": 24302
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.602070536666956e-06,
+      "loss": 0.4756,
+      "step": 24303
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.6014937855062445e-06,
+      "loss": 0.3892,
+      "step": 24304
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.600917034345532e-06,
+      "loss": 0.4475,
+      "step": 24305
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.600340283184821e-06,
+      "loss": 0.3336,
+      "step": 24306
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.599763532024108e-06,
+      "loss": 0.4167,
+      "step": 24307
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.599186780863397e-06,
+      "loss": 0.4206,
+      "step": 24308
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5986100297026855e-06,
+      "loss": 0.3805,
+      "step": 24309
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.598033278541973e-06,
+      "loss": 0.364,
+      "step": 24310
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.597456527381262e-06,
+      "loss": 0.3278,
+      "step": 24311
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.59687977622055e-06,
+      "loss": 0.4313,
+      "step": 24312
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.596303025059838e-06,
+      "loss": 0.321,
+      "step": 24313
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5957262738991265e-06,
+      "loss": 0.622,
+      "step": 24314
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.595149522738416e-06,
+      "loss": 0.4689,
+      "step": 24315
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.594572771577703e-06,
+      "loss": 0.4686,
+      "step": 24316
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.593996020416992e-06,
+      "loss": 0.3887,
+      "step": 24317
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.593419269256281e-06,
+      "loss": 0.344,
+      "step": 24318
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.592842518095568e-06,
+      "loss": 0.4153,
+      "step": 24319
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.592265766934857e-06,
+      "loss": 0.3205,
+      "step": 24320
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.591689015774145e-06,
+      "loss": 0.3738,
+      "step": 24321
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.591112264613433e-06,
+      "loss": 0.4417,
+      "step": 24322
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5905355134527215e-06,
+      "loss": 0.4528,
+      "step": 24323
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.589958762292009e-06,
+      "loss": 0.3917,
+      "step": 24324
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.589382011131298e-06,
+      "loss": 0.5053,
+      "step": 24325
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.588805259970586e-06,
+      "loss": 0.4956,
+      "step": 24326
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.588228508809874e-06,
+      "loss": 0.3556,
+      "step": 24327
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5876517576491625e-06,
+      "loss": 0.4169,
+      "step": 24328
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.587075006488451e-06,
+      "loss": 0.4429,
+      "step": 24329
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.586498255327739e-06,
+      "loss": 0.4378,
+      "step": 24330
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.585921504167028e-06,
+      "loss": 0.4048,
+      "step": 24331
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.585344753006317e-06,
+      "loss": 0.2404,
+      "step": 24332
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584768001845604e-06,
+      "loss": 0.408,
+      "step": 24333
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.584191250684893e-06,
+      "loss": 0.4239,
+      "step": 24334
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.583614499524181e-06,
+      "loss": 0.384,
+      "step": 24335
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.583037748363469e-06,
+      "loss": 0.3482,
+      "step": 24336
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.582460997202758e-06,
+      "loss": 0.386,
+      "step": 24337
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.581884246042046e-06,
+      "loss": 0.377,
+      "step": 24338
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.581307494881334e-06,
+      "loss": 0.4039,
+      "step": 24339
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.580730743720622e-06,
+      "loss": 0.5038,
+      "step": 24340
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.58015399255991e-06,
+      "loss": 0.3589,
+      "step": 24341
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.579577241399199e-06,
+      "loss": 0.467,
+      "step": 24342
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.579000490238487e-06,
+      "loss": 0.4686,
+      "step": 24343
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.578423739077775e-06,
+      "loss": 0.4694,
+      "step": 24344
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.577846987917063e-06,
+      "loss": 0.4183,
+      "step": 24345
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.577270236756353e-06,
+      "loss": 0.3982,
+      "step": 24346
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5766934855956395e-06,
+      "loss": 0.3187,
+      "step": 24347
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.576116734434929e-06,
+      "loss": 0.3627,
+      "step": 24348
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5755399832742175e-06,
+      "loss": 0.4089,
+      "step": 24349
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.574963232113505e-06,
+      "loss": 0.4394,
+      "step": 24350
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.574386480952794e-06,
+      "loss": 0.3743,
+      "step": 24351
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.573809729792082e-06,
+      "loss": 0.4069,
+      "step": 24352
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.57323297863137e-06,
+      "loss": 0.4716,
+      "step": 24353
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.572656227470658e-06,
+      "loss": 0.4374,
+      "step": 24354
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.572079476309947e-06,
+      "loss": 0.4246,
+      "step": 24355
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.571502725149235e-06,
+      "loss": 0.5569,
+      "step": 24356
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.570925973988523e-06,
+      "loss": 0.4916,
+      "step": 24357
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.570349222827811e-06,
+      "loss": 0.4854,
+      "step": 24358
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.569772471667099e-06,
+      "loss": 0.4903,
+      "step": 24359
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.569195720506388e-06,
+      "loss": 0.471,
+      "step": 24360
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.568618969345676e-06,
+      "loss": 0.6266,
+      "step": 24361
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.568042218184965e-06,
+      "loss": 0.4999,
+      "step": 24362
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5674654670242535e-06,
+      "loss": 0.3926,
+      "step": 24363
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.566888715863541e-06,
+      "loss": 0.4404,
+      "step": 24364
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.56631196470283e-06,
+      "loss": 0.4079,
+      "step": 24365
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.565735213542118e-06,
+      "loss": 0.4895,
+      "step": 24366
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.565158462381406e-06,
+      "loss": 0.3914,
+      "step": 24367
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5645817112206945e-06,
+      "loss": 0.4832,
+      "step": 24368
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.564004960059983e-06,
+      "loss": 0.4831,
+      "step": 24369
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.563428208899271e-06,
+      "loss": 0.3244,
+      "step": 24370
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.562851457738559e-06,
+      "loss": 0.3454,
+      "step": 24371
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.562274706577848e-06,
+      "loss": 0.4447,
+      "step": 24372
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5616979554171355e-06,
+      "loss": 0.4846,
+      "step": 24373
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.561121204256424e-06,
+      "loss": 0.4654,
+      "step": 24374
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.560544453095712e-06,
+      "loss": 0.4403,
+      "step": 24375
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.559967701935e-06,
+      "loss": 0.3915,
+      "step": 24376
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.55939095077429e-06,
+      "loss": 0.4088,
+      "step": 24377
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.558814199613576e-06,
+      "loss": 0.3491,
+      "step": 24378
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.558237448452866e-06,
+      "loss": 0.5623,
+      "step": 24379
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.557660697292154e-06,
+      "loss": 0.4009,
+      "step": 24380
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.557083946131442e-06,
+      "loss": 0.4217,
+      "step": 24381
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5565071949707305e-06,
+      "loss": 0.4196,
+      "step": 24382
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.555930443810019e-06,
+      "loss": 0.4117,
+      "step": 24383
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.555353692649307e-06,
+      "loss": 0.3825,
+      "step": 24384
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.554776941488595e-06,
+      "loss": 0.4701,
+      "step": 24385
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.554200190327884e-06,
+      "loss": 0.5087,
+      "step": 24386
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5536234391671715e-06,
+      "loss": 0.4497,
+      "step": 24387
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.55304668800646e-06,
+      "loss": 0.3326,
+      "step": 24388
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5524699368457486e-06,
+      "loss": 0.4566,
+      "step": 24389
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.551893185685036e-06,
+      "loss": 0.4363,
+      "step": 24390
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.551316434524325e-06,
+      "loss": 0.3869,
+      "step": 24391
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.5507396833636125e-06,
+      "loss": 0.4545,
+      "step": 24392
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.550162932202902e-06,
+      "loss": 0.5118,
+      "step": 24393
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.54958618104219e-06,
+      "loss": 0.3364,
+      "step": 24394
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.549009429881478e-06,
+      "loss": 0.5375,
+      "step": 24395
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.548432678720767e-06,
+      "loss": 0.4532,
+      "step": 24396
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.547855927560055e-06,
+      "loss": 0.4592,
+      "step": 24397
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.547279176399343e-06,
+      "loss": 0.4153,
+      "step": 24398
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 6.546702425238631e-06,
+      "loss": 0.4379,
+      "step": 24399
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54612567407792e-06,
+      "loss": 0.4936,
+      "step": 24400
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5455489229172076e-06,
+      "loss": 0.4507,
+      "step": 24401
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.544972171756496e-06,
+      "loss": 0.3824,
+      "step": 24402
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.544395420595785e-06,
+      "loss": 0.3792,
+      "step": 24403
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.543818669435072e-06,
+      "loss": 0.431,
+      "step": 24404
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.543241918274361e-06,
+      "loss": 0.5328,
+      "step": 24405
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.54266516711365e-06,
+      "loss": 0.5282,
+      "step": 24406
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.542088415952937e-06,
+      "loss": 0.471,
+      "step": 24407
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5415116647922265e-06,
+      "loss": 0.2925,
+      "step": 24408
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.540934913631513e-06,
+      "loss": 0.4216,
+      "step": 24409
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.540358162470803e-06,
+      "loss": 0.448,
+      "step": 24410
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.539781411310091e-06,
+      "loss": 0.4041,
+      "step": 24411
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.539204660149379e-06,
+      "loss": 0.5031,
+      "step": 24412
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.538627908988667e-06,
+      "loss": 0.3784,
+      "step": 24413
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.538051157827956e-06,
+      "loss": 0.3601,
+      "step": 24414
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.537474406667244e-06,
+      "loss": 0.4107,
+      "step": 24415
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.536897655506532e-06,
+      "loss": 0.4689,
+      "step": 24416
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.536320904345821e-06,
+      "loss": 0.4535,
+      "step": 24417
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.535744153185108e-06,
+      "loss": 0.5165,
+      "step": 24418
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.535167402024397e-06,
+      "loss": 0.51,
+      "step": 24419
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5345906508636854e-06,
+      "loss": 0.4453,
+      "step": 24420
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.534013899702973e-06,
+      "loss": 0.4909,
+      "step": 24421
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.533437148542262e-06,
+      "loss": 0.3741,
+      "step": 24422
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.532860397381551e-06,
+      "loss": 0.472,
+      "step": 24423
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.532283646220839e-06,
+      "loss": 0.4825,
+      "step": 24424
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.531706895060127e-06,
+      "loss": 0.4641,
+      "step": 24425
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.531130143899416e-06,
+      "loss": 0.557,
+      "step": 24426
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5305533927387035e-06,
+      "loss": 0.4784,
+      "step": 24427
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.529976641577992e-06,
+      "loss": 0.4568,
+      "step": 24428
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.52939989041728e-06,
+      "loss": 0.3978,
+      "step": 24429
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.528823139256568e-06,
+      "loss": 0.4761,
+      "step": 24430
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.528246388095857e-06,
+      "loss": 0.4862,
+      "step": 24431
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5276696369351444e-06,
+      "loss": 0.4518,
+      "step": 24432
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.527092885774433e-06,
+      "loss": 0.3723,
+      "step": 24433
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5265161346137215e-06,
+      "loss": 0.3644,
+      "step": 24434
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.525939383453009e-06,
+      "loss": 0.4607,
+      "step": 24435
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.525362632292298e-06,
+      "loss": 0.3652,
+      "step": 24436
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.524785881131587e-06,
+      "loss": 0.4785,
+      "step": 24437
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.524209129970874e-06,
+      "loss": 0.322,
+      "step": 24438
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.523632378810163e-06,
+      "loss": 0.4807,
+      "step": 24439
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.523055627649452e-06,
+      "loss": 0.3279,
+      "step": 24440
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5224788764887395e-06,
+      "loss": 0.4211,
+      "step": 24441
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.521902125328028e-06,
+      "loss": 0.4045,
+      "step": 24442
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.521325374167317e-06,
+      "loss": 0.2651,
+      "step": 24443
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.520748623006604e-06,
+      "loss": 0.4561,
+      "step": 24444
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.520171871845893e-06,
+      "loss": 0.4735,
+      "step": 24445
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5195951206851805e-06,
+      "loss": 0.4441,
+      "step": 24446
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.519018369524469e-06,
+      "loss": 0.4564,
+      "step": 24447
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5184416183637576e-06,
+      "loss": 0.5561,
+      "step": 24448
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.517864867203045e-06,
+      "loss": 0.2595,
+      "step": 24449
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.517288116042334e-06,
+      "loss": 0.4528,
+      "step": 24450
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.516711364881622e-06,
+      "loss": 0.4398,
+      "step": 24451
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.51613461372091e-06,
+      "loss": 0.4216,
+      "step": 24452
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5155578625601985e-06,
+      "loss": 0.4819,
+      "step": 24453
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.514981111399488e-06,
+      "loss": 0.4314,
+      "step": 24454
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.514404360238776e-06,
+      "loss": 0.3901,
+      "step": 24455
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.513827609078064e-06,
+      "loss": 0.3692,
+      "step": 24456
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.513250857917353e-06,
+      "loss": 0.3351,
+      "step": 24457
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.51267410675664e-06,
+      "loss": 0.4925,
+      "step": 24458
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.512097355595929e-06,
+      "loss": 0.4516,
+      "step": 24459
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.511520604435217e-06,
+      "loss": 0.3506,
+      "step": 24460
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.510943853274505e-06,
+      "loss": 0.4202,
+      "step": 24461
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.510367102113794e-06,
+      "loss": 0.4115,
+      "step": 24462
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.509790350953081e-06,
+      "loss": 0.4268,
+      "step": 24463
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.50921359979237e-06,
+      "loss": 0.4341,
+      "step": 24464
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.508636848631658e-06,
+      "loss": 0.4661,
+      "step": 24465
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.508060097470946e-06,
+      "loss": 0.397,
+      "step": 24466
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.507483346310235e-06,
+      "loss": 0.4396,
+      "step": 24467
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.506906595149524e-06,
+      "loss": 0.3957,
+      "step": 24468
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.506329843988811e-06,
+      "loss": 0.4984,
+      "step": 24469
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5057530928281e-06,
+      "loss": 0.3545,
+      "step": 24470
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.505176341667389e-06,
+      "loss": 0.4438,
+      "step": 24471
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.504599590506676e-06,
+      "loss": 0.5132,
+      "step": 24472
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.504022839345965e-06,
+      "loss": 0.4206,
+      "step": 24473
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5034460881852535e-06,
+      "loss": 0.3503,
+      "step": 24474
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.502869337024541e-06,
+      "loss": 0.3488,
+      "step": 24475
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.50229258586383e-06,
+      "loss": 0.4587,
+      "step": 24476
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.501715834703118e-06,
+      "loss": 0.3986,
+      "step": 24477
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.501139083542406e-06,
+      "loss": 0.4041,
+      "step": 24478
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.5005623323816944e-06,
+      "loss": 0.4612,
+      "step": 24479
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.499985581220982e-06,
+      "loss": 0.4227,
+      "step": 24480
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.499408830060271e-06,
+      "loss": 0.2825,
+      "step": 24481
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.498832078899559e-06,
+      "loss": 0.4792,
+      "step": 24482
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.498255327738847e-06,
+      "loss": 0.4743,
+      "step": 24483
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.497678576578135e-06,
+      "loss": 0.4611,
+      "step": 24484
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.497101825417425e-06,
+      "loss": 0.447,
+      "step": 24485
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4965250742567125e-06,
+      "loss": 0.463,
+      "step": 24486
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.495948323096001e-06,
+      "loss": 0.488,
+      "step": 24487
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4953715719352895e-06,
+      "loss": 0.3375,
+      "step": 24488
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.494794820774577e-06,
+      "loss": 0.4801,
+      "step": 24489
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.494218069613866e-06,
+      "loss": 0.4704,
+      "step": 24490
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.493641318453154e-06,
+      "loss": 0.567,
+      "step": 24491
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.493064567292442e-06,
+      "loss": 0.4211,
+      "step": 24492
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4924878161317305e-06,
+      "loss": 0.3936,
+      "step": 24493
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.491911064971019e-06,
+      "loss": 0.3783,
+      "step": 24494
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.491334313810307e-06,
+      "loss": 0.3549,
+      "step": 24495
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.490757562649595e-06,
+      "loss": 0.3699,
+      "step": 24496
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.490180811488883e-06,
+      "loss": 0.3859,
+      "step": 24497
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4896040603281715e-06,
+      "loss": 0.4435,
+      "step": 24498
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.489027309167461e-06,
+      "loss": 0.3057,
+      "step": 24499
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.488450558006748e-06,
+      "loss": 0.4289,
+      "step": 24500
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.487873806846037e-06,
+      "loss": 0.3778,
+      "step": 24501
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.487297055685326e-06,
+      "loss": 0.4398,
+      "step": 24502
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.486720304524613e-06,
+      "loss": 0.3635,
+      "step": 24503
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.486143553363902e-06,
+      "loss": 0.297,
+      "step": 24504
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.48556680220319e-06,
+      "loss": 0.4768,
+      "step": 24505
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.484990051042478e-06,
+      "loss": 0.4292,
+      "step": 24506
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4844132998817666e-06,
+      "loss": 0.4246,
+      "step": 24507
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.483836548721055e-06,
+      "loss": 0.4035,
+      "step": 24508
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.483259797560343e-06,
+      "loss": 0.2885,
+      "step": 24509
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.482683046399631e-06,
+      "loss": 0.3339,
+      "step": 24510
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.48210629523892e-06,
+      "loss": 0.4402,
+      "step": 24511
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4815295440782075e-06,
+      "loss": 0.3151,
+      "step": 24512
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.480952792917496e-06,
+      "loss": 0.5482,
+      "step": 24513
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.480376041756784e-06,
+      "loss": 0.3736,
+      "step": 24514
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.479799290596073e-06,
+      "loss": 0.389,
+      "step": 24515
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.479222539435362e-06,
+      "loss": 0.441,
+      "step": 24516
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.478645788274649e-06,
+      "loss": 0.4655,
+      "step": 24517
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.478069037113938e-06,
+      "loss": 0.3829,
+      "step": 24518
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.477492285953226e-06,
+      "loss": 0.4354,
+      "step": 24519
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.476915534792514e-06,
+      "loss": 0.3639,
+      "step": 24520
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.476338783631803e-06,
+      "loss": 0.5783,
+      "step": 24521
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.475762032471091e-06,
+      "loss": 0.391,
+      "step": 24522
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.475185281310379e-06,
+      "loss": 0.5054,
+      "step": 24523
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.474608530149667e-06,
+      "loss": 0.379,
+      "step": 24524
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.474031778988956e-06,
+      "loss": 0.3608,
+      "step": 24525
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.473455027828244e-06,
+      "loss": 0.4161,
+      "step": 24526
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.472878276667532e-06,
+      "loss": 0.3788,
+      "step": 24527
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4723015255068215e-06,
+      "loss": 0.5225,
+      "step": 24528
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.471724774346108e-06,
+      "loss": 0.3764,
+      "step": 24529
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.471148023185398e-06,
+      "loss": 0.3177,
+      "step": 24530
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4705712720246846e-06,
+      "loss": 0.4226,
+      "step": 24531
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.469994520863974e-06,
+      "loss": 0.3824,
+      "step": 24532
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4694177697032625e-06,
+      "loss": 0.4538,
+      "step": 24533
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.46884101854255e-06,
+      "loss": 0.4757,
+      "step": 24534
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.468264267381839e-06,
+      "loss": 0.5545,
+      "step": 24535
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.467687516221127e-06,
+      "loss": 0.3469,
+      "step": 24536
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.467110765060415e-06,
+      "loss": 0.5045,
+      "step": 24537
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4665340138997034e-06,
+      "loss": 0.3949,
+      "step": 24538
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.465957262738992e-06,
+      "loss": 0.4146,
+      "step": 24539
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.46538051157828e-06,
+      "loss": 0.5052,
+      "step": 24540
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.464803760417568e-06,
+      "loss": 0.541,
+      "step": 24541
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.464227009256857e-06,
+      "loss": 0.4862,
+      "step": 24542
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.463650258096144e-06,
+      "loss": 0.4726,
+      "step": 24543
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.463073506935433e-06,
+      "loss": 0.4096,
+      "step": 24544
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.462496755774722e-06,
+      "loss": 0.3563,
+      "step": 24545
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.46192000461401e-06,
+      "loss": 0.4158,
+      "step": 24546
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4613432534532985e-06,
+      "loss": 0.4346,
+      "step": 24547
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460766502292586e-06,
+      "loss": 0.4592,
+      "step": 24548
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.460189751131875e-06,
+      "loss": 0.3956,
+      "step": 24549
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.459612999971163e-06,
+      "loss": 0.3914,
+      "step": 24550
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.459036248810451e-06,
+      "loss": 0.467,
+      "step": 24551
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4584594976497395e-06,
+      "loss": 0.3956,
+      "step": 24552
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.457882746489028e-06,
+      "loss": 0.383,
+      "step": 24553
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.457305995328316e-06,
+      "loss": 0.4684,
+      "step": 24554
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.456729244167604e-06,
+      "loss": 0.2983,
+      "step": 24555
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.456152493006893e-06,
+      "loss": 0.3745,
+      "step": 24556
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4555757418461805e-06,
+      "loss": 0.465,
+      "step": 24557
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.454998990685469e-06,
+      "loss": 0.4295,
+      "step": 24558
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.454422239524758e-06,
+      "loss": 0.3795,
+      "step": 24559
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.453845488364045e-06,
+      "loss": 0.4322,
+      "step": 24560
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.453268737203335e-06,
+      "loss": 0.4486,
+      "step": 24561
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.452691986042623e-06,
+      "loss": 0.3874,
+      "step": 24562
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.452115234881911e-06,
+      "loss": 0.4088,
+      "step": 24563
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.451538483721199e-06,
+      "loss": 0.4225,
+      "step": 24564
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.450961732560488e-06,
+      "loss": 0.4666,
+      "step": 24565
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4503849813997756e-06,
+      "loss": 0.45,
+      "step": 24566
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.449808230239064e-06,
+      "loss": 0.4756,
+      "step": 24567
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.449231479078352e-06,
+      "loss": 0.4413,
+      "step": 24568
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.44865472791764e-06,
+      "loss": 0.3544,
+      "step": 24569
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.448077976756929e-06,
+      "loss": 0.4182,
+      "step": 24570
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.4475012255962165e-06,
+      "loss": 0.3888,
+      "step": 24571
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.446924474435505e-06,
+      "loss": 0.3739,
+      "step": 24572
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.446347723274794e-06,
+      "loss": 0.3104,
+      "step": 24573
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.445770972114081e-06,
+      "loss": 0.3677,
+      "step": 24574
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.44519422095337e-06,
+      "loss": 0.421,
+      "step": 24575
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.444617469792659e-06,
+      "loss": 0.4786,
+      "step": 24576
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.444040718631947e-06,
+      "loss": 0.4184,
+      "step": 24577
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 6.443463967471235e-06,
+      "loss": 0.4921,
+      "step": 24578
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.442887216310524e-06,
+      "loss": 0.3049,
+      "step": 24579
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.442310465149812e-06,
+      "loss": 0.4588,
+      "step": 24580
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4417337139891e-06,
+      "loss": 0.4953,
+      "step": 24581
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.441156962828389e-06,
+      "loss": 0.6076,
+      "step": 24582
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440580211667676e-06,
+      "loss": 0.3723,
+      "step": 24583
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.440003460506965e-06,
+      "loss": 0.425,
+      "step": 24584
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.439426709346253e-06,
+      "loss": 0.2546,
+      "step": 24585
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.438849958185541e-06,
+      "loss": 0.5024,
+      "step": 24586
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.43827320702483e-06,
+      "loss": 0.5176,
+      "step": 24587
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.437696455864117e-06,
+      "loss": 0.3705,
+      "step": 24588
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.437119704703406e-06,
+      "loss": 0.3681,
+      "step": 24589
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.436542953542695e-06,
+      "loss": 0.4631,
+      "step": 24590
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.435966202381982e-06,
+      "loss": 0.4804,
+      "step": 24591
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4353894512212715e-06,
+      "loss": 0.417,
+      "step": 24592
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.43481270006056e-06,
+      "loss": 0.4315,
+      "step": 24593
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.434235948899848e-06,
+      "loss": 0.3461,
+      "step": 24594
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.433659197739136e-06,
+      "loss": 0.4146,
+      "step": 24595
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.433082446578425e-06,
+      "loss": 0.5111,
+      "step": 24596
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4325056954177124e-06,
+      "loss": 0.4765,
+      "step": 24597
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.431928944257001e-06,
+      "loss": 0.4829,
+      "step": 24598
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4313521930962895e-06,
+      "loss": 0.2993,
+      "step": 24599
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.430775441935577e-06,
+      "loss": 0.4631,
+      "step": 24600
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.430198690774866e-06,
+      "loss": 0.3284,
+      "step": 24601
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.429621939614153e-06,
+      "loss": 0.4418,
+      "step": 24602
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.429045188453442e-06,
+      "loss": 0.4217,
+      "step": 24603
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4284684372927305e-06,
+      "loss": 0.4253,
+      "step": 24604
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.427891686132018e-06,
+      "loss": 0.3708,
+      "step": 24605
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.427314934971307e-06,
+      "loss": 0.3832,
+      "step": 24606
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.426738183810596e-06,
+      "loss": 0.4827,
+      "step": 24607
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.426161432649884e-06,
+      "loss": 0.4295,
+      "step": 24608
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.425584681489172e-06,
+      "loss": 0.3224,
+      "step": 24609
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.425007930328461e-06,
+      "loss": 0.4325,
+      "step": 24610
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4244311791677485e-06,
+      "loss": 0.4257,
+      "step": 24611
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.423854428007037e-06,
+      "loss": 0.4223,
+      "step": 24612
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4232776768463256e-06,
+      "loss": 0.5264,
+      "step": 24613
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.422700925685613e-06,
+      "loss": 0.353,
+      "step": 24614
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.422124174524902e-06,
+      "loss": 0.478,
+      "step": 24615
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.42154742336419e-06,
+      "loss": 0.3926,
+      "step": 24616
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.420970672203478e-06,
+      "loss": 0.4656,
+      "step": 24617
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4203939210427665e-06,
+      "loss": 0.4225,
+      "step": 24618
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419817169882054e-06,
+      "loss": 0.3426,
+      "step": 24619
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.419240418721343e-06,
+      "loss": 0.3252,
+      "step": 24620
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.418663667560632e-06,
+      "loss": 0.5259,
+      "step": 24621
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.418086916399919e-06,
+      "loss": 0.3499,
+      "step": 24622
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.417510165239208e-06,
+      "loss": 0.3918,
+      "step": 24623
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.416933414078497e-06,
+      "loss": 0.5631,
+      "step": 24624
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4163566629177846e-06,
+      "loss": 0.4444,
+      "step": 24625
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.415779911757073e-06,
+      "loss": 0.3809,
+      "step": 24626
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.415203160596362e-06,
+      "loss": 0.3991,
+      "step": 24627
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.414626409435649e-06,
+      "loss": 0.3331,
+      "step": 24628
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.414049658274938e-06,
+      "loss": 0.3579,
+      "step": 24629
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.413472907114226e-06,
+      "loss": 0.2913,
+      "step": 24630
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.412896155953514e-06,
+      "loss": 0.5073,
+      "step": 24631
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.412319404792803e-06,
+      "loss": 0.4169,
+      "step": 24632
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.411742653632091e-06,
+      "loss": 0.4083,
+      "step": 24633
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.411165902471379e-06,
+      "loss": 0.4773,
+      "step": 24634
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.410589151310667e-06,
+      "loss": 0.4153,
+      "step": 24635
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.410012400149955e-06,
+      "loss": 0.373,
+      "step": 24636
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4094356489892436e-06,
+      "loss": 0.4144,
+      "step": 24637
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.408858897828533e-06,
+      "loss": 0.3662,
+      "step": 24638
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.408282146667821e-06,
+      "loss": 0.4283,
+      "step": 24639
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.407705395507109e-06,
+      "loss": 0.4376,
+      "step": 24640
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.407128644346398e-06,
+      "loss": 0.441,
+      "step": 24641
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.406551893185685e-06,
+      "loss": 0.4942,
+      "step": 24642
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.405975142024974e-06,
+      "loss": 0.3997,
+      "step": 24643
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.4053983908642624e-06,
+      "loss": 0.3592,
+      "step": 24644
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.40482163970355e-06,
+      "loss": 0.3854,
+      "step": 24645
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.404244888542839e-06,
+      "loss": 0.3846,
+      "step": 24646
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.403668137382127e-06,
+      "loss": 0.4078,
+      "step": 24647
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.403091386221415e-06,
+      "loss": 0.3833,
+      "step": 24648
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.402514635060703e-06,
+      "loss": 0.3191,
+      "step": 24649
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.401937883899993e-06,
+      "loss": 0.4201,
+      "step": 24650
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.40136113273928e-06,
+      "loss": 0.3817,
+      "step": 24651
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.400784381578569e-06,
+      "loss": 0.4372,
+      "step": 24652
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.400207630417856e-06,
+      "loss": 0.5077,
+      "step": 24653
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.399630879257145e-06,
+      "loss": 0.4586,
+      "step": 24654
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.399054128096434e-06,
+      "loss": 0.4535,
+      "step": 24655
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3984773769357214e-06,
+      "loss": 0.4365,
+      "step": 24656
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39790062577501e-06,
+      "loss": 0.3249,
+      "step": 24657
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3973238746142985e-06,
+      "loss": 0.6016,
+      "step": 24658
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.396747123453586e-06,
+      "loss": 0.5089,
+      "step": 24659
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.396170372292875e-06,
+      "loss": 0.5806,
+      "step": 24660
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.395593621132163e-06,
+      "loss": 0.4213,
+      "step": 24661
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.395016869971451e-06,
+      "loss": 0.463,
+      "step": 24662
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3944401188107395e-06,
+      "loss": 0.4143,
+      "step": 24663
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.393863367650028e-06,
+      "loss": 0.4346,
+      "step": 24664
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.393286616489316e-06,
+      "loss": 0.4728,
+      "step": 24665
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.392709865328604e-06,
+      "loss": 0.4179,
+      "step": 24666
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.392133114167894e-06,
+      "loss": 0.3814,
+      "step": 24667
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3915563630071804e-06,
+      "loss": 0.4539,
+      "step": 24668
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.39097961184647e-06,
+      "loss": 0.414,
+      "step": 24669
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3904028606857575e-06,
+      "loss": 0.3707,
+      "step": 24670
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.389826109525046e-06,
+      "loss": 0.4206,
+      "step": 24671
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3892493583643346e-06,
+      "loss": 0.5113,
+      "step": 24672
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.388672607203622e-06,
+      "loss": 0.4645,
+      "step": 24673
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.388095856042911e-06,
+      "loss": 0.4914,
+      "step": 24674
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.387519104882199e-06,
+      "loss": 0.418,
+      "step": 24675
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.386942353721487e-06,
+      "loss": 0.4121,
+      "step": 24676
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3863656025607755e-06,
+      "loss": 0.4268,
+      "step": 24677
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.385788851400064e-06,
+      "loss": 0.4969,
+      "step": 24678
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.385212100239352e-06,
+      "loss": 0.4812,
+      "step": 24679
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.38463534907864e-06,
+      "loss": 0.3726,
+      "step": 24680
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.38405859791793e-06,
+      "loss": 0.3262,
+      "step": 24681
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3834818467572165e-06,
+      "loss": 0.4435,
+      "step": 24682
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.382905095596506e-06,
+      "loss": 0.3055,
+      "step": 24683
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.382328344435794e-06,
+      "loss": 0.4965,
+      "step": 24684
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.381751593275082e-06,
+      "loss": 0.3699,
+      "step": 24685
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.381174842114371e-06,
+      "loss": 0.4561,
+      "step": 24686
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.380598090953659e-06,
+      "loss": 0.4038,
+      "step": 24687
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.380021339792947e-06,
+      "loss": 0.3893,
+      "step": 24688
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.379444588632235e-06,
+      "loss": 0.523,
+      "step": 24689
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378867837471523e-06,
+      "loss": 0.4778,
+      "step": 24690
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.378291086310812e-06,
+      "loss": 0.2758,
+      "step": 24691
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3777143351501e-06,
+      "loss": 0.5129,
+      "step": 24692
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.377137583989388e-06,
+      "loss": 0.415,
+      "step": 24693
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.376560832828676e-06,
+      "loss": 0.4892,
+      "step": 24694
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.375984081667965e-06,
+      "loss": 0.5261,
+      "step": 24695
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3754073305072526e-06,
+      "loss": 0.4278,
+      "step": 24696
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.374830579346541e-06,
+      "loss": 0.497,
+      "step": 24697
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3742538281858305e-06,
+      "loss": 0.3788,
+      "step": 24698
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.373677077025118e-06,
+      "loss": 0.4328,
+      "step": 24699
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.373100325864407e-06,
+      "loss": 0.4253,
+      "step": 24700
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.372523574703695e-06,
+      "loss": 0.3696,
+      "step": 24701
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.371946823542983e-06,
+      "loss": 0.561,
+      "step": 24702
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3713700723822714e-06,
+      "loss": 0.4622,
+      "step": 24703
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.37079332122156e-06,
+      "loss": 0.5512,
+      "step": 24704
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.370216570060848e-06,
+      "loss": 0.3459,
+      "step": 24705
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.369639818900136e-06,
+      "loss": 0.3923,
+      "step": 24706
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.369063067739424e-06,
+      "loss": 0.4949,
+      "step": 24707
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.368486316578712e-06,
+      "loss": 0.4772,
+      "step": 24708
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.367909565418001e-06,
+      "loss": 0.3806,
+      "step": 24709
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.367332814257289e-06,
+      "loss": 0.4877,
+      "step": 24710
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.366756063096577e-06,
+      "loss": 0.4846,
+      "step": 24711
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3661793119358665e-06,
+      "loss": 0.4415,
+      "step": 24712
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.365602560775153e-06,
+      "loss": 0.4782,
+      "step": 24713
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.365025809614443e-06,
+      "loss": 0.5062,
+      "step": 24714
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.364449058453731e-06,
+      "loss": 0.4764,
+      "step": 24715
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.363872307293019e-06,
+      "loss": 0.4585,
+      "step": 24716
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3632955561323075e-06,
+      "loss": 0.4135,
+      "step": 24717
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.362718804971596e-06,
+      "loss": 0.4188,
+      "step": 24718
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.362142053810884e-06,
+      "loss": 0.3319,
+      "step": 24719
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.361565302650172e-06,
+      "loss": 0.4316,
+      "step": 24720
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.360988551489461e-06,
+      "loss": 0.467,
+      "step": 24721
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3604118003287485e-06,
+      "loss": 0.5098,
+      "step": 24722
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.359835049168037e-06,
+      "loss": 0.3743,
+      "step": 24723
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.359258298007325e-06,
+      "loss": 0.4326,
+      "step": 24724
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358681546846613e-06,
+      "loss": 0.4465,
+      "step": 24725
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.358104795685902e-06,
+      "loss": 0.4519,
+      "step": 24726
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3575280445251894e-06,
+      "loss": 0.4443,
+      "step": 24727
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.356951293364478e-06,
+      "loss": 0.4003,
+      "step": 24728
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.356374542203767e-06,
+      "loss": 0.3936,
+      "step": 24729
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.355797791043055e-06,
+      "loss": 0.3256,
+      "step": 24730
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3552210398823436e-06,
+      "loss": 0.3968,
+      "step": 24731
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.354644288721632e-06,
+      "loss": 0.4905,
+      "step": 24732
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.35406753756092e-06,
+      "loss": 0.413,
+      "step": 24733
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.353490786400208e-06,
+      "loss": 0.4157,
+      "step": 24734
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.352914035239497e-06,
+      "loss": 0.4703,
+      "step": 24735
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3523372840787845e-06,
+      "loss": 0.5003,
+      "step": 24736
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.351760532918073e-06,
+      "loss": 0.4741,
+      "step": 24737
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.351183781757362e-06,
+      "loss": 0.3834,
+      "step": 24738
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.350607030596649e-06,
+      "loss": 0.4057,
+      "step": 24739
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.350030279435938e-06,
+      "loss": 0.4136,
+      "step": 24740
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.3494535282752255e-06,
+      "loss": 0.3785,
+      "step": 24741
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.348876777114514e-06,
+      "loss": 0.4382,
+      "step": 24742
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.348300025953803e-06,
+      "loss": 0.4634,
+      "step": 24743
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.34772327479309e-06,
+      "loss": 0.4059,
+      "step": 24744
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.34714652363238e-06,
+      "loss": 0.3758,
+      "step": 24745
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.346569772471668e-06,
+      "loss": 0.4584,
+      "step": 24746
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.345993021310956e-06,
+      "loss": 0.4483,
+      "step": 24747
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.345416270150244e-06,
+      "loss": 0.4405,
+      "step": 24748
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.344839518989533e-06,
+      "loss": 0.4812,
+      "step": 24749
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.344262767828821e-06,
+      "loss": 0.5202,
+      "step": 24750
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.343686016668109e-06,
+      "loss": 0.5502,
+      "step": 24751
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.343109265507398e-06,
+      "loss": 0.4519,
+      "step": 24752
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.342532514346685e-06,
+      "loss": 0.4787,
+      "step": 24753
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.341955763185974e-06,
+      "loss": 0.4677,
+      "step": 24754
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.341379012025262e-06,
+      "loss": 0.5308,
+      "step": 24755
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 6.34080226086455e-06,
+      "loss": 0.3194,
+      "step": 24756
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.340225509703839e-06,
+      "loss": 0.4648,
+      "step": 24757
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.339648758543126e-06,
+      "loss": 0.4252,
+      "step": 24758
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.339072007382415e-06,
+      "loss": 0.5364,
+      "step": 24759
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.338495256221704e-06,
+      "loss": 0.3949,
+      "step": 24760
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.337918505060992e-06,
+      "loss": 0.3861,
+      "step": 24761
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3373417539002804e-06,
+      "loss": 0.2473,
+      "step": 24762
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.336765002739569e-06,
+      "loss": 0.4848,
+      "step": 24763
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.336188251578857e-06,
+      "loss": 0.392,
+      "step": 24764
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.335611500418145e-06,
+      "loss": 0.3643,
+      "step": 24765
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.335034749257434e-06,
+      "loss": 0.5084,
+      "step": 24766
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.334457998096721e-06,
+      "loss": 0.5384,
+      "step": 24767
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.33388124693601e-06,
+      "loss": 0.5121,
+      "step": 24768
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3333044957752985e-06,
+      "loss": 0.4161,
+      "step": 24769
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.332727744614586e-06,
+      "loss": 0.4537,
+      "step": 24770
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.332150993453875e-06,
+      "loss": 0.4628,
+      "step": 24771
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.331574242293164e-06,
+      "loss": 0.5431,
+      "step": 24772
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.330997491132451e-06,
+      "loss": 0.384,
+      "step": 24773
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.33042073997174e-06,
+      "loss": 0.3989,
+      "step": 24774
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.329843988811027e-06,
+      "loss": 0.3667,
+      "step": 24775
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3292672376503165e-06,
+      "loss": 0.3769,
+      "step": 24776
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.328690486489605e-06,
+      "loss": 0.4579,
+      "step": 24777
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.328113735328893e-06,
+      "loss": 0.3055,
+      "step": 24778
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.327536984168181e-06,
+      "loss": 0.4968,
+      "step": 24779
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.32696023300747e-06,
+      "loss": 0.3899,
+      "step": 24780
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3263834818467575e-06,
+      "loss": 0.4264,
+      "step": 24781
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.325806730686046e-06,
+      "loss": 0.3623,
+      "step": 24782
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3252299795253345e-06,
+      "loss": 0.4155,
+      "step": 24783
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.324653228364622e-06,
+      "loss": 0.4937,
+      "step": 24784
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.324076477203911e-06,
+      "loss": 0.545,
+      "step": 24785
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.323499726043199e-06,
+      "loss": 0.4546,
+      "step": 24786
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.322922974882487e-06,
+      "loss": 0.319,
+      "step": 24787
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3223462237217755e-06,
+      "loss": 0.35,
+      "step": 24788
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.321769472561065e-06,
+      "loss": 0.3537,
+      "step": 24789
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.321192721400352e-06,
+      "loss": 0.4267,
+      "step": 24790
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.320615970239641e-06,
+      "loss": 0.4301,
+      "step": 24791
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.320039219078929e-06,
+      "loss": 0.3723,
+      "step": 24792
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.319462467918217e-06,
+      "loss": 0.2449,
+      "step": 24793
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.318885716757506e-06,
+      "loss": 0.5106,
+      "step": 24794
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3183089655967935e-06,
+      "loss": 0.3986,
+      "step": 24795
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.317732214436082e-06,
+      "loss": 0.2795,
+      "step": 24796
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.317155463275371e-06,
+      "loss": 0.5106,
+      "step": 24797
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.316578712114658e-06,
+      "loss": 0.4593,
+      "step": 24798
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.316001960953947e-06,
+      "loss": 0.4494,
+      "step": 24799
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.315425209793235e-06,
+      "loss": 0.4857,
+      "step": 24800
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.314848458632523e-06,
+      "loss": 0.4035,
+      "step": 24801
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3142717074718116e-06,
+      "loss": 0.3595,
+      "step": 24802
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.313694956311101e-06,
+      "loss": 0.3655,
+      "step": 24803
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.313118205150388e-06,
+      "loss": 0.377,
+      "step": 24804
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.312541453989677e-06,
+      "loss": 0.43,
+      "step": 24805
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.311964702828966e-06,
+      "loss": 0.3363,
+      "step": 24806
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.311387951668253e-06,
+      "loss": 0.3594,
+      "step": 24807
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.310811200507542e-06,
+      "loss": 0.4546,
+      "step": 24808
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.31023444934683e-06,
+      "loss": 0.3769,
+      "step": 24809
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.309657698186118e-06,
+      "loss": 0.3644,
+      "step": 24810
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.309080947025407e-06,
+      "loss": 0.4126,
+      "step": 24811
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.308504195864694e-06,
+      "loss": 0.4149,
+      "step": 24812
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.307927444703983e-06,
+      "loss": 0.4199,
+      "step": 24813
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.307350693543271e-06,
+      "loss": 0.2693,
+      "step": 24814
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.306773942382559e-06,
+      "loss": 0.5051,
+      "step": 24815
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.306197191221848e-06,
+      "loss": 0.4582,
+      "step": 24816
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.305620440061136e-06,
+      "loss": 0.4344,
+      "step": 24817
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.305043688900424e-06,
+      "loss": 0.5184,
+      "step": 24818
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.304466937739712e-06,
+      "loss": 0.3738,
+      "step": 24819
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.303890186579002e-06,
+      "loss": 0.3266,
+      "step": 24820
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.303313435418289e-06,
+      "loss": 0.4238,
+      "step": 24821
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.302736684257578e-06,
+      "loss": 0.5324,
+      "step": 24822
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.3021599330968665e-06,
+      "loss": 0.4403,
+      "step": 24823
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.301583181936154e-06,
+      "loss": 0.4974,
+      "step": 24824
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.301006430775443e-06,
+      "loss": 0.4,
+      "step": 24825
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.300429679614731e-06,
+      "loss": 0.3647,
+      "step": 24826
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.299852928454019e-06,
+      "loss": 0.425,
+      "step": 24827
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2992761772933075e-06,
+      "loss": 0.3183,
+      "step": 24828
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.298699426132595e-06,
+      "loss": 0.4202,
+      "step": 24829
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.298122674971884e-06,
+      "loss": 0.2782,
+      "step": 24830
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.297545923811172e-06,
+      "loss": 0.3691,
+      "step": 24831
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.29696917265046e-06,
+      "loss": 0.385,
+      "step": 24832
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2963924214897484e-06,
+      "loss": 0.3745,
+      "step": 24833
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.295815670329038e-06,
+      "loss": 0.4542,
+      "step": 24834
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.295238919168325e-06,
+      "loss": 0.4505,
+      "step": 24835
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.294662168007614e-06,
+      "loss": 0.3763,
+      "step": 24836
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2940854168469026e-06,
+      "loss": 0.4766,
+      "step": 24837
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.29350866568619e-06,
+      "loss": 0.4096,
+      "step": 24838
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.292931914525479e-06,
+      "loss": 0.3539,
+      "step": 24839
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.292355163364767e-06,
+      "loss": 0.6296,
+      "step": 24840
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.291778412204055e-06,
+      "loss": 0.5112,
+      "step": 24841
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2912016610433435e-06,
+      "loss": 0.2884,
+      "step": 24842
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.290624909882632e-06,
+      "loss": 0.465,
+      "step": 24843
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.29004815872192e-06,
+      "loss": 0.4476,
+      "step": 24844
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.289471407561208e-06,
+      "loss": 0.3974,
+      "step": 24845
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.288894656400496e-06,
+      "loss": 0.4477,
+      "step": 24846
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2883179052397845e-06,
+      "loss": 0.3966,
+      "step": 24847
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.287741154079073e-06,
+      "loss": 0.4551,
+      "step": 24848
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.287164402918361e-06,
+      "loss": 0.38,
+      "step": 24849
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.286587651757649e-06,
+      "loss": 0.5022,
+      "step": 24850
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.286010900596939e-06,
+      "loss": 0.4737,
+      "step": 24851
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2854341494362255e-06,
+      "loss": 0.4222,
+      "step": 24852
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.284857398275515e-06,
+      "loss": 0.4364,
+      "step": 24853
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.284280647114803e-06,
+      "loss": 0.2888,
+      "step": 24854
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.283703895954091e-06,
+      "loss": 0.4162,
+      "step": 24855
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.28312714479338e-06,
+      "loss": 0.4518,
+      "step": 24856
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.282550393632668e-06,
+      "loss": 0.4889,
+      "step": 24857
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.281973642471956e-06,
+      "loss": 0.4689,
+      "step": 24858
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.281396891311244e-06,
+      "loss": 0.3149,
+      "step": 24859
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.280820140150533e-06,
+      "loss": 0.4395,
+      "step": 24860
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2802433889898205e-06,
+      "loss": 0.4131,
+      "step": 24861
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.279666637829109e-06,
+      "loss": 0.4165,
+      "step": 24862
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.279089886668397e-06,
+      "loss": 0.4702,
+      "step": 24863
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.278513135507685e-06,
+      "loss": 0.4749,
+      "step": 24864
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.277936384346975e-06,
+      "loss": 0.4996,
+      "step": 24865
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2773596331862615e-06,
+      "loss": 0.3616,
+      "step": 24866
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.276782882025551e-06,
+      "loss": 0.4815,
+      "step": 24867
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2762061308648394e-06,
+      "loss": 0.3152,
+      "step": 24868
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275629379704127e-06,
+      "loss": 0.4211,
+      "step": 24869
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.275052628543416e-06,
+      "loss": 0.522,
+      "step": 24870
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.274475877382704e-06,
+      "loss": 0.4353,
+      "step": 24871
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.273899126221992e-06,
+      "loss": 0.5319,
+      "step": 24872
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.27332237506128e-06,
+      "loss": 0.4117,
+      "step": 24873
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.272745623900569e-06,
+      "loss": 0.4844,
+      "step": 24874
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.272168872739857e-06,
+      "loss": 0.3932,
+      "step": 24875
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.271592121579145e-06,
+      "loss": 0.4231,
+      "step": 24876
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.271015370418434e-06,
+      "loss": 0.3775,
+      "step": 24877
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.270438619257721e-06,
+      "loss": 0.5136,
+      "step": 24878
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.26986186809701e-06,
+      "loss": 0.4246,
+      "step": 24879
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.269285116936298e-06,
+      "loss": 0.4365,
+      "step": 24880
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.268708365775586e-06,
+      "loss": 0.4442,
+      "step": 24881
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2681316146148755e-06,
+      "loss": 0.4259,
+      "step": 24882
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.267554863454163e-06,
+      "loss": 0.3807,
+      "step": 24883
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.266978112293452e-06,
+      "loss": 0.5458,
+      "step": 24884
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.26640136113274e-06,
+      "loss": 0.4793,
+      "step": 24885
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.265824609972028e-06,
+      "loss": 0.4811,
+      "step": 24886
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2652478588113165e-06,
+      "loss": 0.2906,
+      "step": 24887
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.264671107650605e-06,
+      "loss": 0.4022,
+      "step": 24888
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.264094356489893e-06,
+      "loss": 0.3946,
+      "step": 24889
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.263517605329181e-06,
+      "loss": 0.3637,
+      "step": 24890
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.26294085416847e-06,
+      "loss": 0.4878,
+      "step": 24891
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.262364103007757e-06,
+      "loss": 0.4441,
+      "step": 24892
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.261787351847046e-06,
+      "loss": 0.4004,
+      "step": 24893
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2612106006863345e-06,
+      "loss": 0.4935,
+      "step": 24894
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.260633849525622e-06,
+      "loss": 0.3631,
+      "step": 24895
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2600570983649115e-06,
+      "loss": 0.4345,
+      "step": 24896
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.259480347204198e-06,
+      "loss": 0.5714,
+      "step": 24897
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.258903596043488e-06,
+      "loss": 0.3446,
+      "step": 24898
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.258326844882776e-06,
+      "loss": 0.332,
+      "step": 24899
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.257750093722064e-06,
+      "loss": 0.4458,
+      "step": 24900
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2571733425613525e-06,
+      "loss": 0.3796,
+      "step": 24901
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.256596591400641e-06,
+      "loss": 0.3805,
+      "step": 24902
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.256019840239929e-06,
+      "loss": 0.4814,
+      "step": 24903
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.255443089079217e-06,
+      "loss": 0.4886,
+      "step": 24904
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.254866337918506e-06,
+      "loss": 0.4427,
+      "step": 24905
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2542895867577935e-06,
+      "loss": 0.3606,
+      "step": 24906
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.253712835597082e-06,
+      "loss": 0.3381,
+      "step": 24907
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2531360844363705e-06,
+      "loss": 0.5581,
+      "step": 24908
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.252559333275658e-06,
+      "loss": 0.5034,
+      "step": 24909
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.251982582114947e-06,
+      "loss": 0.4812,
+      "step": 24910
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.251405830954236e-06,
+      "loss": 0.4345,
+      "step": 24911
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.250829079793523e-06,
+      "loss": 0.4408,
+      "step": 24912
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.250252328632812e-06,
+      "loss": 0.3568,
+      "step": 24913
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2496755774721e-06,
+      "loss": 0.48,
+      "step": 24914
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.249098826311389e-06,
+      "loss": 0.5059,
+      "step": 24915
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.248522075150677e-06,
+      "loss": 0.502,
+      "step": 24916
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.247945323989965e-06,
+      "loss": 0.4469,
+      "step": 24917
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.247368572829253e-06,
+      "loss": 0.491,
+      "step": 24918
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.246791821668542e-06,
+      "loss": 0.529,
+      "step": 24919
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.2462150705078295e-06,
+      "loss": 0.3274,
+      "step": 24920
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.245638319347118e-06,
+      "loss": 0.3866,
+      "step": 24921
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.245061568186407e-06,
+      "loss": 0.3666,
+      "step": 24922
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.244484817025694e-06,
+      "loss": 0.4574,
+      "step": 24923
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.243908065864983e-06,
+      "loss": 0.501,
+      "step": 24924
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.243331314704272e-06,
+      "loss": 0.4111,
+      "step": 24925
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.242754563543559e-06,
+      "loss": 0.4355,
+      "step": 24926
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.242177812382848e-06,
+      "loss": 0.3565,
+      "step": 24927
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.241601061222137e-06,
+      "loss": 0.3464,
+      "step": 24928
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.241024310061425e-06,
+      "loss": 0.4292,
+      "step": 24929
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.240447558900713e-06,
+      "loss": 0.3295,
+      "step": 24930
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.239870807740001e-06,
+      "loss": 0.3847,
+      "step": 24931
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.239294056579289e-06,
+      "loss": 0.5355,
+      "step": 24932
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.238717305418578e-06,
+      "loss": 0.4001,
+      "step": 24933
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.238140554257866e-06,
+      "loss": 0.4676,
+      "step": 24934
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 6.237563803097154e-06,
+      "loss": 0.5057,
+      "step": 24935
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.236987051936443e-06,
+      "loss": 0.4881,
+      "step": 24936
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.23641030077573e-06,
+      "loss": 0.4428,
+      "step": 24937
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.235833549615019e-06,
+      "loss": 0.4971,
+      "step": 24938
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.235256798454307e-06,
+      "loss": 0.5071,
+      "step": 24939
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234680047293595e-06,
+      "loss": 0.5215,
+      "step": 24940
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.234103296132884e-06,
+      "loss": 0.5493,
+      "step": 24941
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.233526544972173e-06,
+      "loss": 0.4022,
+      "step": 24942
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.23294979381146e-06,
+      "loss": 0.4839,
+      "step": 24943
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.232373042650749e-06,
+      "loss": 0.5006,
+      "step": 24944
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.231796291490038e-06,
+      "loss": 0.4035,
+      "step": 24945
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2312195403293255e-06,
+      "loss": 0.436,
+      "step": 24946
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.230642789168614e-06,
+      "loss": 0.5468,
+      "step": 24947
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2300660380079025e-06,
+      "loss": 0.405,
+      "step": 24948
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.22948928684719e-06,
+      "loss": 0.4976,
+      "step": 24949
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.228912535686479e-06,
+      "loss": 0.3828,
+      "step": 24950
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.228335784525766e-06,
+      "loss": 0.4813,
+      "step": 24951
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.227759033365055e-06,
+      "loss": 0.364,
+      "step": 24952
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2271822822043435e-06,
+      "loss": 0.4609,
+      "step": 24953
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.226605531043631e-06,
+      "loss": 0.4018,
+      "step": 24954
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.22602877988292e-06,
+      "loss": 0.4345,
+      "step": 24955
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.225452028722209e-06,
+      "loss": 0.4051,
+      "step": 24956
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.224875277561496e-06,
+      "loss": 0.4888,
+      "step": 24957
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.224298526400785e-06,
+      "loss": 0.5114,
+      "step": 24958
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.223721775240074e-06,
+      "loss": 0.416,
+      "step": 24959
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2231450240793615e-06,
+      "loss": 0.4504,
+      "step": 24960
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.22256827291865e-06,
+      "loss": 0.4919,
+      "step": 24961
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.221991521757939e-06,
+      "loss": 0.3851,
+      "step": 24962
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.221414770597226e-06,
+      "loss": 0.4601,
+      "step": 24963
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.220838019436515e-06,
+      "loss": 0.3448,
+      "step": 24964
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.220261268275803e-06,
+      "loss": 0.3995,
+      "step": 24965
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.219684517115091e-06,
+      "loss": 0.4646,
+      "step": 24966
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2191077659543795e-06,
+      "loss": 0.6374,
+      "step": 24967
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.218531014793667e-06,
+      "loss": 0.3979,
+      "step": 24968
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.217954263632956e-06,
+      "loss": 0.4266,
+      "step": 24969
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.217377512472244e-06,
+      "loss": 0.49,
+      "step": 24970
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.216800761311532e-06,
+      "loss": 0.4755,
+      "step": 24971
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2162240101508205e-06,
+      "loss": 0.4407,
+      "step": 24972
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.21564725899011e-06,
+      "loss": 0.4693,
+      "step": 24973
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.215070507829397e-06,
+      "loss": 0.4268,
+      "step": 24974
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.214493756668686e-06,
+      "loss": 0.4987,
+      "step": 24975
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213917005507975e-06,
+      "loss": 0.3866,
+      "step": 24976
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.213340254347262e-06,
+      "loss": 0.4785,
+      "step": 24977
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.212763503186551e-06,
+      "loss": 0.4839,
+      "step": 24978
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.212186752025839e-06,
+      "loss": 0.3659,
+      "step": 24979
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.211610000865127e-06,
+      "loss": 0.3549,
+      "step": 24980
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.211033249704416e-06,
+      "loss": 0.4857,
+      "step": 24981
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.210456498543704e-06,
+      "loss": 0.3732,
+      "step": 24982
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.209879747382992e-06,
+      "loss": 0.3552,
+      "step": 24983
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.20930299622228e-06,
+      "loss": 0.4241,
+      "step": 24984
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.208726245061568e-06,
+      "loss": 0.3353,
+      "step": 24985
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2081494939008566e-06,
+      "loss": 0.3988,
+      "step": 24986
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.207572742740146e-06,
+      "loss": 0.437,
+      "step": 24987
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.206995991579433e-06,
+      "loss": 0.3786,
+      "step": 24988
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.206419240418722e-06,
+      "loss": 0.3628,
+      "step": 24989
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.205842489258011e-06,
+      "loss": 0.498,
+      "step": 24990
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.205265738097298e-06,
+      "loss": 0.4475,
+      "step": 24991
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.204688986936587e-06,
+      "loss": 0.3763,
+      "step": 24992
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.2041122357758754e-06,
+      "loss": 0.4877,
+      "step": 24993
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.203535484615163e-06,
+      "loss": 0.3799,
+      "step": 24994
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.202958733454452e-06,
+      "loss": 0.4367,
+      "step": 24995
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.20238198229374e-06,
+      "loss": 0.4655,
+      "step": 24996
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.201805231133028e-06,
+      "loss": 0.4821,
+      "step": 24997
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.201228479972316e-06,
+      "loss": 0.4425,
+      "step": 24998
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.200651728811605e-06,
+      "loss": 0.3893,
+      "step": 24999
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.200074977650893e-06,
+      "loss": 0.3239,
+      "step": 25000
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.199498226490181e-06,
+      "loss": 0.467,
+      "step": 25001
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.198921475329469e-06,
+      "loss": 0.4039,
+      "step": 25002
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.198344724168757e-06,
+      "loss": 0.3127,
+      "step": 25003
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.197767973008047e-06,
+      "loss": 0.3936,
+      "step": 25004
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.197191221847334e-06,
+      "loss": 0.464,
+      "step": 25005
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.196614470686623e-06,
+      "loss": 0.389,
+      "step": 25006
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1960377195259115e-06,
+      "loss": 0.3992,
+      "step": 25007
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.195460968365199e-06,
+      "loss": 0.442,
+      "step": 25008
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.194884217204488e-06,
+      "loss": 0.5281,
+      "step": 25009
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.194307466043776e-06,
+      "loss": 0.3458,
+      "step": 25010
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.193730714883064e-06,
+      "loss": 0.4785,
+      "step": 25011
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1931539637223525e-06,
+      "loss": 0.3695,
+      "step": 25012
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.192577212561641e-06,
+      "loss": 0.3147,
+      "step": 25013
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.192000461400929e-06,
+      "loss": 0.454,
+      "step": 25014
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.191423710240217e-06,
+      "loss": 0.4777,
+      "step": 25015
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.190846959079506e-06,
+      "loss": 0.5814,
+      "step": 25016
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1902702079187934e-06,
+      "loss": 0.4907,
+      "step": 25017
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.189693456758083e-06,
+      "loss": 0.4449,
+      "step": 25018
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.18911670559737e-06,
+      "loss": 0.5256,
+      "step": 25019
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.188539954436659e-06,
+      "loss": 0.4243,
+      "step": 25020
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1879632032759476e-06,
+      "loss": 0.5619,
+      "step": 25021
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.187386452115235e-06,
+      "loss": 0.4148,
+      "step": 25022
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.186809700954524e-06,
+      "loss": 0.316,
+      "step": 25023
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.186232949793812e-06,
+      "loss": 0.4,
+      "step": 25024
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1856561986331e-06,
+      "loss": 0.4894,
+      "step": 25025
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1850794474723885e-06,
+      "loss": 0.3147,
+      "step": 25026
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.184502696311677e-06,
+      "loss": 0.2903,
+      "step": 25027
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.183925945150965e-06,
+      "loss": 0.3585,
+      "step": 25028
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.183349193990253e-06,
+      "loss": 0.4402,
+      "step": 25029
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.182772442829542e-06,
+      "loss": 0.3831,
+      "step": 25030
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1821956916688295e-06,
+      "loss": 0.3693,
+      "step": 25031
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.181618940508118e-06,
+      "loss": 0.4212,
+      "step": 25032
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.181042189347407e-06,
+      "loss": 0.4718,
+      "step": 25033
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.180465438186694e-06,
+      "loss": 0.4954,
+      "step": 25034
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.179888687025984e-06,
+      "loss": 0.4048,
+      "step": 25035
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1793119358652705e-06,
+      "loss": 0.4367,
+      "step": 25036
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17873518470456e-06,
+      "loss": 0.4193,
+      "step": 25037
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.178158433543848e-06,
+      "loss": 0.4424,
+      "step": 25038
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.177581682383136e-06,
+      "loss": 0.4271,
+      "step": 25039
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.177004931222425e-06,
+      "loss": 0.4612,
+      "step": 25040
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.176428180061713e-06,
+      "loss": 0.3708,
+      "step": 25041
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.175851428901001e-06,
+      "loss": 0.388,
+      "step": 25042
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.175274677740289e-06,
+      "loss": 0.522,
+      "step": 25043
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.174697926579578e-06,
+      "loss": 0.4114,
+      "step": 25044
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1741211754188656e-06,
+      "loss": 0.4317,
+      "step": 25045
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.173544424258154e-06,
+      "loss": 0.3929,
+      "step": 25046
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.172967673097443e-06,
+      "loss": 0.4403,
+      "step": 25047
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17239092193673e-06,
+      "loss": 0.423,
+      "step": 25048
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.17181417077602e-06,
+      "loss": 0.5279,
+      "step": 25049
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.171237419615308e-06,
+      "loss": 0.3802,
+      "step": 25050
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.170660668454596e-06,
+      "loss": 0.4656,
+      "step": 25051
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1700839172938844e-06,
+      "loss": 0.5023,
+      "step": 25052
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.169507166133172e-06,
+      "loss": 0.4221,
+      "step": 25053
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.168930414972461e-06,
+      "loss": 0.339,
+      "step": 25054
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.168353663811749e-06,
+      "loss": 0.3819,
+      "step": 25055
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.167776912651037e-06,
+      "loss": 0.5185,
+      "step": 25056
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.167200161490325e-06,
+      "loss": 0.4483,
+      "step": 25057
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.166623410329614e-06,
+      "loss": 0.2979,
+      "step": 25058
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.166046659168902e-06,
+      "loss": 0.4685,
+      "step": 25059
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.16546990800819e-06,
+      "loss": 0.5241,
+      "step": 25060
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.164893156847479e-06,
+      "loss": 0.4313,
+      "step": 25061
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.164316405686766e-06,
+      "loss": 0.4516,
+      "step": 25062
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.163739654526055e-06,
+      "loss": 0.4564,
+      "step": 25063
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.163162903365344e-06,
+      "loss": 0.3643,
+      "step": 25064
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.162586152204631e-06,
+      "loss": 0.3727,
+      "step": 25065
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1620094010439205e-06,
+      "loss": 0.4425,
+      "step": 25066
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.161432649883209e-06,
+      "loss": 0.4496,
+      "step": 25067
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.160855898722497e-06,
+      "loss": 0.4572,
+      "step": 25068
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.160279147561785e-06,
+      "loss": 0.3944,
+      "step": 25069
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.159702396401074e-06,
+      "loss": 0.3675,
+      "step": 25070
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1591256452403615e-06,
+      "loss": 0.4171,
+      "step": 25071
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.15854889407965e-06,
+      "loss": 0.4117,
+      "step": 25072
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.157972142918938e-06,
+      "loss": 0.4439,
+      "step": 25073
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.157395391758226e-06,
+      "loss": 0.4615,
+      "step": 25074
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.156818640597515e-06,
+      "loss": 0.4273,
+      "step": 25075
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1562418894368024e-06,
+      "loss": 0.4626,
+      "step": 25076
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.155665138276091e-06,
+      "loss": 0.5079,
+      "step": 25077
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1550883871153795e-06,
+      "loss": 0.3676,
+      "step": 25078
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.154511635954667e-06,
+      "loss": 0.422,
+      "step": 25079
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1539348847939566e-06,
+      "loss": 0.4651,
+      "step": 25080
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.153358133633245e-06,
+      "loss": 0.4274,
+      "step": 25081
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.152781382472533e-06,
+      "loss": 0.5084,
+      "step": 25082
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.152204631311821e-06,
+      "loss": 0.4397,
+      "step": 25083
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.15162788015111e-06,
+      "loss": 0.4876,
+      "step": 25084
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1510511289903975e-06,
+      "loss": 0.4373,
+      "step": 25085
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.150474377829686e-06,
+      "loss": 0.3215,
+      "step": 25086
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.149897626668975e-06,
+      "loss": 0.4768,
+      "step": 25087
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.149320875508262e-06,
+      "loss": 0.2958,
+      "step": 25088
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.148744124347551e-06,
+      "loss": 0.3795,
+      "step": 25089
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1481673731868385e-06,
+      "loss": 0.4675,
+      "step": 25090
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.147590622026127e-06,
+      "loss": 0.3731,
+      "step": 25091
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1470138708654156e-06,
+      "loss": 0.4849,
+      "step": 25092
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.146437119704703e-06,
+      "loss": 0.4511,
+      "step": 25093
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.145860368543992e-06,
+      "loss": 0.3589,
+      "step": 25094
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.145283617383281e-06,
+      "loss": 0.4259,
+      "step": 25095
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.144706866222568e-06,
+      "loss": 0.311,
+      "step": 25096
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.144130115061857e-06,
+      "loss": 0.4778,
+      "step": 25097
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.143553363901146e-06,
+      "loss": 0.3712,
+      "step": 25098
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.142976612740434e-06,
+      "loss": 0.4809,
+      "step": 25099
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.142399861579722e-06,
+      "loss": 0.4298,
+      "step": 25100
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.141823110419011e-06,
+      "loss": 0.4018,
+      "step": 25101
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.141246359258298e-06,
+      "loss": 0.4231,
+      "step": 25102
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.140669608097587e-06,
+      "loss": 0.364,
+      "step": 25103
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.140092856936875e-06,
+      "loss": 0.4562,
+      "step": 25104
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.139516105776163e-06,
+      "loss": 0.3667,
+      "step": 25105
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.138939354615452e-06,
+      "loss": 0.4956,
+      "step": 25106
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.138362603454739e-06,
+      "loss": 0.4019,
+      "step": 25107
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.137785852294028e-06,
+      "loss": 0.4868,
+      "step": 25108
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.137209101133317e-06,
+      "loss": 0.3961,
+      "step": 25109
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.136632349972604e-06,
+      "loss": 0.3597,
+      "step": 25110
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.1360555988118934e-06,
+      "loss": 0.461,
+      "step": 25111
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.135478847651182e-06,
+      "loss": 0.4804,
+      "step": 25112
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.13490209649047e-06,
+      "loss": 0.3474,
+      "step": 25113
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 6.134325345329758e-06,
+      "loss": 0.5333,
+      "step": 25114
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.133748594169047e-06,
+      "loss": 0.4307,
+      "step": 25115
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.133171843008334e-06,
+      "loss": 0.3593,
+      "step": 25116
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.132595091847623e-06,
+      "loss": 0.326,
+      "step": 25117
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1320183406869115e-06,
+      "loss": 0.5606,
+      "step": 25118
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.131441589526199e-06,
+      "loss": 0.4399,
+      "step": 25119
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.130864838365488e-06,
+      "loss": 0.3112,
+      "step": 25120
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.130288087204776e-06,
+      "loss": 0.4395,
+      "step": 25121
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.129711336044064e-06,
+      "loss": 0.382,
+      "step": 25122
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1291345848833524e-06,
+      "loss": 0.4283,
+      "step": 25123
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.12855783372264e-06,
+      "loss": 0.4726,
+      "step": 25124
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.127981082561929e-06,
+      "loss": 0.4062,
+      "step": 25125
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.127404331401218e-06,
+      "loss": 0.4093,
+      "step": 25126
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.126827580240505e-06,
+      "loss": 0.4545,
+      "step": 25127
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.126250829079794e-06,
+      "loss": 0.4642,
+      "step": 25128
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.125674077919083e-06,
+      "loss": 0.4566,
+      "step": 25129
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1250973267583705e-06,
+      "loss": 0.3785,
+      "step": 25130
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.124520575597659e-06,
+      "loss": 0.3542,
+      "step": 25131
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1239438244369475e-06,
+      "loss": 0.4109,
+      "step": 25132
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.123367073276235e-06,
+      "loss": 0.3675,
+      "step": 25133
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.122790322115524e-06,
+      "loss": 0.3806,
+      "step": 25134
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.122213570954812e-06,
+      "loss": 0.4548,
+      "step": 25135
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1216368197941e-06,
+      "loss": 0.4898,
+      "step": 25136
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1210600686333885e-06,
+      "loss": 0.3966,
+      "step": 25137
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.120483317472677e-06,
+      "loss": 0.4852,
+      "step": 25138
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.119906566311965e-06,
+      "loss": 0.4627,
+      "step": 25139
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.119329815151254e-06,
+      "loss": 0.3243,
+      "step": 25140
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.118753063990541e-06,
+      "loss": 0.4381,
+      "step": 25141
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.11817631282983e-06,
+      "loss": 0.394,
+      "step": 25142
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.117599561669119e-06,
+      "loss": 0.4349,
+      "step": 25143
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1170228105084065e-06,
+      "loss": 0.4491,
+      "step": 25144
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.116446059347695e-06,
+      "loss": 0.4942,
+      "step": 25145
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.115869308186984e-06,
+      "loss": 0.4418,
+      "step": 25146
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.115292557026271e-06,
+      "loss": 0.3774,
+      "step": 25147
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.11471580586556e-06,
+      "loss": 0.4451,
+      "step": 25148
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.114139054704848e-06,
+      "loss": 0.4841,
+      "step": 25149
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.113562303544136e-06,
+      "loss": 0.5289,
+      "step": 25150
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1129855523834246e-06,
+      "loss": 0.5195,
+      "step": 25151
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.112408801222713e-06,
+      "loss": 0.4046,
+      "step": 25152
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.111832050062001e-06,
+      "loss": 0.3928,
+      "step": 25153
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.111255298901289e-06,
+      "loss": 0.4476,
+      "step": 25154
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.110678547740579e-06,
+      "loss": 0.3396,
+      "step": 25155
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1101017965798655e-06,
+      "loss": 0.3777,
+      "step": 25156
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.109525045419155e-06,
+      "loss": 0.5002,
+      "step": 25157
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.108948294258442e-06,
+      "loss": 0.3811,
+      "step": 25158
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.108371543097731e-06,
+      "loss": 0.3595,
+      "step": 25159
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.10779479193702e-06,
+      "loss": 0.4392,
+      "step": 25160
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.107218040776307e-06,
+      "loss": 0.4192,
+      "step": 25161
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.106641289615596e-06,
+      "loss": 0.4843,
+      "step": 25162
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.106064538454884e-06,
+      "loss": 0.3891,
+      "step": 25163
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.105487787294172e-06,
+      "loss": 0.3852,
+      "step": 25164
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.104911036133461e-06,
+      "loss": 0.4541,
+      "step": 25165
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.104334284972749e-06,
+      "loss": 0.5332,
+      "step": 25166
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.103757533812037e-06,
+      "loss": 0.4742,
+      "step": 25167
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.103180782651325e-06,
+      "loss": 0.5007,
+      "step": 25168
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.102604031490614e-06,
+      "loss": 0.5118,
+      "step": 25169
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.102027280329902e-06,
+      "loss": 0.3825,
+      "step": 25170
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.101450529169191e-06,
+      "loss": 0.4093,
+      "step": 25171
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.1008737780084795e-06,
+      "loss": 0.3934,
+      "step": 25172
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.100297026847767e-06,
+      "loss": 0.3671,
+      "step": 25173
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.099720275687056e-06,
+      "loss": 0.3907,
+      "step": 25174
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.099143524526343e-06,
+      "loss": 0.5245,
+      "step": 25175
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.098566773365632e-06,
+      "loss": 0.3784,
+      "step": 25176
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0979900222049205e-06,
+      "loss": 0.6503,
+      "step": 25177
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.097413271044208e-06,
+      "loss": 0.5415,
+      "step": 25178
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.096836519883497e-06,
+      "loss": 0.4232,
+      "step": 25179
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.096259768722785e-06,
+      "loss": 0.4549,
+      "step": 25180
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.095683017562073e-06,
+      "loss": 0.4551,
+      "step": 25181
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0951062664013614e-06,
+      "loss": 0.5292,
+      "step": 25182
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.09452951524065e-06,
+      "loss": 0.4423,
+      "step": 25183
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.093952764079938e-06,
+      "loss": 0.4101,
+      "step": 25184
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.093376012919226e-06,
+      "loss": 0.4189,
+      "step": 25185
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0927992617585156e-06,
+      "loss": 0.5127,
+      "step": 25186
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.092222510597802e-06,
+      "loss": 0.4769,
+      "step": 25187
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.091645759437092e-06,
+      "loss": 0.4209,
+      "step": 25188
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.09106900827638e-06,
+      "loss": 0.3244,
+      "step": 25189
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.090492257115668e-06,
+      "loss": 0.2769,
+      "step": 25190
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0899155059549565e-06,
+      "loss": 0.4875,
+      "step": 25191
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.089338754794244e-06,
+      "loss": 0.3382,
+      "step": 25192
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.088762003633533e-06,
+      "loss": 0.5087,
+      "step": 25193
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.088185252472821e-06,
+      "loss": 0.5024,
+      "step": 25194
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.087608501312109e-06,
+      "loss": 0.4293,
+      "step": 25195
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0870317501513975e-06,
+      "loss": 0.3871,
+      "step": 25196
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.086454998990686e-06,
+      "loss": 0.4527,
+      "step": 25197
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.085878247829974e-06,
+      "loss": 0.464,
+      "step": 25198
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.085301496669262e-06,
+      "loss": 0.4944,
+      "step": 25199
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.084724745508551e-06,
+      "loss": 0.3807,
+      "step": 25200
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0841479943478385e-06,
+      "loss": 0.3753,
+      "step": 25201
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.083571243187128e-06,
+      "loss": 0.4141,
+      "step": 25202
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.082994492026416e-06,
+      "loss": 0.4268,
+      "step": 25203
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.082417740865704e-06,
+      "loss": 0.383,
+      "step": 25204
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.081840989704993e-06,
+      "loss": 0.5174,
+      "step": 25205
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.081264238544281e-06,
+      "loss": 0.4879,
+      "step": 25206
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.080687487383569e-06,
+      "loss": 0.4233,
+      "step": 25207
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.080110736222857e-06,
+      "loss": 0.3738,
+      "step": 25208
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.079533985062146e-06,
+      "loss": 0.4542,
+      "step": 25209
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0789572339014336e-06,
+      "loss": 0.3899,
+      "step": 25210
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.078380482740722e-06,
+      "loss": 0.4214,
+      "step": 25211
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.07780373158001e-06,
+      "loss": 0.3671,
+      "step": 25212
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.077226980419298e-06,
+      "loss": 0.4199,
+      "step": 25213
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.076650229258587e-06,
+      "loss": 0.4944,
+      "step": 25214
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0760734780978745e-06,
+      "loss": 0.3792,
+      "step": 25215
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.075496726937163e-06,
+      "loss": 0.3438,
+      "step": 25216
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0749199757764524e-06,
+      "loss": 0.4625,
+      "step": 25217
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.074343224615739e-06,
+      "loss": 0.5668,
+      "step": 25218
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.073766473455029e-06,
+      "loss": 0.4467,
+      "step": 25219
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.073189722294317e-06,
+      "loss": 0.4822,
+      "step": 25220
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.072612971133605e-06,
+      "loss": 0.5061,
+      "step": 25221
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.072036219972893e-06,
+      "loss": 0.3724,
+      "step": 25222
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.071459468812182e-06,
+      "loss": 0.4512,
+      "step": 25223
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.07088271765147e-06,
+      "loss": 0.4093,
+      "step": 25224
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.070305966490758e-06,
+      "loss": 0.454,
+      "step": 25225
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.069729215330047e-06,
+      "loss": 0.3867,
+      "step": 25226
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.069152464169334e-06,
+      "loss": 0.3764,
+      "step": 25227
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.068575713008623e-06,
+      "loss": 0.4811,
+      "step": 25228
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.067998961847911e-06,
+      "loss": 0.4234,
+      "step": 25229
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.067422210687199e-06,
+      "loss": 0.4544,
+      "step": 25230
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.066845459526488e-06,
+      "loss": 0.5173,
+      "step": 25231
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.066268708365775e-06,
+      "loss": 0.3743,
+      "step": 25232
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.065691957205065e-06,
+      "loss": 0.507,
+      "step": 25233
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.065115206044353e-06,
+      "loss": 0.3262,
+      "step": 25234
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.064538454883641e-06,
+      "loss": 0.3532,
+      "step": 25235
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0639617037229295e-06,
+      "loss": 0.4637,
+      "step": 25236
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.063384952562218e-06,
+      "loss": 0.4366,
+      "step": 25237
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.062808201401506e-06,
+      "loss": 0.4537,
+      "step": 25238
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.062231450240794e-06,
+      "loss": 0.4427,
+      "step": 25239
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.061654699080083e-06,
+      "loss": 0.508,
+      "step": 25240
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0610779479193704e-06,
+      "loss": 0.4098,
+      "step": 25241
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.060501196758659e-06,
+      "loss": 0.4682,
+      "step": 25242
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0599244455979475e-06,
+      "loss": 0.3735,
+      "step": 25243
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.059347694437235e-06,
+      "loss": 0.5207,
+      "step": 25244
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.058770943276524e-06,
+      "loss": 0.3677,
+      "step": 25245
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.058194192115811e-06,
+      "loss": 0.3557,
+      "step": 25246
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0576174409551e-06,
+      "loss": 0.3424,
+      "step": 25247
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.057040689794389e-06,
+      "loss": 0.3627,
+      "step": 25248
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.056463938633676e-06,
+      "loss": 0.3723,
+      "step": 25249
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0558871874729655e-06,
+      "loss": 0.4416,
+      "step": 25250
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.055310436312254e-06,
+      "loss": 0.4123,
+      "step": 25251
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.054733685151542e-06,
+      "loss": 0.5,
+      "step": 25252
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.05415693399083e-06,
+      "loss": 0.3185,
+      "step": 25253
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.053580182830119e-06,
+      "loss": 0.4972,
+      "step": 25254
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0530034316694065e-06,
+      "loss": 0.4392,
+      "step": 25255
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.052426680508695e-06,
+      "loss": 0.4335,
+      "step": 25256
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0518499293479836e-06,
+      "loss": 0.4305,
+      "step": 25257
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.051273178187271e-06,
+      "loss": 0.4567,
+      "step": 25258
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.05069642702656e-06,
+      "loss": 0.4326,
+      "step": 25259
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.050119675865848e-06,
+      "loss": 0.422,
+      "step": 25260
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.049542924705136e-06,
+      "loss": 0.5506,
+      "step": 25261
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0489661735444245e-06,
+      "loss": 0.4707,
+      "step": 25262
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.048389422383712e-06,
+      "loss": 0.4872,
+      "step": 25263
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.047812671223002e-06,
+      "loss": 0.3667,
+      "step": 25264
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.04723592006229e-06,
+      "loss": 0.5255,
+      "step": 25265
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.046659168901578e-06,
+      "loss": 0.4554,
+      "step": 25266
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.046082417740866e-06,
+      "loss": 0.3243,
+      "step": 25267
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.045505666580155e-06,
+      "loss": 0.462,
+      "step": 25268
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0449289154194426e-06,
+      "loss": 0.4806,
+      "step": 25269
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.044352164258731e-06,
+      "loss": 0.3145,
+      "step": 25270
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.04377541309802e-06,
+      "loss": 0.3722,
+      "step": 25271
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.043198661937307e-06,
+      "loss": 0.4097,
+      "step": 25272
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.042621910776596e-06,
+      "loss": 0.3408,
+      "step": 25273
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.042045159615884e-06,
+      "loss": 0.2958,
+      "step": 25274
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.041468408455172e-06,
+      "loss": 0.4111,
+      "step": 25275
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.040891657294461e-06,
+      "loss": 0.4724,
+      "step": 25276
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.04031490613375e-06,
+      "loss": 0.3581,
+      "step": 25277
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.039738154973037e-06,
+      "loss": 0.5086,
+      "step": 25278
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.039161403812326e-06,
+      "loss": 0.2206,
+      "step": 25279
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.038584652651613e-06,
+      "loss": 0.3918,
+      "step": 25280
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.038007901490902e-06,
+      "loss": 0.45,
+      "step": 25281
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.037431150330191e-06,
+      "loss": 0.4874,
+      "step": 25282
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.036854399169479e-06,
+      "loss": 0.3654,
+      "step": 25283
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.036277648008767e-06,
+      "loss": 0.3665,
+      "step": 25284
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.035700896848056e-06,
+      "loss": 0.3285,
+      "step": 25285
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.035124145687343e-06,
+      "loss": 0.4141,
+      "step": 25286
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.034547394526632e-06,
+      "loss": 0.3707,
+      "step": 25287
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.0339706433659204e-06,
+      "loss": 0.3879,
+      "step": 25288
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.033393892205208e-06,
+      "loss": 0.4729,
+      "step": 25289
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.032817141044497e-06,
+      "loss": 0.4083,
+      "step": 25290
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.032240389883785e-06,
+      "loss": 0.3217,
+      "step": 25291
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.031663638723073e-06,
+      "loss": 0.4411,
+      "step": 25292
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 6.031086887562362e-06,
+      "loss": 0.4407,
+      "step": 25293
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.030510136401651e-06,
+      "loss": 0.5119,
+      "step": 25294
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0299333852409385e-06,
+      "loss": 0.4932,
+      "step": 25295
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.029356634080227e-06,
+      "loss": 0.3608,
+      "step": 25296
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.028779882919515e-06,
+      "loss": 0.4322,
+      "step": 25297
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.028203131758803e-06,
+      "loss": 0.2902,
+      "step": 25298
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.027626380598092e-06,
+      "loss": 0.4321,
+      "step": 25299
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0270496294373794e-06,
+      "loss": 0.4708,
+      "step": 25300
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.026472878276668e-06,
+      "loss": 0.4617,
+      "step": 25301
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0258961271159565e-06,
+      "loss": 0.491,
+      "step": 25302
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.025319375955244e-06,
+      "loss": 0.422,
+      "step": 25303
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.024742624794533e-06,
+      "loss": 0.3524,
+      "step": 25304
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.024165873633821e-06,
+      "loss": 0.4056,
+      "step": 25305
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.023589122473109e-06,
+      "loss": 0.5062,
+      "step": 25306
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0230123713123975e-06,
+      "loss": 0.5969,
+      "step": 25307
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.022435620151687e-06,
+      "loss": 0.294,
+      "step": 25308
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.021858868990974e-06,
+      "loss": 0.4493,
+      "step": 25309
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.021282117830263e-06,
+      "loss": 0.354,
+      "step": 25310
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.020705366669552e-06,
+      "loss": 0.394,
+      "step": 25311
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.020128615508839e-06,
+      "loss": 0.3721,
+      "step": 25312
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.019551864348128e-06,
+      "loss": 0.3958,
+      "step": 25313
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0189751131874155e-06,
+      "loss": 0.5159,
+      "step": 25314
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.018398362026704e-06,
+      "loss": 0.425,
+      "step": 25315
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0178216108659926e-06,
+      "loss": 0.4676,
+      "step": 25316
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.01724485970528e-06,
+      "loss": 0.3273,
+      "step": 25317
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.016668108544569e-06,
+      "loss": 0.4644,
+      "step": 25318
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.016091357383857e-06,
+      "loss": 0.2959,
+      "step": 25319
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.015514606223145e-06,
+      "loss": 0.4315,
+      "step": 25320
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0149378550624335e-06,
+      "loss": 0.5245,
+      "step": 25321
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.014361103901722e-06,
+      "loss": 0.3541,
+      "step": 25322
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.01378435274101e-06,
+      "loss": 0.3574,
+      "step": 25323
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.013207601580299e-06,
+      "loss": 0.5828,
+      "step": 25324
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.012630850419588e-06,
+      "loss": 0.3118,
+      "step": 25325
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.012054099258875e-06,
+      "loss": 0.3458,
+      "step": 25326
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.011477348098164e-06,
+      "loss": 0.4436,
+      "step": 25327
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.010900596937452e-06,
+      "loss": 0.3669,
+      "step": 25328
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.01032384577674e-06,
+      "loss": 0.5213,
+      "step": 25329
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.009747094616029e-06,
+      "loss": 0.5161,
+      "step": 25330
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.009170343455317e-06,
+      "loss": 0.4939,
+      "step": 25331
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.008593592294605e-06,
+      "loss": 0.3153,
+      "step": 25332
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.008016841133893e-06,
+      "loss": 0.5405,
+      "step": 25333
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.007440089973181e-06,
+      "loss": 0.4566,
+      "step": 25334
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.00686333881247e-06,
+      "loss": 0.4686,
+      "step": 25335
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.006286587651758e-06,
+      "loss": 0.3825,
+      "step": 25336
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.005709836491046e-06,
+      "loss": 0.4506,
+      "step": 25337
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.005133085330334e-06,
+      "loss": 0.4122,
+      "step": 25338
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.004556334169624e-06,
+      "loss": 0.5002,
+      "step": 25339
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0039795830089106e-06,
+      "loss": 0.4089,
+      "step": 25340
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0034028318482e-06,
+      "loss": 0.4805,
+      "step": 25341
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.0028260806874885e-06,
+      "loss": 0.4961,
+      "step": 25342
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.002249329526776e-06,
+      "loss": 0.4199,
+      "step": 25343
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.001672578366065e-06,
+      "loss": 0.3232,
+      "step": 25344
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.001095827205353e-06,
+      "loss": 0.4143,
+      "step": 25345
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 6.000519076044641e-06,
+      "loss": 0.3661,
+      "step": 25346
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9999423248839294e-06,
+      "loss": 0.4332,
+      "step": 25347
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.999365573723218e-06,
+      "loss": 0.4004,
+      "step": 25348
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.998788822562506e-06,
+      "loss": 0.4783,
+      "step": 25349
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.998212071401794e-06,
+      "loss": 0.3578,
+      "step": 25350
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.997635320241082e-06,
+      "loss": 0.4468,
+      "step": 25351
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.99705856908037e-06,
+      "loss": 0.4164,
+      "step": 25352
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.996481817919659e-06,
+      "loss": 0.4055,
+      "step": 25353
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.995905066758947e-06,
+      "loss": 0.3977,
+      "step": 25354
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.995328315598236e-06,
+      "loss": 0.6523,
+      "step": 25355
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9947515644375245e-06,
+      "loss": 0.5484,
+      "step": 25356
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.994174813276812e-06,
+      "loss": 0.3148,
+      "step": 25357
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.993598062116101e-06,
+      "loss": 0.5042,
+      "step": 25358
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.993021310955389e-06,
+      "loss": 0.4174,
+      "step": 25359
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.992444559794677e-06,
+      "loss": 0.4117,
+      "step": 25360
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9918678086339655e-06,
+      "loss": 0.3619,
+      "step": 25361
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.991291057473254e-06,
+      "loss": 0.5344,
+      "step": 25362
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.990714306312542e-06,
+      "loss": 0.4126,
+      "step": 25363
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.99013755515183e-06,
+      "loss": 0.3797,
+      "step": 25364
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.989560803991119e-06,
+      "loss": 0.4568,
+      "step": 25365
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9889840528304065e-06,
+      "loss": 0.3978,
+      "step": 25366
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.988407301669695e-06,
+      "loss": 0.3965,
+      "step": 25367
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.987830550508983e-06,
+      "loss": 0.3502,
+      "step": 25368
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.987253799348271e-06,
+      "loss": 0.3914,
+      "step": 25369
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.986677048187561e-06,
+      "loss": 0.4474,
+      "step": 25370
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9861002970268474e-06,
+      "loss": 0.4793,
+      "step": 25371
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.985523545866137e-06,
+      "loss": 0.3584,
+      "step": 25372
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.984946794705425e-06,
+      "loss": 0.4652,
+      "step": 25373
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.984370043544713e-06,
+      "loss": 0.6165,
+      "step": 25374
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9837932923840016e-06,
+      "loss": 0.4164,
+      "step": 25375
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.98321654122329e-06,
+      "loss": 0.3917,
+      "step": 25376
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.982639790062578e-06,
+      "loss": 0.4405,
+      "step": 25377
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.982063038901866e-06,
+      "loss": 0.4995,
+      "step": 25378
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.981486287741155e-06,
+      "loss": 0.3886,
+      "step": 25379
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9809095365804425e-06,
+      "loss": 0.4236,
+      "step": 25380
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.980332785419731e-06,
+      "loss": 0.518,
+      "step": 25381
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.97975603425902e-06,
+      "loss": 0.4671,
+      "step": 25382
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.979179283098307e-06,
+      "loss": 0.4936,
+      "step": 25383
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.978602531937596e-06,
+      "loss": 0.4155,
+      "step": 25384
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9780257807768835e-06,
+      "loss": 0.3723,
+      "step": 25385
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.977449029616173e-06,
+      "loss": 0.4601,
+      "step": 25386
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.976872278455461e-06,
+      "loss": 0.4342,
+      "step": 25387
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.976295527294749e-06,
+      "loss": 0.4667,
+      "step": 25388
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.975718776134038e-06,
+      "loss": 0.3315,
+      "step": 25389
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.975142024973326e-06,
+      "loss": 0.3485,
+      "step": 25390
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.974565273812614e-06,
+      "loss": 0.341,
+      "step": 25391
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.973988522651902e-06,
+      "loss": 0.4685,
+      "step": 25392
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.973411771491191e-06,
+      "loss": 0.4102,
+      "step": 25393
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.972835020330479e-06,
+      "loss": 0.4014,
+      "step": 25394
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.972258269169767e-06,
+      "loss": 0.3672,
+      "step": 25395
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.971681518009056e-06,
+      "loss": 0.4566,
+      "step": 25396
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.971104766848343e-06,
+      "loss": 0.3954,
+      "step": 25397
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.970528015687632e-06,
+      "loss": 0.382,
+      "step": 25398
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.969951264526921e-06,
+      "loss": 0.4303,
+      "step": 25399
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.969374513366208e-06,
+      "loss": 0.3883,
+      "step": 25400
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9687977622054975e-06,
+      "loss": 0.4011,
+      "step": 25401
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.968221011044784e-06,
+      "loss": 0.4763,
+      "step": 25402
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.967644259884074e-06,
+      "loss": 0.4457,
+      "step": 25403
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.967067508723362e-06,
+      "loss": 0.4037,
+      "step": 25404
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.96649075756265e-06,
+      "loss": 0.4953,
+      "step": 25405
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9659140064019384e-06,
+      "loss": 0.3717,
+      "step": 25406
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.965337255241227e-06,
+      "loss": 0.3648,
+      "step": 25407
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.964760504080515e-06,
+      "loss": 0.4758,
+      "step": 25408
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.964183752919803e-06,
+      "loss": 0.511,
+      "step": 25409
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.963607001759092e-06,
+      "loss": 0.4699,
+      "step": 25410
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.963030250598379e-06,
+      "loss": 0.4795,
+      "step": 25411
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.962453499437668e-06,
+      "loss": 0.3797,
+      "step": 25412
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9618767482769565e-06,
+      "loss": 0.3895,
+      "step": 25413
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.961299997116244e-06,
+      "loss": 0.4484,
+      "step": 25414
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.960723245955533e-06,
+      "loss": 0.4701,
+      "step": 25415
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.960146494794822e-06,
+      "loss": 0.4799,
+      "step": 25416
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.95956974363411e-06,
+      "loss": 0.5185,
+      "step": 25417
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.958992992473398e-06,
+      "loss": 0.3767,
+      "step": 25418
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.958416241312686e-06,
+      "loss": 0.433,
+      "step": 25419
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9578394901519745e-06,
+      "loss": 0.4964,
+      "step": 25420
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.957262738991263e-06,
+      "loss": 0.4261,
+      "step": 25421
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.956685987830551e-06,
+      "loss": 0.5262,
+      "step": 25422
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.956109236669839e-06,
+      "loss": 0.3639,
+      "step": 25423
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.955532485509128e-06,
+      "loss": 0.419,
+      "step": 25424
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9549557343484155e-06,
+      "loss": 0.5074,
+      "step": 25425
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.954378983187704e-06,
+      "loss": 0.3024,
+      "step": 25426
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9538022320269925e-06,
+      "loss": 0.3315,
+      "step": 25427
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.95322548086628e-06,
+      "loss": 0.3484,
+      "step": 25428
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.952648729705569e-06,
+      "loss": 0.4316,
+      "step": 25429
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.952071978544858e-06,
+      "loss": 0.5338,
+      "step": 25430
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.951495227384145e-06,
+      "loss": 0.3978,
+      "step": 25431
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.950918476223434e-06,
+      "loss": 0.2878,
+      "step": 25432
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.950341725062723e-06,
+      "loss": 0.3501,
+      "step": 25433
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9497649739020105e-06,
+      "loss": 0.423,
+      "step": 25434
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.949188222741299e-06,
+      "loss": 0.28,
+      "step": 25435
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.948611471580587e-06,
+      "loss": 0.3579,
+      "step": 25436
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.948034720419875e-06,
+      "loss": 0.3886,
+      "step": 25437
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.947457969259164e-06,
+      "loss": 0.4139,
+      "step": 25438
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9468812180984515e-06,
+      "loss": 0.3226,
+      "step": 25439
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.94630446693774e-06,
+      "loss": 0.4018,
+      "step": 25440
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.945727715777029e-06,
+      "loss": 0.3067,
+      "step": 25441
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.945150964616316e-06,
+      "loss": 0.4186,
+      "step": 25442
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.944574213455605e-06,
+      "loss": 0.4435,
+      "step": 25443
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.943997462294893e-06,
+      "loss": 0.3154,
+      "step": 25444
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.943420711134181e-06,
+      "loss": 0.31,
+      "step": 25445
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9428439599734695e-06,
+      "loss": 0.4609,
+      "step": 25446
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.942267208812759e-06,
+      "loss": 0.4031,
+      "step": 25447
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.941690457652047e-06,
+      "loss": 0.4374,
+      "step": 25448
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.941113706491335e-06,
+      "loss": 0.3964,
+      "step": 25449
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.940536955330624e-06,
+      "loss": 0.4491,
+      "step": 25450
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.939960204169911e-06,
+      "loss": 0.4364,
+      "step": 25451
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9393834530092e-06,
+      "loss": 0.4767,
+      "step": 25452
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9388067018484884e-06,
+      "loss": 0.4237,
+      "step": 25453
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.938229950687776e-06,
+      "loss": 0.5357,
+      "step": 25454
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.937653199527065e-06,
+      "loss": 0.477,
+      "step": 25455
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.937076448366352e-06,
+      "loss": 0.4096,
+      "step": 25456
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.936499697205641e-06,
+      "loss": 0.3303,
+      "step": 25457
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.935922946044929e-06,
+      "loss": 0.4559,
+      "step": 25458
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.935346194884217e-06,
+      "loss": 0.4943,
+      "step": 25459
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.934769443723506e-06,
+      "loss": 0.368,
+      "step": 25460
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.934192692562795e-06,
+      "loss": 0.4844,
+      "step": 25461
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.933615941402082e-06,
+      "loss": 0.354,
+      "step": 25462
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.933039190241371e-06,
+      "loss": 0.4454,
+      "step": 25463
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.93246243908066e-06,
+      "loss": 0.3835,
+      "step": 25464
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.931885687919947e-06,
+      "loss": 0.3715,
+      "step": 25465
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.931308936759236e-06,
+      "loss": 0.3389,
+      "step": 25466
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.9307321855985245e-06,
+      "loss": 0.4409,
+      "step": 25467
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.930155434437812e-06,
+      "loss": 0.4568,
+      "step": 25468
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.929578683277101e-06,
+      "loss": 0.3605,
+      "step": 25469
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.929001932116389e-06,
+      "loss": 0.4211,
+      "step": 25470
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 5.928425180955677e-06,
+      "loss": 0.4437,
+      "step": 25471
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9278484297949655e-06,
+      "loss": 0.3586,
+      "step": 25472
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.927271678634253e-06,
+      "loss": 0.3885,
+      "step": 25473
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.926694927473542e-06,
+      "loss": 0.4186,
+      "step": 25474
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.92611817631283e-06,
+      "loss": 0.4015,
+      "step": 25475
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.925541425152118e-06,
+      "loss": 0.5737,
+      "step": 25476
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.924964673991407e-06,
+      "loss": 0.4787,
+      "step": 25477
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.924387922830696e-06,
+      "loss": 0.4557,
+      "step": 25478
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9238111716699835e-06,
+      "loss": 0.4787,
+      "step": 25479
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.923234420509272e-06,
+      "loss": 0.4034,
+      "step": 25480
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9226576693485605e-06,
+      "loss": 0.3497,
+      "step": 25481
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.922080918187848e-06,
+      "loss": 0.4155,
+      "step": 25482
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.921504167027137e-06,
+      "loss": 0.5233,
+      "step": 25483
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.920927415866425e-06,
+      "loss": 0.4491,
+      "step": 25484
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.920350664705713e-06,
+      "loss": 0.4954,
+      "step": 25485
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9197739135450015e-06,
+      "loss": 0.402,
+      "step": 25486
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.91919716238429e-06,
+      "loss": 0.4609,
+      "step": 25487
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.918620411223578e-06,
+      "loss": 0.4992,
+      "step": 25488
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.918043660062866e-06,
+      "loss": 0.3595,
+      "step": 25489
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.917466908902154e-06,
+      "loss": 0.4121,
+      "step": 25490
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9168901577414425e-06,
+      "loss": 0.4711,
+      "step": 25491
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.916313406580732e-06,
+      "loss": 0.4708,
+      "step": 25492
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.915736655420019e-06,
+      "loss": 0.3936,
+      "step": 25493
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.915159904259308e-06,
+      "loss": 0.3487,
+      "step": 25494
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.914583153098597e-06,
+      "loss": 0.3917,
+      "step": 25495
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.914006401937884e-06,
+      "loss": 0.4276,
+      "step": 25496
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.913429650777173e-06,
+      "loss": 0.3456,
+      "step": 25497
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.912852899616461e-06,
+      "loss": 0.5421,
+      "step": 25498
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.912276148455749e-06,
+      "loss": 0.344,
+      "step": 25499
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9116993972950376e-06,
+      "loss": 0.5188,
+      "step": 25500
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.911122646134326e-06,
+      "loss": 0.4755,
+      "step": 25501
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.910545894973614e-06,
+      "loss": 0.3614,
+      "step": 25502
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.909969143812902e-06,
+      "loss": 0.3785,
+      "step": 25503
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.909392392652191e-06,
+      "loss": 0.4749,
+      "step": 25504
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.9088156414914785e-06,
+      "loss": 0.3749,
+      "step": 25505
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.908238890330767e-06,
+      "loss": 0.3372,
+      "step": 25506
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.907662139170055e-06,
+      "loss": 0.4685,
+      "step": 25507
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.907085388009344e-06,
+      "loss": 0.3532,
+      "step": 25508
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.906508636848633e-06,
+      "loss": 0.4591,
+      "step": 25509
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.90593188568792e-06,
+      "loss": 0.5413,
+      "step": 25510
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.905355134527209e-06,
+      "loss": 0.4462,
+      "step": 25511
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.904778383366497e-06,
+      "loss": 0.4281,
+      "step": 25512
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.904201632205785e-06,
+      "loss": 0.464,
+      "step": 25513
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.903624881045074e-06,
+      "loss": 0.3603,
+      "step": 25514
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.903048129884362e-06,
+      "loss": 0.3485,
+      "step": 25515
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.90247137872365e-06,
+      "loss": 0.4376,
+      "step": 25516
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.901894627562938e-06,
+      "loss": 0.3131,
+      "step": 25517
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.901317876402227e-06,
+      "loss": 0.303,
+      "step": 25518
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.900741125241515e-06,
+      "loss": 0.4361,
+      "step": 25519
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.900164374080803e-06,
+      "loss": 0.4962,
+      "step": 25520
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8995876229200925e-06,
+      "loss": 0.4961,
+      "step": 25521
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.899010871759379e-06,
+      "loss": 0.4142,
+      "step": 25522
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.898434120598669e-06,
+      "loss": 0.3997,
+      "step": 25523
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8978573694379556e-06,
+      "loss": 0.4174,
+      "step": 25524
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.897280618277245e-06,
+      "loss": 0.483,
+      "step": 25525
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8967038671165335e-06,
+      "loss": 0.4747,
+      "step": 25526
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.896127115955821e-06,
+      "loss": 0.4194,
+      "step": 25527
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.89555036479511e-06,
+      "loss": 0.4245,
+      "step": 25528
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.894973613634398e-06,
+      "loss": 0.4369,
+      "step": 25529
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.894396862473686e-06,
+      "loss": 0.4471,
+      "step": 25530
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8938201113129744e-06,
+      "loss": 0.4285,
+      "step": 25531
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.893243360152263e-06,
+      "loss": 0.2736,
+      "step": 25532
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.892666608991551e-06,
+      "loss": 0.4112,
+      "step": 25533
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.892089857830839e-06,
+      "loss": 0.4806,
+      "step": 25534
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.891513106670128e-06,
+      "loss": 0.3358,
+      "step": 25535
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.890936355509415e-06,
+      "loss": 0.3975,
+      "step": 25536
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.890359604348704e-06,
+      "loss": 0.4193,
+      "step": 25537
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.889782853187993e-06,
+      "loss": 0.3278,
+      "step": 25538
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.889206102027281e-06,
+      "loss": 0.2915,
+      "step": 25539
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8886293508665695e-06,
+      "loss": 0.3398,
+      "step": 25540
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.888052599705857e-06,
+      "loss": 0.4448,
+      "step": 25541
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.887475848545146e-06,
+      "loss": 0.3215,
+      "step": 25542
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.886899097384434e-06,
+      "loss": 0.354,
+      "step": 25543
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.886322346223722e-06,
+      "loss": 0.5627,
+      "step": 25544
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8857455950630105e-06,
+      "loss": 0.3724,
+      "step": 25545
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.885168843902299e-06,
+      "loss": 0.4236,
+      "step": 25546
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.884592092741587e-06,
+      "loss": 0.4273,
+      "step": 25547
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.884015341580875e-06,
+      "loss": 0.4231,
+      "step": 25548
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.883438590420164e-06,
+      "loss": 0.3803,
+      "step": 25549
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8828618392594515e-06,
+      "loss": 0.4075,
+      "step": 25550
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.88228508809874e-06,
+      "loss": 0.3151,
+      "step": 25551
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.881708336938029e-06,
+      "loss": 0.4423,
+      "step": 25552
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.881131585777316e-06,
+      "loss": 0.4348,
+      "step": 25553
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.880554834616606e-06,
+      "loss": 0.4389,
+      "step": 25554
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.879978083455894e-06,
+      "loss": 0.4335,
+      "step": 25555
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.879401332295182e-06,
+      "loss": 0.3853,
+      "step": 25556
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.87882458113447e-06,
+      "loss": 0.3136,
+      "step": 25557
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.878247829973758e-06,
+      "loss": 0.4104,
+      "step": 25558
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8776710788130466e-06,
+      "loss": 0.4041,
+      "step": 25559
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.877094327652335e-06,
+      "loss": 0.3971,
+      "step": 25560
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.876517576491623e-06,
+      "loss": 0.5442,
+      "step": 25561
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.875940825330911e-06,
+      "loss": 0.348,
+      "step": 25562
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8753640741702e-06,
+      "loss": 0.4561,
+      "step": 25563
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8747873230094875e-06,
+      "loss": 0.4687,
+      "step": 25564
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.874210571848776e-06,
+      "loss": 0.4519,
+      "step": 25565
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.873633820688065e-06,
+      "loss": 0.3985,
+      "step": 25566
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.873057069527352e-06,
+      "loss": 0.5658,
+      "step": 25567
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.872480318366641e-06,
+      "loss": 0.4269,
+      "step": 25568
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.87190356720593e-06,
+      "loss": 0.3201,
+      "step": 25569
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.871326816045218e-06,
+      "loss": 0.4277,
+      "step": 25570
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.870750064884506e-06,
+      "loss": 0.4194,
+      "step": 25571
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.870173313723795e-06,
+      "loss": 0.3832,
+      "step": 25572
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.869596562563083e-06,
+      "loss": 0.4491,
+      "step": 25573
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.869019811402371e-06,
+      "loss": 0.3657,
+      "step": 25574
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.868443060241659e-06,
+      "loss": 0.4291,
+      "step": 25575
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.867866309080947e-06,
+      "loss": 0.4001,
+      "step": 25576
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.867289557920236e-06,
+      "loss": 0.3455,
+      "step": 25577
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.866712806759524e-06,
+      "loss": 0.4524,
+      "step": 25578
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.866136055598812e-06,
+      "loss": 0.3649,
+      "step": 25579
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.865559304438101e-06,
+      "loss": 0.4089,
+      "step": 25580
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.864982553277388e-06,
+      "loss": 0.3749,
+      "step": 25581
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.864405802116677e-06,
+      "loss": 0.5669,
+      "step": 25582
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.863829050955966e-06,
+      "loss": 0.2844,
+      "step": 25583
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.863252299795253e-06,
+      "loss": 0.4255,
+      "step": 25584
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8626755486345425e-06,
+      "loss": 0.357,
+      "step": 25585
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.862098797473831e-06,
+      "loss": 0.3278,
+      "step": 25586
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.861522046313119e-06,
+      "loss": 0.4307,
+      "step": 25587
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.860945295152407e-06,
+      "loss": 0.3402,
+      "step": 25588
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.860368543991696e-06,
+      "loss": 0.3467,
+      "step": 25589
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8597917928309834e-06,
+      "loss": 0.394,
+      "step": 25590
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.859215041670272e-06,
+      "loss": 0.343,
+      "step": 25591
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8586382905095605e-06,
+      "loss": 0.412,
+      "step": 25592
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.858061539348848e-06,
+      "loss": 0.4173,
+      "step": 25593
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.857484788188137e-06,
+      "loss": 0.4058,
+      "step": 25594
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.856908037027424e-06,
+      "loss": 0.3266,
+      "step": 25595
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.856331285866713e-06,
+      "loss": 0.3101,
+      "step": 25596
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8557545347060015e-06,
+      "loss": 0.4672,
+      "step": 25597
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.855177783545289e-06,
+      "loss": 0.4242,
+      "step": 25598
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.854601032384578e-06,
+      "loss": 0.4924,
+      "step": 25599
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.854024281223867e-06,
+      "loss": 0.42,
+      "step": 25600
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.853447530063155e-06,
+      "loss": 0.3467,
+      "step": 25601
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.852870778902443e-06,
+      "loss": 0.5349,
+      "step": 25602
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.852294027741732e-06,
+      "loss": 0.4046,
+      "step": 25603
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8517172765810195e-06,
+      "loss": 0.4286,
+      "step": 25604
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.851140525420308e-06,
+      "loss": 0.4389,
+      "step": 25605
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8505637742595966e-06,
+      "loss": 0.4827,
+      "step": 25606
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.849987023098884e-06,
+      "loss": 0.4946,
+      "step": 25607
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.849410271938173e-06,
+      "loss": 0.3266,
+      "step": 25608
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.848833520777461e-06,
+      "loss": 0.4312,
+      "step": 25609
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.848256769616749e-06,
+      "loss": 0.4461,
+      "step": 25610
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8476800184560375e-06,
+      "loss": 0.437,
+      "step": 25611
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.847103267295325e-06,
+      "loss": 0.3994,
+      "step": 25612
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.846526516134614e-06,
+      "loss": 0.3818,
+      "step": 25613
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.845949764973903e-06,
+      "loss": 0.4852,
+      "step": 25614
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.84537301381319e-06,
+      "loss": 0.3571,
+      "step": 25615
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.844796262652479e-06,
+      "loss": 0.439,
+      "step": 25616
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.844219511491768e-06,
+      "loss": 0.5358,
+      "step": 25617
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8436427603310556e-06,
+      "loss": 0.4744,
+      "step": 25618
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.843066009170344e-06,
+      "loss": 0.4116,
+      "step": 25619
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.842489258009633e-06,
+      "loss": 0.4858,
+      "step": 25620
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.84191250684892e-06,
+      "loss": 0.3472,
+      "step": 25621
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.841335755688209e-06,
+      "loss": 0.4874,
+      "step": 25622
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.840759004527497e-06,
+      "loss": 0.3579,
+      "step": 25623
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.840182253366785e-06,
+      "loss": 0.4297,
+      "step": 25624
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.839605502206074e-06,
+      "loss": 0.4643,
+      "step": 25625
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.839028751045362e-06,
+      "loss": 0.4735,
+      "step": 25626
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.83845199988465e-06,
+      "loss": 0.5127,
+      "step": 25627
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.837875248723938e-06,
+      "loss": 0.3644,
+      "step": 25628
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.837298497563226e-06,
+      "loss": 0.4396,
+      "step": 25629
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8367217464025146e-06,
+      "loss": 0.4251,
+      "step": 25630
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.836144995241804e-06,
+      "loss": 0.4606,
+      "step": 25631
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.835568244081092e-06,
+      "loss": 0.3427,
+      "step": 25632
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.83499149292038e-06,
+      "loss": 0.3404,
+      "step": 25633
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.834414741759669e-06,
+      "loss": 0.509,
+      "step": 25634
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.833837990598956e-06,
+      "loss": 0.3687,
+      "step": 25635
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.833261239438245e-06,
+      "loss": 0.3554,
+      "step": 25636
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8326844882775334e-06,
+      "loss": 0.3935,
+      "step": 25637
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.832107737116821e-06,
+      "loss": 0.4728,
+      "step": 25638
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.83153098595611e-06,
+      "loss": 0.3583,
+      "step": 25639
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.830954234795398e-06,
+      "loss": 0.2934,
+      "step": 25640
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.830377483634686e-06,
+      "loss": 0.5529,
+      "step": 25641
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.829800732473974e-06,
+      "loss": 0.4966,
+      "step": 25642
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.829223981313264e-06,
+      "loss": 0.4176,
+      "step": 25643
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.828647230152551e-06,
+      "loss": 0.4519,
+      "step": 25644
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.82807047899184e-06,
+      "loss": 0.4064,
+      "step": 25645
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.827493727831127e-06,
+      "loss": 0.3996,
+      "step": 25646
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.826916976670416e-06,
+      "loss": 0.3396,
+      "step": 25647
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.826340225509705e-06,
+      "loss": 0.4255,
+      "step": 25648
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.8257634743489924e-06,
+      "loss": 0.3712,
+      "step": 25649
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 5.825186723188281e-06,
+      "loss": 0.4068,
+      "step": 25650
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8246099720275695e-06,
+      "loss": 0.4349,
+      "step": 25651
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.824033220866857e-06,
+      "loss": 0.4319,
+      "step": 25652
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.823456469706146e-06,
+      "loss": 0.4444,
+      "step": 25653
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.822879718545434e-06,
+      "loss": 0.3629,
+      "step": 25654
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.822302967384722e-06,
+      "loss": 0.4603,
+      "step": 25655
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8217262162240105e-06,
+      "loss": 0.4633,
+      "step": 25656
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.821149465063299e-06,
+      "loss": 0.4845,
+      "step": 25657
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.820572713902587e-06,
+      "loss": 0.3174,
+      "step": 25658
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.819995962741875e-06,
+      "loss": 0.3728,
+      "step": 25659
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.819419211581165e-06,
+      "loss": 0.5001,
+      "step": 25660
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.818842460420452e-06,
+      "loss": 0.5407,
+      "step": 25661
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.818265709259741e-06,
+      "loss": 0.4252,
+      "step": 25662
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8176889580990285e-06,
+      "loss": 0.3129,
+      "step": 25663
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.817112206938317e-06,
+      "loss": 0.4847,
+      "step": 25664
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8165354557776056e-06,
+      "loss": 0.3674,
+      "step": 25665
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.815958704616893e-06,
+      "loss": 0.2943,
+      "step": 25666
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.815381953456182e-06,
+      "loss": 0.5188,
+      "step": 25667
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.81480520229547e-06,
+      "loss": 0.4515,
+      "step": 25668
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.814228451134758e-06,
+      "loss": 0.4821,
+      "step": 25669
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8136516999740465e-06,
+      "loss": 0.4516,
+      "step": 25670
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.813074948813335e-06,
+      "loss": 0.6114,
+      "step": 25671
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.812498197652623e-06,
+      "loss": 0.5034,
+      "step": 25672
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.811921446491911e-06,
+      "loss": 0.4629,
+      "step": 25673
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.811344695331201e-06,
+      "loss": 0.4868,
+      "step": 25674
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8107679441704875e-06,
+      "loss": 0.3288,
+      "step": 25675
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.810191193009777e-06,
+      "loss": 0.5471,
+      "step": 25676
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.809614441849065e-06,
+      "loss": 0.474,
+      "step": 25677
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.809037690688353e-06,
+      "loss": 0.3941,
+      "step": 25678
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.808460939527642e-06,
+      "loss": 0.4395,
+      "step": 25679
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.807884188366929e-06,
+      "loss": 0.5161,
+      "step": 25680
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.807307437206218e-06,
+      "loss": 0.4651,
+      "step": 25681
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.806730686045506e-06,
+      "loss": 0.3759,
+      "step": 25682
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.806153934884794e-06,
+      "loss": 0.3867,
+      "step": 25683
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.805577183724083e-06,
+      "loss": 0.5457,
+      "step": 25684
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.805000432563371e-06,
+      "loss": 0.381,
+      "step": 25685
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.804423681402659e-06,
+      "loss": 0.4804,
+      "step": 25686
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.803846930241947e-06,
+      "loss": 0.4671,
+      "step": 25687
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.803270179081236e-06,
+      "loss": 0.4307,
+      "step": 25688
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8026934279205236e-06,
+      "loss": 0.4064,
+      "step": 25689
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.802116676759812e-06,
+      "loss": 0.4472,
+      "step": 25690
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.8015399255991015e-06,
+      "loss": 0.4892,
+      "step": 25691
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.800963174438389e-06,
+      "loss": 0.4198,
+      "step": 25692
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.800386423277678e-06,
+      "loss": 0.4388,
+      "step": 25693
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.799809672116966e-06,
+      "loss": 0.4707,
+      "step": 25694
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.799232920956254e-06,
+      "loss": 0.397,
+      "step": 25695
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7986561697955424e-06,
+      "loss": 0.4276,
+      "step": 25696
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.79807941863483e-06,
+      "loss": 0.4792,
+      "step": 25697
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.797502667474119e-06,
+      "loss": 0.3989,
+      "step": 25698
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.796925916313407e-06,
+      "loss": 0.324,
+      "step": 25699
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.796349165152695e-06,
+      "loss": 0.4869,
+      "step": 25700
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.795772413991983e-06,
+      "loss": 0.4314,
+      "step": 25701
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.795195662831272e-06,
+      "loss": 0.4613,
+      "step": 25702
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.79461891167056e-06,
+      "loss": 0.4343,
+      "step": 25703
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.794042160509848e-06,
+      "loss": 0.4354,
+      "step": 25704
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7934654093491375e-06,
+      "loss": 0.4036,
+      "step": 25705
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.792888658188424e-06,
+      "loss": 0.3713,
+      "step": 25706
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.792311907027714e-06,
+      "loss": 0.3251,
+      "step": 25707
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.791735155867002e-06,
+      "loss": 0.4652,
+      "step": 25708
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.79115840470629e-06,
+      "loss": 0.2634,
+      "step": 25709
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7905816535455785e-06,
+      "loss": 0.4332,
+      "step": 25710
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.790004902384867e-06,
+      "loss": 0.4513,
+      "step": 25711
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.789428151224155e-06,
+      "loss": 0.3696,
+      "step": 25712
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.788851400063443e-06,
+      "loss": 0.3735,
+      "step": 25713
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.788274648902732e-06,
+      "loss": 0.3344,
+      "step": 25714
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7876978977420195e-06,
+      "loss": 0.404,
+      "step": 25715
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.787121146581308e-06,
+      "loss": 0.4034,
+      "step": 25716
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.786544395420596e-06,
+      "loss": 0.4978,
+      "step": 25717
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.785967644259884e-06,
+      "loss": 0.3707,
+      "step": 25718
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.785390893099173e-06,
+      "loss": 0.3127,
+      "step": 25719
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7848141419384604e-06,
+      "loss": 0.3959,
+      "step": 25720
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.784237390777749e-06,
+      "loss": 0.418,
+      "step": 25721
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.783660639617038e-06,
+      "loss": 0.4326,
+      "step": 25722
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.783083888456326e-06,
+      "loss": 0.3886,
+      "step": 25723
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7825071372956146e-06,
+      "loss": 0.4145,
+      "step": 25724
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.781930386134903e-06,
+      "loss": 0.4046,
+      "step": 25725
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.781353634974191e-06,
+      "loss": 0.4745,
+      "step": 25726
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.780776883813479e-06,
+      "loss": 0.4559,
+      "step": 25727
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.780200132652768e-06,
+      "loss": 0.3717,
+      "step": 25728
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7796233814920555e-06,
+      "loss": 0.4163,
+      "step": 25729
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.779046630331344e-06,
+      "loss": 0.3428,
+      "step": 25730
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.778469879170633e-06,
+      "loss": 0.4273,
+      "step": 25731
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.77789312800992e-06,
+      "loss": 0.421,
+      "step": 25732
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.777316376849209e-06,
+      "loss": 0.4427,
+      "step": 25733
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7767396256884965e-06,
+      "loss": 0.4465,
+      "step": 25734
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.776162874527785e-06,
+      "loss": 0.3416,
+      "step": 25735
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.775586123367074e-06,
+      "loss": 0.3965,
+      "step": 25736
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.775009372206361e-06,
+      "loss": 0.3478,
+      "step": 25737
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.774432621045651e-06,
+      "loss": 0.5091,
+      "step": 25738
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.773855869884939e-06,
+      "loss": 0.3713,
+      "step": 25739
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.773279118724227e-06,
+      "loss": 0.4142,
+      "step": 25740
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.772702367563515e-06,
+      "loss": 0.3406,
+      "step": 25741
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.772125616402804e-06,
+      "loss": 0.4801,
+      "step": 25742
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.771548865242092e-06,
+      "loss": 0.5011,
+      "step": 25743
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.77097211408138e-06,
+      "loss": 0.4015,
+      "step": 25744
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.770395362920669e-06,
+      "loss": 0.3544,
+      "step": 25745
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.769818611759956e-06,
+      "loss": 0.4511,
+      "step": 25746
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.769241860599245e-06,
+      "loss": 0.3707,
+      "step": 25747
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.768665109438533e-06,
+      "loss": 0.5381,
+      "step": 25748
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.768088358277821e-06,
+      "loss": 0.5503,
+      "step": 25749
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.76751160711711e-06,
+      "loss": 0.4352,
+      "step": 25750
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.766934855956397e-06,
+      "loss": 0.3535,
+      "step": 25751
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.766358104795686e-06,
+      "loss": 0.452,
+      "step": 25752
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.765781353634975e-06,
+      "loss": 0.4042,
+      "step": 25753
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.765204602474263e-06,
+      "loss": 0.5181,
+      "step": 25754
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7646278513135514e-06,
+      "loss": 0.3858,
+      "step": 25755
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.76405110015284e-06,
+      "loss": 0.5025,
+      "step": 25756
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.763474348992128e-06,
+      "loss": 0.49,
+      "step": 25757
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.762897597831416e-06,
+      "loss": 0.4016,
+      "step": 25758
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.762320846670705e-06,
+      "loss": 0.4604,
+      "step": 25759
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.761744095509992e-06,
+      "loss": 0.4391,
+      "step": 25760
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.761167344349281e-06,
+      "loss": 0.3426,
+      "step": 25761
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7605905931885695e-06,
+      "loss": 0.3855,
+      "step": 25762
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.760013842027857e-06,
+      "loss": 0.3788,
+      "step": 25763
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.759437090867146e-06,
+      "loss": 0.5235,
+      "step": 25764
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.758860339706435e-06,
+      "loss": 0.5017,
+      "step": 25765
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.758283588545722e-06,
+      "loss": 0.5073,
+      "step": 25766
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.757706837385011e-06,
+      "loss": 0.4041,
+      "step": 25767
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.757130086224298e-06,
+      "loss": 0.3603,
+      "step": 25768
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7565533350635875e-06,
+      "loss": 0.5632,
+      "step": 25769
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.755976583902876e-06,
+      "loss": 0.4891,
+      "step": 25770
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.755399832742164e-06,
+      "loss": 0.4347,
+      "step": 25771
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.754823081581452e-06,
+      "loss": 0.3198,
+      "step": 25772
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.754246330420741e-06,
+      "loss": 0.5815,
+      "step": 25773
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7536695792600285e-06,
+      "loss": 0.4267,
+      "step": 25774
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.753092828099317e-06,
+      "loss": 0.4691,
+      "step": 25775
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7525160769386055e-06,
+      "loss": 0.5479,
+      "step": 25776
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.751939325777893e-06,
+      "loss": 0.3053,
+      "step": 25777
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.751362574617182e-06,
+      "loss": 0.4058,
+      "step": 25778
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.75078582345647e-06,
+      "loss": 0.3614,
+      "step": 25779
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.750209072295758e-06,
+      "loss": 0.4389,
+      "step": 25780
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7496323211350465e-06,
+      "loss": 0.3356,
+      "step": 25781
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.749055569974336e-06,
+      "loss": 0.3619,
+      "step": 25782
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.748478818813623e-06,
+      "loss": 0.3248,
+      "step": 25783
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.747902067652912e-06,
+      "loss": 0.4071,
+      "step": 25784
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7473253164922e-06,
+      "loss": 0.3512,
+      "step": 25785
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.746748565331488e-06,
+      "loss": 0.3236,
+      "step": 25786
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.746171814170777e-06,
+      "loss": 0.4429,
+      "step": 25787
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7455950630100645e-06,
+      "loss": 0.505,
+      "step": 25788
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.745018311849353e-06,
+      "loss": 0.5054,
+      "step": 25789
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.744441560688642e-06,
+      "loss": 0.3599,
+      "step": 25790
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.743864809527929e-06,
+      "loss": 0.5249,
+      "step": 25791
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.743288058367218e-06,
+      "loss": 0.5589,
+      "step": 25792
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.742711307206506e-06,
+      "loss": 0.338,
+      "step": 25793
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.742134556045794e-06,
+      "loss": 0.4035,
+      "step": 25794
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7415578048850826e-06,
+      "loss": 0.3683,
+      "step": 25795
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.740981053724372e-06,
+      "loss": 0.4196,
+      "step": 25796
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.740404302563659e-06,
+      "loss": 0.5113,
+      "step": 25797
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.739827551402948e-06,
+      "loss": 0.4403,
+      "step": 25798
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.739250800242237e-06,
+      "loss": 0.441,
+      "step": 25799
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.738674049081524e-06,
+      "loss": 0.429,
+      "step": 25800
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.738097297920813e-06,
+      "loss": 0.5194,
+      "step": 25801
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.737520546760101e-06,
+      "loss": 0.511,
+      "step": 25802
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.736943795599389e-06,
+      "loss": 0.2866,
+      "step": 25803
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.736367044438678e-06,
+      "loss": 0.3039,
+      "step": 25804
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.735790293277965e-06,
+      "loss": 0.3826,
+      "step": 25805
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.735213542117254e-06,
+      "loss": 0.4462,
+      "step": 25806
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.734636790956542e-06,
+      "loss": 0.4947,
+      "step": 25807
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.73406003979583e-06,
+      "loss": 0.4756,
+      "step": 25808
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.733483288635119e-06,
+      "loss": 0.3582,
+      "step": 25809
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.732906537474407e-06,
+      "loss": 0.407,
+      "step": 25810
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.732329786313695e-06,
+      "loss": 0.387,
+      "step": 25811
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.731753035152983e-06,
+      "loss": 0.457,
+      "step": 25812
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.731176283992273e-06,
+      "loss": 0.4659,
+      "step": 25813
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.73059953283156e-06,
+      "loss": 0.3279,
+      "step": 25814
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.730022781670849e-06,
+      "loss": 0.4361,
+      "step": 25815
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7294460305101375e-06,
+      "loss": 0.4449,
+      "step": 25816
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.728869279349425e-06,
+      "loss": 0.3952,
+      "step": 25817
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.728292528188714e-06,
+      "loss": 0.6406,
+      "step": 25818
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.727715777028001e-06,
+      "loss": 0.3806,
+      "step": 25819
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.72713902586729e-06,
+      "loss": 0.5434,
+      "step": 25820
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7265622747065785e-06,
+      "loss": 0.4325,
+      "step": 25821
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.725985523545866e-06,
+      "loss": 0.4501,
+      "step": 25822
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.725408772385155e-06,
+      "loss": 0.4254,
+      "step": 25823
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.724832021224443e-06,
+      "loss": 0.4029,
+      "step": 25824
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.724255270063731e-06,
+      "loss": 0.5331,
+      "step": 25825
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.7236785189030194e-06,
+      "loss": 0.5262,
+      "step": 25826
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.723101767742309e-06,
+      "loss": 0.3934,
+      "step": 25827
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.722525016581596e-06,
+      "loss": 0.3751,
+      "step": 25828
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 5.721948265420885e-06,
+      "loss": 0.4493,
+      "step": 25829
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7213715142601736e-06,
+      "loss": 0.2896,
+      "step": 25830
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.720794763099461e-06,
+      "loss": 0.3044,
+      "step": 25831
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.72021801193875e-06,
+      "loss": 0.3339,
+      "step": 25832
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.719641260778038e-06,
+      "loss": 0.4521,
+      "step": 25833
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.719064509617326e-06,
+      "loss": 0.4598,
+      "step": 25834
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7184877584566145e-06,
+      "loss": 0.3992,
+      "step": 25835
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.717911007295903e-06,
+      "loss": 0.5029,
+      "step": 25836
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.717334256135191e-06,
+      "loss": 0.4905,
+      "step": 25837
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.716757504974479e-06,
+      "loss": 0.4638,
+      "step": 25838
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.716180753813767e-06,
+      "loss": 0.3815,
+      "step": 25839
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7156040026530555e-06,
+      "loss": 0.4494,
+      "step": 25840
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.715027251492344e-06,
+      "loss": 0.4984,
+      "step": 25841
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.714450500331632e-06,
+      "loss": 0.4615,
+      "step": 25842
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.71387374917092e-06,
+      "loss": 0.4333,
+      "step": 25843
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.71329699801021e-06,
+      "loss": 0.5563,
+      "step": 25844
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.712720246849497e-06,
+      "loss": 0.47,
+      "step": 25845
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.712143495688786e-06,
+      "loss": 0.4891,
+      "step": 25846
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.711566744528074e-06,
+      "loss": 0.5344,
+      "step": 25847
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.710989993367362e-06,
+      "loss": 0.404,
+      "step": 25848
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.710413242206651e-06,
+      "loss": 0.4777,
+      "step": 25849
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.709836491045939e-06,
+      "loss": 0.383,
+      "step": 25850
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.709259739885227e-06,
+      "loss": 0.5166,
+      "step": 25851
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.708682988724515e-06,
+      "loss": 0.3539,
+      "step": 25852
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.708106237563804e-06,
+      "loss": 0.4653,
+      "step": 25853
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7075294864030916e-06,
+      "loss": 0.4846,
+      "step": 25854
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.70695273524238e-06,
+      "loss": 0.3674,
+      "step": 25855
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.706375984081668e-06,
+      "loss": 0.4545,
+      "step": 25856
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.705799232920956e-06,
+      "loss": 0.4253,
+      "step": 25857
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.705222481760246e-06,
+      "loss": 0.4371,
+      "step": 25858
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7046457305995325e-06,
+      "loss": 0.4287,
+      "step": 25859
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.704068979438822e-06,
+      "loss": 0.4149,
+      "step": 25860
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.7034922282781104e-06,
+      "loss": 0.4234,
+      "step": 25861
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.702915477117398e-06,
+      "loss": 0.4206,
+      "step": 25862
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.702338725956687e-06,
+      "loss": 0.5025,
+      "step": 25863
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.701761974795975e-06,
+      "loss": 0.3465,
+      "step": 25864
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.701185223635263e-06,
+      "loss": 0.5541,
+      "step": 25865
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.700608472474551e-06,
+      "loss": 0.3651,
+      "step": 25866
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.70003172131384e-06,
+      "loss": 0.4091,
+      "step": 25867
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.699454970153128e-06,
+      "loss": 0.4144,
+      "step": 25868
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.698878218992416e-06,
+      "loss": 0.39,
+      "step": 25869
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.698301467831705e-06,
+      "loss": 0.4269,
+      "step": 25870
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.697724716670992e-06,
+      "loss": 0.3503,
+      "step": 25871
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.697147965510281e-06,
+      "loss": 0.3577,
+      "step": 25872
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.696571214349569e-06,
+      "loss": 0.3917,
+      "step": 25873
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.695994463188857e-06,
+      "loss": 0.3835,
+      "step": 25874
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6954177120281465e-06,
+      "loss": 0.4845,
+      "step": 25875
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.694840960867434e-06,
+      "loss": 0.2994,
+      "step": 25876
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.694264209706723e-06,
+      "loss": 0.3241,
+      "step": 25877
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.693687458546011e-06,
+      "loss": 0.3194,
+      "step": 25878
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.693110707385299e-06,
+      "loss": 0.4996,
+      "step": 25879
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6925339562245875e-06,
+      "loss": 0.5707,
+      "step": 25880
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.691957205063876e-06,
+      "loss": 0.3577,
+      "step": 25881
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.691380453903164e-06,
+      "loss": 0.3485,
+      "step": 25882
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.690803702742452e-06,
+      "loss": 0.4721,
+      "step": 25883
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.690226951581741e-06,
+      "loss": 0.493,
+      "step": 25884
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6896502004210284e-06,
+      "loss": 0.3887,
+      "step": 25885
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.689073449260317e-06,
+      "loss": 0.4485,
+      "step": 25886
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.688496698099606e-06,
+      "loss": 0.4559,
+      "step": 25887
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.687919946938893e-06,
+      "loss": 0.3918,
+      "step": 25888
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6873431957781826e-06,
+      "loss": 0.3603,
+      "step": 25889
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.686766444617469e-06,
+      "loss": 0.4104,
+      "step": 25890
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.686189693456759e-06,
+      "loss": 0.4771,
+      "step": 25891
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.685612942296047e-06,
+      "loss": 0.5238,
+      "step": 25892
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.685036191135335e-06,
+      "loss": 0.466,
+      "step": 25893
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6844594399746235e-06,
+      "loss": 0.5796,
+      "step": 25894
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.683882688813912e-06,
+      "loss": 0.4134,
+      "step": 25895
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6833059376532e-06,
+      "loss": 0.4893,
+      "step": 25896
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.682729186492488e-06,
+      "loss": 0.3778,
+      "step": 25897
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.682152435331777e-06,
+      "loss": 0.4617,
+      "step": 25898
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6815756841710645e-06,
+      "loss": 0.4539,
+      "step": 25899
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.680998933010353e-06,
+      "loss": 0.4603,
+      "step": 25900
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6804221818496416e-06,
+      "loss": 0.5392,
+      "step": 25901
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.679845430688929e-06,
+      "loss": 0.3868,
+      "step": 25902
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.679268679528218e-06,
+      "loss": 0.4927,
+      "step": 25903
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.678691928367507e-06,
+      "loss": 0.4986,
+      "step": 25904
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.678115177206794e-06,
+      "loss": 0.372,
+      "step": 25905
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.677538426046083e-06,
+      "loss": 0.4578,
+      "step": 25906
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.676961674885371e-06,
+      "loss": 0.5006,
+      "step": 25907
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.67638492372466e-06,
+      "loss": 0.3986,
+      "step": 25908
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.675808172563948e-06,
+      "loss": 0.4349,
+      "step": 25909
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.675231421403236e-06,
+      "loss": 0.3761,
+      "step": 25910
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.674654670242524e-06,
+      "loss": 0.4111,
+      "step": 25911
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.674077919081813e-06,
+      "loss": 0.4827,
+      "step": 25912
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6735011679211006e-06,
+      "loss": 0.4287,
+      "step": 25913
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.672924416760389e-06,
+      "loss": 0.457,
+      "step": 25914
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.672347665599678e-06,
+      "loss": 0.3791,
+      "step": 25915
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.671770914438965e-06,
+      "loss": 0.3728,
+      "step": 25916
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.671194163278254e-06,
+      "loss": 0.4167,
+      "step": 25917
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.670617412117543e-06,
+      "loss": 0.4133,
+      "step": 25918
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.67004066095683e-06,
+      "loss": 0.4507,
+      "step": 25919
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6694639097961194e-06,
+      "loss": 0.4743,
+      "step": 25920
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.668887158635408e-06,
+      "loss": 0.4164,
+      "step": 25921
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.668310407474696e-06,
+      "loss": 0.3771,
+      "step": 25922
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.667733656313984e-06,
+      "loss": 0.4401,
+      "step": 25923
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.667156905153272e-06,
+      "loss": 0.5115,
+      "step": 25924
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.66658015399256e-06,
+      "loss": 0.4106,
+      "step": 25925
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.666003402831849e-06,
+      "loss": 0.321,
+      "step": 25926
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.665426651671137e-06,
+      "loss": 0.4285,
+      "step": 25927
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.664849900510425e-06,
+      "loss": 0.3554,
+      "step": 25928
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.664273149349714e-06,
+      "loss": 0.3426,
+      "step": 25929
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.663696398189001e-06,
+      "loss": 0.5187,
+      "step": 25930
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.66311964702829e-06,
+      "loss": 0.4413,
+      "step": 25931
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6625428958675784e-06,
+      "loss": 0.3387,
+      "step": 25932
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.661966144706866e-06,
+      "loss": 0.4303,
+      "step": 25933
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.661389393546155e-06,
+      "loss": 0.4248,
+      "step": 25934
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.660812642385444e-06,
+      "loss": 0.4595,
+      "step": 25935
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.660235891224731e-06,
+      "loss": 0.4158,
+      "step": 25936
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65965914006402e-06,
+      "loss": 0.4307,
+      "step": 25937
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.659082388903309e-06,
+      "loss": 0.3899,
+      "step": 25938
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6585056377425965e-06,
+      "loss": 0.4498,
+      "step": 25939
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.657928886581885e-06,
+      "loss": 0.3958,
+      "step": 25940
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.657352135421173e-06,
+      "loss": 0.3852,
+      "step": 25941
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.656775384260461e-06,
+      "loss": 0.5198,
+      "step": 25942
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65619863309975e-06,
+      "loss": 0.3856,
+      "step": 25943
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6556218819390374e-06,
+      "loss": 0.4587,
+      "step": 25944
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.655045130778326e-06,
+      "loss": 0.4717,
+      "step": 25945
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6544683796176145e-06,
+      "loss": 0.3463,
+      "step": 25946
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.653891628456902e-06,
+      "loss": 0.3501,
+      "step": 25947
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.653314877296191e-06,
+      "loss": 0.4261,
+      "step": 25948
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.65273812613548e-06,
+      "loss": 0.408,
+      "step": 25949
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.652161374974767e-06,
+      "loss": 0.3409,
+      "step": 25950
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.651584623814056e-06,
+      "loss": 0.3752,
+      "step": 25951
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.651007872653345e-06,
+      "loss": 0.4239,
+      "step": 25952
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6504311214926325e-06,
+      "loss": 0.4165,
+      "step": 25953
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.649854370331921e-06,
+      "loss": 0.4874,
+      "step": 25954
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.64927761917121e-06,
+      "loss": 0.4707,
+      "step": 25955
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.648700868010497e-06,
+      "loss": 0.3674,
+      "step": 25956
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.648124116849786e-06,
+      "loss": 0.4265,
+      "step": 25957
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6475473656890735e-06,
+      "loss": 0.4631,
+      "step": 25958
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.646970614528362e-06,
+      "loss": 0.5532,
+      "step": 25959
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6463938633676506e-06,
+      "loss": 0.3748,
+      "step": 25960
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.645817112206938e-06,
+      "loss": 0.4479,
+      "step": 25961
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.645240361046227e-06,
+      "loss": 0.3908,
+      "step": 25962
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.644663609885515e-06,
+      "loss": 0.3413,
+      "step": 25963
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.644086858724803e-06,
+      "loss": 0.3699,
+      "step": 25964
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6435101075640915e-06,
+      "loss": 0.4466,
+      "step": 25965
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.642933356403381e-06,
+      "loss": 0.3572,
+      "step": 25966
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.642356605242668e-06,
+      "loss": 0.4001,
+      "step": 25967
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.641779854081957e-06,
+      "loss": 0.4316,
+      "step": 25968
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.641203102921246e-06,
+      "loss": 0.2996,
+      "step": 25969
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.640626351760533e-06,
+      "loss": 0.4201,
+      "step": 25970
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.640049600599822e-06,
+      "loss": 0.486,
+      "step": 25971
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.63947284943911e-06,
+      "loss": 0.4554,
+      "step": 25972
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.638896098278398e-06,
+      "loss": 0.4037,
+      "step": 25973
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.638319347117687e-06,
+      "loss": 0.4765,
+      "step": 25974
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.637742595956975e-06,
+      "loss": 0.457,
+      "step": 25975
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.637165844796263e-06,
+      "loss": 0.3873,
+      "step": 25976
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.636589093635551e-06,
+      "loss": 0.353,
+      "step": 25977
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.636012342474839e-06,
+      "loss": 0.4427,
+      "step": 25978
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.635435591314128e-06,
+      "loss": 0.3955,
+      "step": 25979
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.634858840153417e-06,
+      "loss": 0.4857,
+      "step": 25980
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.634282088992704e-06,
+      "loss": 0.4251,
+      "step": 25981
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.633705337831993e-06,
+      "loss": 0.4079,
+      "step": 25982
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.633128586671282e-06,
+      "loss": 0.4953,
+      "step": 25983
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.632551835510569e-06,
+      "loss": 0.3968,
+      "step": 25984
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.631975084349858e-06,
+      "loss": 0.4106,
+      "step": 25985
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6313983331891465e-06,
+      "loss": 0.3951,
+      "step": 25986
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.630821582028434e-06,
+      "loss": 0.5506,
+      "step": 25987
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.630244830867723e-06,
+      "loss": 0.3307,
+      "step": 25988
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.629668079707011e-06,
+      "loss": 0.4255,
+      "step": 25989
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.629091328546299e-06,
+      "loss": 0.4583,
+      "step": 25990
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.628514577385587e-06,
+      "loss": 0.4596,
+      "step": 25991
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.627937826224876e-06,
+      "loss": 0.3269,
+      "step": 25992
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.627361075064164e-06,
+      "loss": 0.4213,
+      "step": 25993
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.626784323903452e-06,
+      "loss": 0.3516,
+      "step": 25994
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.62620757274274e-06,
+      "loss": 0.3705,
+      "step": 25995
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.625630821582028e-06,
+      "loss": 0.3484,
+      "step": 25996
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.625054070421318e-06,
+      "loss": 0.4974,
+      "step": 25997
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.624477319260605e-06,
+      "loss": 0.3205,
+      "step": 25998
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.623900568099894e-06,
+      "loss": 0.4061,
+      "step": 25999
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6233238169391825e-06,
+      "loss": 0.5764,
+      "step": 26000
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.62274706577847e-06,
+      "loss": 0.4348,
+      "step": 26001
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.622170314617759e-06,
+      "loss": 0.5753,
+      "step": 26002
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.621593563457047e-06,
+      "loss": 0.3409,
+      "step": 26003
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.621016812296335e-06,
+      "loss": 0.5202,
+      "step": 26004
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6204400611356235e-06,
+      "loss": 0.5621,
+      "step": 26005
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.619863309974912e-06,
+      "loss": 0.3725,
+      "step": 26006
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.6192865588142e-06,
+      "loss": 0.4831,
+      "step": 26007
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 5.618709807653488e-06,
+      "loss": 0.4307,
+      "step": 26008
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.618133056492777e-06,
+      "loss": 0.3811,
+      "step": 26009
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.6175563053320645e-06,
+      "loss": 0.5079,
+      "step": 26010
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.616979554171354e-06,
+      "loss": 0.4811,
+      "step": 26011
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.616402803010641e-06,
+      "loss": 0.4999,
+      "step": 26012
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.61582605184993e-06,
+      "loss": 0.462,
+      "step": 26013
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.615249300689219e-06,
+      "loss": 0.3916,
+      "step": 26014
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.614672549528506e-06,
+      "loss": 0.4541,
+      "step": 26015
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.614095798367795e-06,
+      "loss": 0.484,
+      "step": 26016
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.613519047207083e-06,
+      "loss": 0.3875,
+      "step": 26017
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.612942296046371e-06,
+      "loss": 0.3357,
+      "step": 26018
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.6123655448856595e-06,
+      "loss": 0.4991,
+      "step": 26019
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.611788793724948e-06,
+      "loss": 0.4221,
+      "step": 26020
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.611212042564236e-06,
+      "loss": 0.389,
+      "step": 26021
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.610635291403524e-06,
+      "loss": 0.4275,
+      "step": 26022
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.610058540242813e-06,
+      "loss": 0.3881,
+      "step": 26023
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.6094817890821005e-06,
+      "loss": 0.3675,
+      "step": 26024
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.608905037921389e-06,
+      "loss": 0.4336,
+      "step": 26025
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.6083282867606784e-06,
+      "loss": 0.4364,
+      "step": 26026
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.607751535599965e-06,
+      "loss": 0.49,
+      "step": 26027
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.607174784439255e-06,
+      "loss": 0.4401,
+      "step": 26028
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.606598033278542e-06,
+      "loss": 0.406,
+      "step": 26029
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.606021282117831e-06,
+      "loss": 0.5974,
+      "step": 26030
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.605444530957119e-06,
+      "loss": 0.2759,
+      "step": 26031
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.604867779796407e-06,
+      "loss": 0.4306,
+      "step": 26032
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.604291028635696e-06,
+      "loss": 0.4235,
+      "step": 26033
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.603714277474984e-06,
+      "loss": 0.3451,
+      "step": 26034
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.603137526314272e-06,
+      "loss": 0.4438,
+      "step": 26035
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.60256077515356e-06,
+      "loss": 0.4004,
+      "step": 26036
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.601984023992849e-06,
+      "loss": 0.3894,
+      "step": 26037
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.6014072728321366e-06,
+      "loss": 0.5656,
+      "step": 26038
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.600830521671425e-06,
+      "loss": 0.4154,
+      "step": 26039
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.600253770510714e-06,
+      "loss": 0.5545,
+      "step": 26040
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.599677019350001e-06,
+      "loss": 0.4226,
+      "step": 26041
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.599100268189291e-06,
+      "loss": 0.3732,
+      "step": 26042
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.598523517028579e-06,
+      "loss": 0.5635,
+      "step": 26043
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.597946765867867e-06,
+      "loss": 0.435,
+      "step": 26044
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5973700147071555e-06,
+      "loss": 0.4007,
+      "step": 26045
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.596793263546443e-06,
+      "loss": 0.4564,
+      "step": 26046
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.596216512385732e-06,
+      "loss": 0.4302,
+      "step": 26047
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.59563976122502e-06,
+      "loss": 0.3159,
+      "step": 26048
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.595063010064308e-06,
+      "loss": 0.4393,
+      "step": 26049
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.594486258903596e-06,
+      "loss": 0.4414,
+      "step": 26050
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.593909507742885e-06,
+      "loss": 0.3297,
+      "step": 26051
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.593332756582173e-06,
+      "loss": 0.4744,
+      "step": 26052
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.592756005421461e-06,
+      "loss": 0.457,
+      "step": 26053
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.59217925426075e-06,
+      "loss": 0.3253,
+      "step": 26054
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.591602503100037e-06,
+      "loss": 0.4758,
+      "step": 26055
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.591025751939326e-06,
+      "loss": 0.4512,
+      "step": 26056
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.590449000778615e-06,
+      "loss": 0.3511,
+      "step": 26057
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.589872249617902e-06,
+      "loss": 0.3634,
+      "step": 26058
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5892954984571915e-06,
+      "loss": 0.4305,
+      "step": 26059
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.58871874729648e-06,
+      "loss": 0.3771,
+      "step": 26060
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.588141996135768e-06,
+      "loss": 0.4766,
+      "step": 26061
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.587565244975056e-06,
+      "loss": 0.447,
+      "step": 26062
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.586988493814344e-06,
+      "loss": 0.5168,
+      "step": 26063
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5864117426536325e-06,
+      "loss": 0.5236,
+      "step": 26064
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.585834991492921e-06,
+      "loss": 0.4123,
+      "step": 26065
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.585258240332209e-06,
+      "loss": 0.4591,
+      "step": 26066
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.584681489171497e-06,
+      "loss": 0.4736,
+      "step": 26067
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.584104738010786e-06,
+      "loss": 0.4096,
+      "step": 26068
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5835279868500734e-06,
+      "loss": 0.3764,
+      "step": 26069
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.582951235689362e-06,
+      "loss": 0.4422,
+      "step": 26070
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.582374484528651e-06,
+      "loss": 0.5217,
+      "step": 26071
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.581797733367938e-06,
+      "loss": 0.2945,
+      "step": 26072
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.581220982207228e-06,
+      "loss": 0.4709,
+      "step": 26073
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.580644231046516e-06,
+      "loss": 0.39,
+      "step": 26074
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.580067479885804e-06,
+      "loss": 0.5402,
+      "step": 26075
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.579490728725092e-06,
+      "loss": 0.3964,
+      "step": 26076
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.578913977564381e-06,
+      "loss": 0.326,
+      "step": 26077
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5783372264036685e-06,
+      "loss": 0.5452,
+      "step": 26078
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.577760475242957e-06,
+      "loss": 0.296,
+      "step": 26079
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.577183724082245e-06,
+      "loss": 0.4905,
+      "step": 26080
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.576606972921533e-06,
+      "loss": 0.4118,
+      "step": 26081
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.576030221760822e-06,
+      "loss": 0.4404,
+      "step": 26082
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5754534706001095e-06,
+      "loss": 0.5129,
+      "step": 26083
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.574876719439398e-06,
+      "loss": 0.326,
+      "step": 26084
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5742999682786866e-06,
+      "loss": 0.5086,
+      "step": 26085
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.573723217117974e-06,
+      "loss": 0.4903,
+      "step": 26086
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.573146465957263e-06,
+      "loss": 0.464,
+      "step": 26087
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.572569714796552e-06,
+      "loss": 0.562,
+      "step": 26088
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.571992963635839e-06,
+      "loss": 0.4427,
+      "step": 26089
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.571416212475128e-06,
+      "loss": 0.4158,
+      "step": 26090
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.570839461314417e-06,
+      "loss": 0.4509,
+      "step": 26091
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.570262710153705e-06,
+      "loss": 0.3108,
+      "step": 26092
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.569685958992993e-06,
+      "loss": 0.405,
+      "step": 26093
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.569109207832282e-06,
+      "loss": 0.3524,
+      "step": 26094
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.568532456671569e-06,
+      "loss": 0.378,
+      "step": 26095
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.567955705510858e-06,
+      "loss": 0.45,
+      "step": 26096
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.567378954350146e-06,
+      "loss": 0.4115,
+      "step": 26097
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.566802203189434e-06,
+      "loss": 0.3724,
+      "step": 26098
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.566225452028723e-06,
+      "loss": 0.3651,
+      "step": 26099
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.56564870086801e-06,
+      "loss": 0.3993,
+      "step": 26100
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.565071949707299e-06,
+      "loss": 0.3072,
+      "step": 26101
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.564495198546588e-06,
+      "loss": 0.6186,
+      "step": 26102
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.563918447385875e-06,
+      "loss": 0.4458,
+      "step": 26103
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5633416962251644e-06,
+      "loss": 0.3477,
+      "step": 26104
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.562764945064453e-06,
+      "loss": 0.4291,
+      "step": 26105
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.562188193903741e-06,
+      "loss": 0.429,
+      "step": 26106
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.561611442743029e-06,
+      "loss": 0.4342,
+      "step": 26107
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.561034691582318e-06,
+      "loss": 0.5307,
+      "step": 26108
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.560457940421605e-06,
+      "loss": 0.4356,
+      "step": 26109
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.559881189260894e-06,
+      "loss": 0.5012,
+      "step": 26110
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5593044381001825e-06,
+      "loss": 0.3201,
+      "step": 26111
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.55872768693947e-06,
+      "loss": 0.4497,
+      "step": 26112
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.558150935778759e-06,
+      "loss": 0.4269,
+      "step": 26113
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.557574184618047e-06,
+      "loss": 0.3567,
+      "step": 26114
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.556997433457335e-06,
+      "loss": 0.4342,
+      "step": 26115
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5564206822966234e-06,
+      "loss": 0.4371,
+      "step": 26116
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.555843931135911e-06,
+      "loss": 0.3533,
+      "step": 26117
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5552671799752e-06,
+      "loss": 0.4466,
+      "step": 26118
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.554690428814489e-06,
+      "loss": 0.3625,
+      "step": 26119
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.554113677653776e-06,
+      "loss": 0.4406,
+      "step": 26120
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.553536926493065e-06,
+      "loss": 0.3696,
+      "step": 26121
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.552960175332354e-06,
+      "loss": 0.316,
+      "step": 26122
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5523834241716415e-06,
+      "loss": 0.4106,
+      "step": 26123
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.55180667301093e-06,
+      "loss": 0.2838,
+      "step": 26124
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5512299218502185e-06,
+      "loss": 0.3751,
+      "step": 26125
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.550653170689506e-06,
+      "loss": 0.3842,
+      "step": 26126
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.550076419528795e-06,
+      "loss": 0.3968,
+      "step": 26127
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.549499668368083e-06,
+      "loss": 0.4636,
+      "step": 26128
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.548922917207371e-06,
+      "loss": 0.3696,
+      "step": 26129
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5483461660466595e-06,
+      "loss": 0.3364,
+      "step": 26130
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.547769414885948e-06,
+      "loss": 0.4297,
+      "step": 26131
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.547192663725236e-06,
+      "loss": 0.4688,
+      "step": 26132
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.546615912564525e-06,
+      "loss": 0.4233,
+      "step": 26133
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.546039161403812e-06,
+      "loss": 0.3947,
+      "step": 26134
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.545462410243101e-06,
+      "loss": 0.4664,
+      "step": 26135
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.54488565908239e-06,
+      "loss": 0.3389,
+      "step": 26136
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5443089079216775e-06,
+      "loss": 0.4475,
+      "step": 26137
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.543732156760966e-06,
+      "loss": 0.4076,
+      "step": 26138
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.543155405600255e-06,
+      "loss": 0.486,
+      "step": 26139
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.542578654439542e-06,
+      "loss": 0.4551,
+      "step": 26140
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.542001903278831e-06,
+      "loss": 0.3673,
+      "step": 26141
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.541425152118119e-06,
+      "loss": 0.3541,
+      "step": 26142
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.540848400957407e-06,
+      "loss": 0.4683,
+      "step": 26143
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5402716497966956e-06,
+      "loss": 0.4186,
+      "step": 26144
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.539694898635984e-06,
+      "loss": 0.4359,
+      "step": 26145
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.539118147475272e-06,
+      "loss": 0.4365,
+      "step": 26146
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.53854139631456e-06,
+      "loss": 0.3997,
+      "step": 26147
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.53796464515385e-06,
+      "loss": 0.4456,
+      "step": 26148
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5373878939931365e-06,
+      "loss": 0.4038,
+      "step": 26149
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.536811142832426e-06,
+      "loss": 0.4818,
+      "step": 26150
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.536234391671713e-06,
+      "loss": 0.3319,
+      "step": 26151
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.535657640511002e-06,
+      "loss": 0.4171,
+      "step": 26152
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.535080889350291e-06,
+      "loss": 0.5468,
+      "step": 26153
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.534504138189578e-06,
+      "loss": 0.3684,
+      "step": 26154
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.533927387028867e-06,
+      "loss": 0.3769,
+      "step": 26155
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.533350635868155e-06,
+      "loss": 0.4975,
+      "step": 26156
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.532773884707443e-06,
+      "loss": 0.4063,
+      "step": 26157
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.532197133546732e-06,
+      "loss": 0.4052,
+      "step": 26158
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.53162038238602e-06,
+      "loss": 0.4523,
+      "step": 26159
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.531043631225308e-06,
+      "loss": 0.2792,
+      "step": 26160
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.530466880064596e-06,
+      "loss": 0.4859,
+      "step": 26161
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.529890128903885e-06,
+      "loss": 0.4538,
+      "step": 26162
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.529313377743173e-06,
+      "loss": 0.4061,
+      "step": 26163
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.528736626582462e-06,
+      "loss": 0.4271,
+      "step": 26164
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5281598754217505e-06,
+      "loss": 0.4199,
+      "step": 26165
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.527583124261038e-06,
+      "loss": 0.411,
+      "step": 26166
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.527006373100327e-06,
+      "loss": 0.2973,
+      "step": 26167
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.526429621939614e-06,
+      "loss": 0.449,
+      "step": 26168
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.525852870778903e-06,
+      "loss": 0.3416,
+      "step": 26169
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5252761196181915e-06,
+      "loss": 0.4793,
+      "step": 26170
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.524699368457479e-06,
+      "loss": 0.4349,
+      "step": 26171
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.524122617296768e-06,
+      "loss": 0.5402,
+      "step": 26172
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.523545866136056e-06,
+      "loss": 0.5202,
+      "step": 26173
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.522969114975344e-06,
+      "loss": 0.3429,
+      "step": 26174
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5223923638146324e-06,
+      "loss": 0.3613,
+      "step": 26175
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.521815612653921e-06,
+      "loss": 0.3153,
+      "step": 26176
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.521238861493209e-06,
+      "loss": 0.3667,
+      "step": 26177
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.520662110332497e-06,
+      "loss": 0.3921,
+      "step": 26178
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5200853591717866e-06,
+      "loss": 0.3033,
+      "step": 26179
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.519508608011073e-06,
+      "loss": 0.4035,
+      "step": 26180
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.518931856850363e-06,
+      "loss": 0.4086,
+      "step": 26181
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.518355105689651e-06,
+      "loss": 0.4411,
+      "step": 26182
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.517778354528939e-06,
+      "loss": 0.5316,
+      "step": 26183
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.5172016033682275e-06,
+      "loss": 0.6216,
+      "step": 26184
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.516624852207515e-06,
+      "loss": 0.4617,
+      "step": 26185
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 5.516048101046804e-06,
+      "loss": 0.3939,
+      "step": 26186
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.515471349886092e-06,
+      "loss": 0.4294,
+      "step": 26187
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.51489459872538e-06,
+      "loss": 0.4291,
+      "step": 26188
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5143178475646685e-06,
+      "loss": 0.419,
+      "step": 26189
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.513741096403957e-06,
+      "loss": 0.4616,
+      "step": 26190
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.513164345243245e-06,
+      "loss": 0.3375,
+      "step": 26191
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.512587594082533e-06,
+      "loss": 0.4042,
+      "step": 26192
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.512010842921822e-06,
+      "loss": 0.3668,
+      "step": 26193
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5114340917611095e-06,
+      "loss": 0.4026,
+      "step": 26194
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.510857340600399e-06,
+      "loss": 0.3295,
+      "step": 26195
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.510280589439687e-06,
+      "loss": 0.3725,
+      "step": 26196
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.509703838278975e-06,
+      "loss": 0.3792,
+      "step": 26197
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.509127087118264e-06,
+      "loss": 0.3389,
+      "step": 26198
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.508550335957552e-06,
+      "loss": 0.3779,
+      "step": 26199
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.50797358479684e-06,
+      "loss": 0.3275,
+      "step": 26200
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.507396833636128e-06,
+      "loss": 0.4268,
+      "step": 26201
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.506820082475416e-06,
+      "loss": 0.4148,
+      "step": 26202
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5062433313147046e-06,
+      "loss": 0.4958,
+      "step": 26203
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.505666580153993e-06,
+      "loss": 0.3645,
+      "step": 26204
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.505089828993281e-06,
+      "loss": 0.4105,
+      "step": 26205
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.504513077832569e-06,
+      "loss": 0.5179,
+      "step": 26206
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.503936326671858e-06,
+      "loss": 0.3553,
+      "step": 26207
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5033595755111455e-06,
+      "loss": 0.3562,
+      "step": 26208
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.502782824350434e-06,
+      "loss": 0.4535,
+      "step": 26209
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5022060731897234e-06,
+      "loss": 0.3946,
+      "step": 26210
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.50162932202901e-06,
+      "loss": 0.3925,
+      "step": 26211
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.5010525708683e-06,
+      "loss": 0.4006,
+      "step": 26212
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.500475819707588e-06,
+      "loss": 0.4423,
+      "step": 26213
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.499899068546876e-06,
+      "loss": 0.5551,
+      "step": 26214
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.499322317386164e-06,
+      "loss": 0.3799,
+      "step": 26215
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.498745566225453e-06,
+      "loss": 0.3511,
+      "step": 26216
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.498168815064741e-06,
+      "loss": 0.4771,
+      "step": 26217
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.497592063904029e-06,
+      "loss": 0.386,
+      "step": 26218
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.497015312743318e-06,
+      "loss": 0.4431,
+      "step": 26219
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.496438561582605e-06,
+      "loss": 0.3274,
+      "step": 26220
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.495861810421894e-06,
+      "loss": 0.3737,
+      "step": 26221
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.495285059261182e-06,
+      "loss": 0.3627,
+      "step": 26222
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.49470830810047e-06,
+      "loss": 0.4239,
+      "step": 26223
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.494131556939759e-06,
+      "loss": 0.4923,
+      "step": 26224
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.493554805779046e-06,
+      "loss": 0.3286,
+      "step": 26225
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.492978054618336e-06,
+      "loss": 0.4309,
+      "step": 26226
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.492401303457624e-06,
+      "loss": 0.4212,
+      "step": 26227
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.491824552296912e-06,
+      "loss": 0.4424,
+      "step": 26228
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4912478011362005e-06,
+      "loss": 0.465,
+      "step": 26229
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.490671049975489e-06,
+      "loss": 0.3784,
+      "step": 26230
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.490094298814777e-06,
+      "loss": 0.424,
+      "step": 26231
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.489517547654065e-06,
+      "loss": 0.4622,
+      "step": 26232
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.488940796493354e-06,
+      "loss": 0.4312,
+      "step": 26233
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4883640453326414e-06,
+      "loss": 0.3605,
+      "step": 26234
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.48778729417193e-06,
+      "loss": 0.4491,
+      "step": 26235
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4872105430112185e-06,
+      "loss": 0.3524,
+      "step": 26236
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.486633791850506e-06,
+      "loss": 0.416,
+      "step": 26237
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.486057040689795e-06,
+      "loss": 0.4204,
+      "step": 26238
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.485480289529082e-06,
+      "loss": 0.5279,
+      "step": 26239
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.484903538368371e-06,
+      "loss": 0.3913,
+      "step": 26240
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.48432678720766e-06,
+      "loss": 0.4527,
+      "step": 26241
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.483750036046947e-06,
+      "loss": 0.411,
+      "step": 26242
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4831732848862365e-06,
+      "loss": 0.4501,
+      "step": 26243
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.482596533725525e-06,
+      "loss": 0.3825,
+      "step": 26244
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.482019782564813e-06,
+      "loss": 0.3454,
+      "step": 26245
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.481443031404101e-06,
+      "loss": 0.4194,
+      "step": 26246
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.48086628024339e-06,
+      "loss": 0.387,
+      "step": 26247
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4802895290826775e-06,
+      "loss": 0.2792,
+      "step": 26248
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.479712777921966e-06,
+      "loss": 0.4447,
+      "step": 26249
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4791360267612546e-06,
+      "loss": 0.3994,
+      "step": 26250
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.478559275600542e-06,
+      "loss": 0.4401,
+      "step": 26251
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.477982524439831e-06,
+      "loss": 0.601,
+      "step": 26252
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.477405773279119e-06,
+      "loss": 0.4109,
+      "step": 26253
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.476829022118407e-06,
+      "loss": 0.3401,
+      "step": 26254
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.476252270957696e-06,
+      "loss": 0.4138,
+      "step": 26255
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.475675519796983e-06,
+      "loss": 0.5485,
+      "step": 26256
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.475098768636273e-06,
+      "loss": 0.4609,
+      "step": 26257
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.474522017475561e-06,
+      "loss": 0.4301,
+      "step": 26258
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.473945266314849e-06,
+      "loss": 0.5817,
+      "step": 26259
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.473368515154137e-06,
+      "loss": 0.3682,
+      "step": 26260
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.472791763993426e-06,
+      "loss": 0.3779,
+      "step": 26261
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4722150128327136e-06,
+      "loss": 0.3951,
+      "step": 26262
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.471638261672002e-06,
+      "loss": 0.3302,
+      "step": 26263
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.471061510511291e-06,
+      "loss": 0.4438,
+      "step": 26264
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.470484759350578e-06,
+      "loss": 0.3614,
+      "step": 26265
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.469908008189867e-06,
+      "loss": 0.4142,
+      "step": 26266
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.469331257029155e-06,
+      "loss": 0.5304,
+      "step": 26267
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.468754505868443e-06,
+      "loss": 0.4192,
+      "step": 26268
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.468177754707732e-06,
+      "loss": 0.3932,
+      "step": 26269
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.467601003547021e-06,
+      "loss": 0.3916,
+      "step": 26270
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.467024252386308e-06,
+      "loss": 0.3865,
+      "step": 26271
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.466447501225597e-06,
+      "loss": 0.499,
+      "step": 26272
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.465870750064884e-06,
+      "loss": 0.359,
+      "step": 26273
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.465293998904173e-06,
+      "loss": 0.4252,
+      "step": 26274
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.464717247743462e-06,
+      "loss": 0.3885,
+      "step": 26275
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.46414049658275e-06,
+      "loss": 0.3829,
+      "step": 26276
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.463563745422038e-06,
+      "loss": 0.2941,
+      "step": 26277
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.462986994261327e-06,
+      "loss": 0.4238,
+      "step": 26278
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.462410243100614e-06,
+      "loss": 0.5129,
+      "step": 26279
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.461833491939903e-06,
+      "loss": 0.3579,
+      "step": 26280
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4612567407791914e-06,
+      "loss": 0.4351,
+      "step": 26281
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.460679989618479e-06,
+      "loss": 0.4723,
+      "step": 26282
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.460103238457768e-06,
+      "loss": 0.3685,
+      "step": 26283
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.459526487297056e-06,
+      "loss": 0.3827,
+      "step": 26284
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.458949736136344e-06,
+      "loss": 0.4185,
+      "step": 26285
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.458372984975633e-06,
+      "loss": 0.4216,
+      "step": 26286
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.457796233814922e-06,
+      "loss": 0.5089,
+      "step": 26287
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4572194826542095e-06,
+      "loss": 0.3584,
+      "step": 26288
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.456642731493498e-06,
+      "loss": 0.4229,
+      "step": 26289
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.456065980332786e-06,
+      "loss": 0.3379,
+      "step": 26290
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.455489229172074e-06,
+      "loss": 0.4504,
+      "step": 26291
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.454912478011363e-06,
+      "loss": 0.4099,
+      "step": 26292
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4543357268506504e-06,
+      "loss": 0.4361,
+      "step": 26293
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.453758975689939e-06,
+      "loss": 0.423,
+      "step": 26294
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4531822245292275e-06,
+      "loss": 0.3186,
+      "step": 26295
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.452605473368515e-06,
+      "loss": 0.4852,
+      "step": 26296
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.452028722207804e-06,
+      "loss": 0.399,
+      "step": 26297
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.451451971047092e-06,
+      "loss": 0.3412,
+      "step": 26298
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.45087521988638e-06,
+      "loss": 0.5165,
+      "step": 26299
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4502984687256685e-06,
+      "loss": 0.4912,
+      "step": 26300
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.449721717564958e-06,
+      "loss": 0.4775,
+      "step": 26301
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.449144966404245e-06,
+      "loss": 0.418,
+      "step": 26302
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.448568215243534e-06,
+      "loss": 0.4333,
+      "step": 26303
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.447991464082823e-06,
+      "loss": 0.3842,
+      "step": 26304
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.44741471292211e-06,
+      "loss": 0.5523,
+      "step": 26305
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.446837961761399e-06,
+      "loss": 0.5051,
+      "step": 26306
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4462612106006865e-06,
+      "loss": 0.4762,
+      "step": 26307
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.445684459439975e-06,
+      "loss": 0.4191,
+      "step": 26308
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4451077082792636e-06,
+      "loss": 0.4221,
+      "step": 26309
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.444530957118551e-06,
+      "loss": 0.3606,
+      "step": 26310
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.44395420595784e-06,
+      "loss": 0.3609,
+      "step": 26311
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.443377454797128e-06,
+      "loss": 0.3511,
+      "step": 26312
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.442800703636416e-06,
+      "loss": 0.6069,
+      "step": 26313
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4422239524757045e-06,
+      "loss": 0.3907,
+      "step": 26314
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.441647201314993e-06,
+      "loss": 0.3561,
+      "step": 26315
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.441070450154281e-06,
+      "loss": 0.392,
+      "step": 26316
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.44049369899357e-06,
+      "loss": 0.5188,
+      "step": 26317
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.439916947832859e-06,
+      "loss": 0.3817,
+      "step": 26318
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.439340196672146e-06,
+      "loss": 0.4563,
+      "step": 26319
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.438763445511435e-06,
+      "loss": 0.4333,
+      "step": 26320
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.438186694350723e-06,
+      "loss": 0.3693,
+      "step": 26321
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.437609943190011e-06,
+      "loss": 0.4545,
+      "step": 26322
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4370331920293e-06,
+      "loss": 0.4656,
+      "step": 26323
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.436456440868587e-06,
+      "loss": 0.4323,
+      "step": 26324
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.435879689707876e-06,
+      "loss": 0.434,
+      "step": 26325
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.435302938547164e-06,
+      "loss": 0.3623,
+      "step": 26326
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.434726187386452e-06,
+      "loss": 0.4381,
+      "step": 26327
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.434149436225741e-06,
+      "loss": 0.4136,
+      "step": 26328
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.433572685065029e-06,
+      "loss": 0.3945,
+      "step": 26329
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.432995933904317e-06,
+      "loss": 0.4608,
+      "step": 26330
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.432419182743605e-06,
+      "loss": 0.4558,
+      "step": 26331
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.431842431582895e-06,
+      "loss": 0.4378,
+      "step": 26332
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4312656804221816e-06,
+      "loss": 0.4305,
+      "step": 26333
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.430688929261471e-06,
+      "loss": 0.4402,
+      "step": 26334
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4301121781007595e-06,
+      "loss": 0.4575,
+      "step": 26335
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.429535426940047e-06,
+      "loss": 0.3388,
+      "step": 26336
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.428958675779336e-06,
+      "loss": 0.4321,
+      "step": 26337
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.428381924618624e-06,
+      "loss": 0.354,
+      "step": 26338
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.427805173457912e-06,
+      "loss": 0.3508,
+      "step": 26339
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4272284222972004e-06,
+      "loss": 0.3264,
+      "step": 26340
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.426651671136488e-06,
+      "loss": 0.3394,
+      "step": 26341
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.426074919975777e-06,
+      "loss": 0.3693,
+      "step": 26342
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.425498168815065e-06,
+      "loss": 0.4419,
+      "step": 26343
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.424921417654353e-06,
+      "loss": 0.5194,
+      "step": 26344
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.424344666493641e-06,
+      "loss": 0.4156,
+      "step": 26345
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.42376791533293e-06,
+      "loss": 0.4239,
+      "step": 26346
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.423191164172218e-06,
+      "loss": 0.5226,
+      "step": 26347
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.422614413011507e-06,
+      "loss": 0.4208,
+      "step": 26348
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4220376618507955e-06,
+      "loss": 0.3614,
+      "step": 26349
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.421460910690083e-06,
+      "loss": 0.3964,
+      "step": 26350
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.420884159529372e-06,
+      "loss": 0.4393,
+      "step": 26351
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.42030740836866e-06,
+      "loss": 0.2893,
+      "step": 26352
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.419730657207948e-06,
+      "loss": 0.5806,
+      "step": 26353
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4191539060472365e-06,
+      "loss": 0.3752,
+      "step": 26354
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.418577154886525e-06,
+      "loss": 0.362,
+      "step": 26355
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.418000403725813e-06,
+      "loss": 0.3997,
+      "step": 26356
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.417423652565101e-06,
+      "loss": 0.3921,
+      "step": 26357
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.41684690140439e-06,
+      "loss": 0.4103,
+      "step": 26358
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4162701502436775e-06,
+      "loss": 0.3289,
+      "step": 26359
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.415693399082966e-06,
+      "loss": 0.4558,
+      "step": 26360
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.415116647922254e-06,
+      "loss": 0.3369,
+      "step": 26361
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.414539896761542e-06,
+      "loss": 0.3931,
+      "step": 26362
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.413963145600832e-06,
+      "loss": 0.4781,
+      "step": 26363
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.4133863944401184e-06,
+      "loss": 0.481,
+      "step": 26364
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 5.412809643279408e-06,
+      "loss": 0.3186,
+      "step": 26365
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.412232892118696e-06,
+      "loss": 0.3885,
+      "step": 26366
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.411656140957984e-06,
+      "loss": 0.4283,
+      "step": 26367
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.4110793897972726e-06,
+      "loss": 0.4989,
+      "step": 26368
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.410502638636561e-06,
+      "loss": 0.5148,
+      "step": 26369
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.409925887475849e-06,
+      "loss": 0.499,
+      "step": 26370
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.409349136315137e-06,
+      "loss": 0.3482,
+      "step": 26371
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.408772385154426e-06,
+      "loss": 0.2515,
+      "step": 26372
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.4081956339937135e-06,
+      "loss": 0.3864,
+      "step": 26373
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.407618882833002e-06,
+      "loss": 0.3418,
+      "step": 26374
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.407042131672291e-06,
+      "loss": 0.409,
+      "step": 26375
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.406465380511578e-06,
+      "loss": 0.3076,
+      "step": 26376
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.405888629350867e-06,
+      "loss": 0.5868,
+      "step": 26377
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.4053118781901545e-06,
+      "loss": 0.496,
+      "step": 26378
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.404735127029444e-06,
+      "loss": 0.3628,
+      "step": 26379
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.404158375868732e-06,
+      "loss": 0.3564,
+      "step": 26380
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.40358162470802e-06,
+      "loss": 0.3358,
+      "step": 26381
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.403004873547309e-06,
+      "loss": 0.277,
+      "step": 26382
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.402428122386597e-06,
+      "loss": 0.4372,
+      "step": 26383
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.401851371225885e-06,
+      "loss": 0.3663,
+      "step": 26384
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.401274620065173e-06,
+      "loss": 0.3806,
+      "step": 26385
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.400697868904462e-06,
+      "loss": 0.434,
+      "step": 26386
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.40012111774375e-06,
+      "loss": 0.2717,
+      "step": 26387
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.399544366583038e-06,
+      "loss": 0.352,
+      "step": 26388
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.398967615422327e-06,
+      "loss": 0.5385,
+      "step": 26389
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.398390864261614e-06,
+      "loss": 0.4641,
+      "step": 26390
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.397814113100903e-06,
+      "loss": 0.38,
+      "step": 26391
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.397237361940192e-06,
+      "loss": 0.4701,
+      "step": 26392
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.396660610779479e-06,
+      "loss": 0.4469,
+      "step": 26393
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3960838596187685e-06,
+      "loss": 0.3946,
+      "step": 26394
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.395507108458055e-06,
+      "loss": 0.437,
+      "step": 26395
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.394930357297345e-06,
+      "loss": 0.4177,
+      "step": 26396
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.394353606136633e-06,
+      "loss": 0.2479,
+      "step": 26397
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.393776854975921e-06,
+      "loss": 0.4342,
+      "step": 26398
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3932001038152094e-06,
+      "loss": 0.4101,
+      "step": 26399
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.392623352654498e-06,
+      "loss": 0.5015,
+      "step": 26400
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.392046601493786e-06,
+      "loss": 0.4949,
+      "step": 26401
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.391469850333074e-06,
+      "loss": 0.3779,
+      "step": 26402
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.390893099172363e-06,
+      "loss": 0.4823,
+      "step": 26403
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.39031634801165e-06,
+      "loss": 0.4235,
+      "step": 26404
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.389739596850939e-06,
+      "loss": 0.3537,
+      "step": 26405
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3891628456902275e-06,
+      "loss": 0.3886,
+      "step": 26406
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.388586094529515e-06,
+      "loss": 0.5164,
+      "step": 26407
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.388009343368804e-06,
+      "loss": 0.3039,
+      "step": 26408
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.387432592208093e-06,
+      "loss": 0.3605,
+      "step": 26409
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.386855841047381e-06,
+      "loss": 0.5866,
+      "step": 26410
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.386279089886669e-06,
+      "loss": 0.4052,
+      "step": 26411
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.385702338725957e-06,
+      "loss": 0.3654,
+      "step": 26412
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3851255875652455e-06,
+      "loss": 0.4618,
+      "step": 26413
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.384548836404534e-06,
+      "loss": 0.4295,
+      "step": 26414
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.383972085243822e-06,
+      "loss": 0.4778,
+      "step": 26415
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.38339533408311e-06,
+      "loss": 0.3993,
+      "step": 26416
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.382818582922399e-06,
+      "loss": 0.3695,
+      "step": 26417
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3822418317616865e-06,
+      "loss": 0.4331,
+      "step": 26418
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.381665080600975e-06,
+      "loss": 0.381,
+      "step": 26419
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3810883294402635e-06,
+      "loss": 0.5086,
+      "step": 26420
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.380511578279551e-06,
+      "loss": 0.3897,
+      "step": 26421
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.37993482711884e-06,
+      "loss": 0.4182,
+      "step": 26422
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.379358075958129e-06,
+      "loss": 0.3825,
+      "step": 26423
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.378781324797416e-06,
+      "loss": 0.4195,
+      "step": 26424
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.378204573636705e-06,
+      "loss": 0.3647,
+      "step": 26425
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.377627822475994e-06,
+      "loss": 0.4542,
+      "step": 26426
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3770510713152816e-06,
+      "loss": 0.3728,
+      "step": 26427
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.37647432015457e-06,
+      "loss": 0.4974,
+      "step": 26428
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.375897568993858e-06,
+      "loss": 0.3344,
+      "step": 26429
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.375320817833146e-06,
+      "loss": 0.4407,
+      "step": 26430
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.374744066672435e-06,
+      "loss": 0.3782,
+      "step": 26431
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3741673155117225e-06,
+      "loss": 0.3416,
+      "step": 26432
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.373590564351011e-06,
+      "loss": 0.3823,
+      "step": 26433
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3730138131903e-06,
+      "loss": 0.4625,
+      "step": 26434
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.372437062029587e-06,
+      "loss": 0.3513,
+      "step": 26435
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.371860310868876e-06,
+      "loss": 0.3576,
+      "step": 26436
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.371283559708164e-06,
+      "loss": 0.4194,
+      "step": 26437
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.370706808547452e-06,
+      "loss": 0.2987,
+      "step": 26438
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.370130057386741e-06,
+      "loss": 0.3995,
+      "step": 26439
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.36955330622603e-06,
+      "loss": 0.514,
+      "step": 26440
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.368976555065318e-06,
+      "loss": 0.4032,
+      "step": 26441
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.368399803904606e-06,
+      "loss": 0.3965,
+      "step": 26442
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.367823052743895e-06,
+      "loss": 0.4844,
+      "step": 26443
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.367246301583182e-06,
+      "loss": 0.4908,
+      "step": 26444
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.366669550422471e-06,
+      "loss": 0.4629,
+      "step": 26445
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.366092799261759e-06,
+      "loss": 0.4179,
+      "step": 26446
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.365516048101047e-06,
+      "loss": 0.3389,
+      "step": 26447
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.364939296940336e-06,
+      "loss": 0.4052,
+      "step": 26448
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.364362545779623e-06,
+      "loss": 0.5101,
+      "step": 26449
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.363785794618912e-06,
+      "loss": 0.4078,
+      "step": 26450
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3632090434582e-06,
+      "loss": 0.3752,
+      "step": 26451
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.362632292297488e-06,
+      "loss": 0.435,
+      "step": 26452
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.362055541136777e-06,
+      "loss": 0.4274,
+      "step": 26453
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.361478789976066e-06,
+      "loss": 0.4932,
+      "step": 26454
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.360902038815353e-06,
+      "loss": 0.5546,
+      "step": 26455
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.360325287654642e-06,
+      "loss": 0.3584,
+      "step": 26456
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.359748536493931e-06,
+      "loss": 0.3775,
+      "step": 26457
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3591717853332184e-06,
+      "loss": 0.2965,
+      "step": 26458
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.358595034172507e-06,
+      "loss": 0.3755,
+      "step": 26459
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3580182830117955e-06,
+      "loss": 0.4987,
+      "step": 26460
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.357441531851083e-06,
+      "loss": 0.4369,
+      "step": 26461
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.356864780690372e-06,
+      "loss": 0.347,
+      "step": 26462
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.356288029529659e-06,
+      "loss": 0.4289,
+      "step": 26463
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.355711278368948e-06,
+      "loss": 0.3948,
+      "step": 26464
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3551345272082365e-06,
+      "loss": 0.3917,
+      "step": 26465
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.354557776047524e-06,
+      "loss": 0.4323,
+      "step": 26466
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.353981024886813e-06,
+      "loss": 0.4204,
+      "step": 26467
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.353404273726101e-06,
+      "loss": 0.49,
+      "step": 26468
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.352827522565389e-06,
+      "loss": 0.4161,
+      "step": 26469
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.352250771404678e-06,
+      "loss": 0.4535,
+      "step": 26470
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.351674020243967e-06,
+      "loss": 0.4373,
+      "step": 26471
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3510972690832545e-06,
+      "loss": 0.4204,
+      "step": 26472
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.350520517922543e-06,
+      "loss": 0.432,
+      "step": 26473
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3499437667618316e-06,
+      "loss": 0.5528,
+      "step": 26474
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.349367015601119e-06,
+      "loss": 0.3735,
+      "step": 26475
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.348790264440408e-06,
+      "loss": 0.4814,
+      "step": 26476
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.348213513279696e-06,
+      "loss": 0.3494,
+      "step": 26477
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.347636762118984e-06,
+      "loss": 0.4854,
+      "step": 26478
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3470600109582725e-06,
+      "loss": 0.3761,
+      "step": 26479
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.346483259797561e-06,
+      "loss": 0.3347,
+      "step": 26480
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.345906508636849e-06,
+      "loss": 0.5345,
+      "step": 26481
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.345329757476137e-06,
+      "loss": 0.5035,
+      "step": 26482
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.344753006315425e-06,
+      "loss": 0.3988,
+      "step": 26483
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3441762551547135e-06,
+      "loss": 0.5181,
+      "step": 26484
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.343599503994003e-06,
+      "loss": 0.3831,
+      "step": 26485
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.34302275283329e-06,
+      "loss": 0.5108,
+      "step": 26486
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.342446001672579e-06,
+      "loss": 0.4653,
+      "step": 26487
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.341869250511868e-06,
+      "loss": 0.4851,
+      "step": 26488
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.341292499351155e-06,
+      "loss": 0.4489,
+      "step": 26489
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.340715748190444e-06,
+      "loss": 0.4427,
+      "step": 26490
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.340138997029732e-06,
+      "loss": 0.4739,
+      "step": 26491
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.33956224586902e-06,
+      "loss": 0.3404,
+      "step": 26492
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.338985494708309e-06,
+      "loss": 0.5093,
+      "step": 26493
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.338408743547597e-06,
+      "loss": 0.3665,
+      "step": 26494
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.337831992386885e-06,
+      "loss": 0.4788,
+      "step": 26495
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.337255241226173e-06,
+      "loss": 0.3579,
+      "step": 26496
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.336678490065462e-06,
+      "loss": 0.3903,
+      "step": 26497
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3361017389047495e-06,
+      "loss": 0.2948,
+      "step": 26498
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.335524987744038e-06,
+      "loss": 0.4562,
+      "step": 26499
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.334948236583326e-06,
+      "loss": 0.5048,
+      "step": 26500
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.334371485422615e-06,
+      "loss": 0.3286,
+      "step": 26501
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.333794734261904e-06,
+      "loss": 0.454,
+      "step": 26502
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.333217983101191e-06,
+      "loss": 0.4152,
+      "step": 26503
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.33264123194048e-06,
+      "loss": 0.4978,
+      "step": 26504
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3320644807797684e-06,
+      "loss": 0.3991,
+      "step": 26505
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.331487729619056e-06,
+      "loss": 0.4179,
+      "step": 26506
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.330910978458345e-06,
+      "loss": 0.5732,
+      "step": 26507
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.330334227297633e-06,
+      "loss": 0.4505,
+      "step": 26508
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.329757476136921e-06,
+      "loss": 0.6128,
+      "step": 26509
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.329180724976209e-06,
+      "loss": 0.3906,
+      "step": 26510
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.328603973815498e-06,
+      "loss": 0.3801,
+      "step": 26511
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.328027222654786e-06,
+      "loss": 0.3338,
+      "step": 26512
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.327450471494074e-06,
+      "loss": 0.2961,
+      "step": 26513
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3268737203333635e-06,
+      "loss": 0.3496,
+      "step": 26514
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.32629696917265e-06,
+      "loss": 0.4574,
+      "step": 26515
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.32572021801194e-06,
+      "loss": 0.4721,
+      "step": 26516
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.325143466851227e-06,
+      "loss": 0.4759,
+      "step": 26517
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.324566715690516e-06,
+      "loss": 0.4409,
+      "step": 26518
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3239899645298045e-06,
+      "loss": 0.4729,
+      "step": 26519
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.323413213369092e-06,
+      "loss": 0.4515,
+      "step": 26520
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.322836462208381e-06,
+      "loss": 0.5287,
+      "step": 26521
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.322259711047669e-06,
+      "loss": 0.5069,
+      "step": 26522
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.321682959886957e-06,
+      "loss": 0.4046,
+      "step": 26523
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3211062087262455e-06,
+      "loss": 0.4528,
+      "step": 26524
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.320529457565534e-06,
+      "loss": 0.3744,
+      "step": 26525
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.319952706404822e-06,
+      "loss": 0.4817,
+      "step": 26526
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.31937595524411e-06,
+      "loss": 0.3948,
+      "step": 26527
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.318799204083399e-06,
+      "loss": 0.438,
+      "step": 26528
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.318222452922686e-06,
+      "loss": 0.431,
+      "step": 26529
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.317645701761975e-06,
+      "loss": 0.43,
+      "step": 26530
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.317068950601264e-06,
+      "loss": 0.3751,
+      "step": 26531
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.316492199440552e-06,
+      "loss": 0.4573,
+      "step": 26532
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3159154482798406e-06,
+      "loss": 0.4217,
+      "step": 26533
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.315338697119128e-06,
+      "loss": 0.4776,
+      "step": 26534
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.314761945958417e-06,
+      "loss": 0.5069,
+      "step": 26535
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.314185194797705e-06,
+      "loss": 0.3348,
+      "step": 26536
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.313608443636993e-06,
+      "loss": 0.3808,
+      "step": 26537
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3130316924762815e-06,
+      "loss": 0.4042,
+      "step": 26538
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.31245494131557e-06,
+      "loss": 0.4217,
+      "step": 26539
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.311878190154858e-06,
+      "loss": 0.4787,
+      "step": 26540
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.311301438994146e-06,
+      "loss": 0.398,
+      "step": 26541
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.310724687833435e-06,
+      "loss": 0.3535,
+      "step": 26542
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.3101479366727225e-06,
+      "loss": 0.3337,
+      "step": 26543
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 5.309571185512011e-06,
+      "loss": 0.4509,
+      "step": 26544
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.3089944343513e-06,
+      "loss": 0.4055,
+      "step": 26545
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.308417683190587e-06,
+      "loss": 0.4029,
+      "step": 26546
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.307840932029877e-06,
+      "loss": 0.4415,
+      "step": 26547
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.307264180869165e-06,
+      "loss": 0.4232,
+      "step": 26548
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.306687429708453e-06,
+      "loss": 0.4869,
+      "step": 26549
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.306110678547741e-06,
+      "loss": 0.4168,
+      "step": 26550
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.305533927387029e-06,
+      "loss": 0.3586,
+      "step": 26551
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.304957176226318e-06,
+      "loss": 0.3987,
+      "step": 26552
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.304380425065606e-06,
+      "loss": 0.4741,
+      "step": 26553
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.303803673904894e-06,
+      "loss": 0.3977,
+      "step": 26554
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.303226922744182e-06,
+      "loss": 0.3622,
+      "step": 26555
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.302650171583471e-06,
+      "loss": 0.4416,
+      "step": 26556
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.3020734204227585e-06,
+      "loss": 0.4088,
+      "step": 26557
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.301496669262047e-06,
+      "loss": 0.4476,
+      "step": 26558
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.300919918101336e-06,
+      "loss": 0.4114,
+      "step": 26559
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.300343166940623e-06,
+      "loss": 0.3784,
+      "step": 26560
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.299766415779912e-06,
+      "loss": 0.3658,
+      "step": 26561
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.299189664619201e-06,
+      "loss": 0.5357,
+      "step": 26562
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.298612913458489e-06,
+      "loss": 0.343,
+      "step": 26563
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2980361622977774e-06,
+      "loss": 0.3727,
+      "step": 26564
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.297459411137066e-06,
+      "loss": 0.4587,
+      "step": 26565
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.296882659976354e-06,
+      "loss": 0.3897,
+      "step": 26566
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.296305908815642e-06,
+      "loss": 0.5453,
+      "step": 26567
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.29572915765493e-06,
+      "loss": 0.3417,
+      "step": 26568
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.295152406494218e-06,
+      "loss": 0.3973,
+      "step": 26569
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.294575655333507e-06,
+      "loss": 0.3394,
+      "step": 26570
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.293998904172795e-06,
+      "loss": 0.4947,
+      "step": 26571
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.293422153012083e-06,
+      "loss": 0.4878,
+      "step": 26572
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.292845401851372e-06,
+      "loss": 0.3215,
+      "step": 26573
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.292268650690659e-06,
+      "loss": 0.3941,
+      "step": 26574
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.291691899529948e-06,
+      "loss": 0.3156,
+      "step": 26575
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.291115148369237e-06,
+      "loss": 0.4022,
+      "step": 26576
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.290538397208524e-06,
+      "loss": 0.4665,
+      "step": 26577
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2899616460478135e-06,
+      "loss": 0.4358,
+      "step": 26578
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.289384894887102e-06,
+      "loss": 0.3769,
+      "step": 26579
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.28880814372639e-06,
+      "loss": 0.3814,
+      "step": 26580
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.288231392565678e-06,
+      "loss": 0.363,
+      "step": 26581
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.287654641404967e-06,
+      "loss": 0.4991,
+      "step": 26582
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2870778902442545e-06,
+      "loss": 0.4486,
+      "step": 26583
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.286501139083543e-06,
+      "loss": 0.3433,
+      "step": 26584
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.285924387922831e-06,
+      "loss": 0.3971,
+      "step": 26585
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.285347636762119e-06,
+      "loss": 0.3005,
+      "step": 26586
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.284770885601408e-06,
+      "loss": 0.3008,
+      "step": 26587
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.284194134440695e-06,
+      "loss": 0.403,
+      "step": 26588
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.283617383279984e-06,
+      "loss": 0.5399,
+      "step": 26589
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2830406321192725e-06,
+      "loss": 0.5092,
+      "step": 26590
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.28246388095856e-06,
+      "loss": 0.3785,
+      "step": 26591
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.281887129797849e-06,
+      "loss": 0.5379,
+      "step": 26592
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.281310378637138e-06,
+      "loss": 0.406,
+      "step": 26593
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.280733627476426e-06,
+      "loss": 0.3816,
+      "step": 26594
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.280156876315714e-06,
+      "loss": 0.5651,
+      "step": 26595
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.279580125155003e-06,
+      "loss": 0.385,
+      "step": 26596
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2790033739942905e-06,
+      "loss": 0.397,
+      "step": 26597
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.278426622833579e-06,
+      "loss": 0.3731,
+      "step": 26598
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.277849871672868e-06,
+      "loss": 0.397,
+      "step": 26599
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.277273120512155e-06,
+      "loss": 0.3364,
+      "step": 26600
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.276696369351444e-06,
+      "loss": 0.3579,
+      "step": 26601
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.276119618190732e-06,
+      "loss": 0.4181,
+      "step": 26602
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.27554286703002e-06,
+      "loss": 0.384,
+      "step": 26603
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2749661158693085e-06,
+      "loss": 0.4234,
+      "step": 26604
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.274389364708596e-06,
+      "loss": 0.447,
+      "step": 26605
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.273812613547885e-06,
+      "loss": 0.4049,
+      "step": 26606
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.273235862387174e-06,
+      "loss": 0.4084,
+      "step": 26607
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.272659111226461e-06,
+      "loss": 0.3151,
+      "step": 26608
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.27208236006575e-06,
+      "loss": 0.4154,
+      "step": 26609
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.271505608905039e-06,
+      "loss": 0.444,
+      "step": 26610
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2709288577443266e-06,
+      "loss": 0.3322,
+      "step": 26611
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.270352106583615e-06,
+      "loss": 0.4167,
+      "step": 26612
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.269775355422904e-06,
+      "loss": 0.3764,
+      "step": 26613
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.269198604262191e-06,
+      "loss": 0.4613,
+      "step": 26614
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.26862185310148e-06,
+      "loss": 0.4154,
+      "step": 26615
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.268045101940768e-06,
+      "loss": 0.4125,
+      "step": 26616
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.267468350780056e-06,
+      "loss": 0.42,
+      "step": 26617
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.266891599619345e-06,
+      "loss": 0.4845,
+      "step": 26618
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.266314848458633e-06,
+      "loss": 0.4042,
+      "step": 26619
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.265738097297921e-06,
+      "loss": 0.4013,
+      "step": 26620
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.265161346137209e-06,
+      "loss": 0.388,
+      "step": 26621
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.264584594976497e-06,
+      "loss": 0.4694,
+      "step": 26622
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.264007843815786e-06,
+      "loss": 0.4121,
+      "step": 26623
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.263431092655075e-06,
+      "loss": 0.4212,
+      "step": 26624
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.262854341494363e-06,
+      "loss": 0.3706,
+      "step": 26625
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.262277590333651e-06,
+      "loss": 0.4282,
+      "step": 26626
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.26170083917294e-06,
+      "loss": 0.3635,
+      "step": 26627
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.261124088012227e-06,
+      "loss": 0.3133,
+      "step": 26628
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.260547336851516e-06,
+      "loss": 0.472,
+      "step": 26629
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2599705856908045e-06,
+      "loss": 0.5246,
+      "step": 26630
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.259393834530092e-06,
+      "loss": 0.3264,
+      "step": 26631
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.258817083369381e-06,
+      "loss": 0.418,
+      "step": 26632
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.258240332208669e-06,
+      "loss": 0.4177,
+      "step": 26633
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.257663581047957e-06,
+      "loss": 0.4076,
+      "step": 26634
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.257086829887245e-06,
+      "loss": 0.3499,
+      "step": 26635
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.256510078726535e-06,
+      "loss": 0.3168,
+      "step": 26636
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.255933327565822e-06,
+      "loss": 0.3344,
+      "step": 26637
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.255356576405111e-06,
+      "loss": 0.4446,
+      "step": 26638
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.254779825244398e-06,
+      "loss": 0.4574,
+      "step": 26639
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.254203074083687e-06,
+      "loss": 0.4366,
+      "step": 26640
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.253626322922976e-06,
+      "loss": 0.3848,
+      "step": 26641
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2530495717622634e-06,
+      "loss": 0.4835,
+      "step": 26642
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.252472820601552e-06,
+      "loss": 0.4184,
+      "step": 26643
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2518960694408405e-06,
+      "loss": 0.4323,
+      "step": 26644
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.251319318280128e-06,
+      "loss": 0.4679,
+      "step": 26645
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.250742567119417e-06,
+      "loss": 0.5563,
+      "step": 26646
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.250165815958705e-06,
+      "loss": 0.4508,
+      "step": 26647
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.249589064797993e-06,
+      "loss": 0.486,
+      "step": 26648
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2490123136372815e-06,
+      "loss": 0.4772,
+      "step": 26649
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.24843556247657e-06,
+      "loss": 0.3799,
+      "step": 26650
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.247858811315858e-06,
+      "loss": 0.405,
+      "step": 26651
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.247282060155146e-06,
+      "loss": 0.3531,
+      "step": 26652
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.246705308994436e-06,
+      "loss": 0.3762,
+      "step": 26653
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.246128557833723e-06,
+      "loss": 0.3492,
+      "step": 26654
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.245551806673012e-06,
+      "loss": 0.3829,
+      "step": 26655
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2449750555122995e-06,
+      "loss": 0.444,
+      "step": 26656
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.244398304351588e-06,
+      "loss": 0.4096,
+      "step": 26657
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2438215531908766e-06,
+      "loss": 0.3593,
+      "step": 26658
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.243244802030164e-06,
+      "loss": 0.3763,
+      "step": 26659
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.242668050869453e-06,
+      "loss": 0.3535,
+      "step": 26660
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.242091299708741e-06,
+      "loss": 0.4194,
+      "step": 26661
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.241514548548029e-06,
+      "loss": 0.3819,
+      "step": 26662
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2409377973873175e-06,
+      "loss": 0.4413,
+      "step": 26663
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.240361046226606e-06,
+      "loss": 0.3781,
+      "step": 26664
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.239784295065894e-06,
+      "loss": 0.3542,
+      "step": 26665
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.239207543905182e-06,
+      "loss": 0.4205,
+      "step": 26666
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.238630792744472e-06,
+      "loss": 0.3687,
+      "step": 26667
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2380540415837585e-06,
+      "loss": 0.3774,
+      "step": 26668
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.237477290423048e-06,
+      "loss": 0.4099,
+      "step": 26669
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.236900539262336e-06,
+      "loss": 0.4671,
+      "step": 26670
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.236323788101624e-06,
+      "loss": 0.4267,
+      "step": 26671
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.235747036940913e-06,
+      "loss": 0.5757,
+      "step": 26672
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2351702857802e-06,
+      "loss": 0.4608,
+      "step": 26673
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.234593534619489e-06,
+      "loss": 0.3546,
+      "step": 26674
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.234016783458777e-06,
+      "loss": 0.418,
+      "step": 26675
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.233440032298065e-06,
+      "loss": 0.4464,
+      "step": 26676
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.232863281137354e-06,
+      "loss": 0.3557,
+      "step": 26677
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.232286529976642e-06,
+      "loss": 0.3683,
+      "step": 26678
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.23170977881593e-06,
+      "loss": 0.411,
+      "step": 26679
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.231133027655218e-06,
+      "loss": 0.3798,
+      "step": 26680
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.230556276494507e-06,
+      "loss": 0.4631,
+      "step": 26681
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2299795253337946e-06,
+      "loss": 0.4454,
+      "step": 26682
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.229402774173083e-06,
+      "loss": 0.37,
+      "step": 26683
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2288260230123725e-06,
+      "loss": 0.4032,
+      "step": 26684
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.22824927185166e-06,
+      "loss": 0.4937,
+      "step": 26685
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.227672520690949e-06,
+      "loss": 0.3927,
+      "step": 26686
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.227095769530237e-06,
+      "loss": 0.3587,
+      "step": 26687
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.226519018369525e-06,
+      "loss": 0.4357,
+      "step": 26688
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2259422672088134e-06,
+      "loss": 0.4757,
+      "step": 26689
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.225365516048101e-06,
+      "loss": 0.4095,
+      "step": 26690
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.22478876488739e-06,
+      "loss": 0.4573,
+      "step": 26691
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.224212013726678e-06,
+      "loss": 0.3574,
+      "step": 26692
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.223635262565966e-06,
+      "loss": 0.4321,
+      "step": 26693
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.223058511405254e-06,
+      "loss": 0.4441,
+      "step": 26694
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.222481760244543e-06,
+      "loss": 0.2756,
+      "step": 26695
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.221905009083831e-06,
+      "loss": 0.4617,
+      "step": 26696
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.221328257923119e-06,
+      "loss": 0.3175,
+      "step": 26697
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2207515067624085e-06,
+      "loss": 0.3789,
+      "step": 26698
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.220174755601695e-06,
+      "loss": 0.3145,
+      "step": 26699
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.219598004440985e-06,
+      "loss": 0.5165,
+      "step": 26700
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.219021253280273e-06,
+      "loss": 0.3034,
+      "step": 26701
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.218444502119561e-06,
+      "loss": 0.2768,
+      "step": 26702
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2178677509588495e-06,
+      "loss": 0.4115,
+      "step": 26703
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.217290999798138e-06,
+      "loss": 0.3749,
+      "step": 26704
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.216714248637426e-06,
+      "loss": 0.4512,
+      "step": 26705
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.216137497476714e-06,
+      "loss": 0.4389,
+      "step": 26706
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.215560746316002e-06,
+      "loss": 0.4475,
+      "step": 26707
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2149839951552905e-06,
+      "loss": 0.5098,
+      "step": 26708
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.214407243994579e-06,
+      "loss": 0.3934,
+      "step": 26709
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.213830492833867e-06,
+      "loss": 0.4196,
+      "step": 26710
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.213253741673155e-06,
+      "loss": 0.3548,
+      "step": 26711
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.212676990512444e-06,
+      "loss": 0.5434,
+      "step": 26712
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2121002393517314e-06,
+      "loss": 0.3468,
+      "step": 26713
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.21152348819102e-06,
+      "loss": 0.4343,
+      "step": 26714
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.210946737030309e-06,
+      "loss": 0.5077,
+      "step": 26715
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.210369985869597e-06,
+      "loss": 0.477,
+      "step": 26716
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2097932347088856e-06,
+      "loss": 0.3807,
+      "step": 26717
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.209216483548174e-06,
+      "loss": 0.4088,
+      "step": 26718
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.208639732387462e-06,
+      "loss": 0.3789,
+      "step": 26719
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.20806298122675e-06,
+      "loss": 0.4816,
+      "step": 26720
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.207486230066039e-06,
+      "loss": 0.3939,
+      "step": 26721
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.2069094789053265e-06,
+      "loss": 0.4002,
+      "step": 26722
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 5.206332727744615e-06,
+      "loss": 0.4399,
+      "step": 26723
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.205755976583903e-06,
+      "loss": 0.3744,
+      "step": 26724
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.205179225423191e-06,
+      "loss": 0.3013,
+      "step": 26725
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.20460247426248e-06,
+      "loss": 0.3186,
+      "step": 26726
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.2040257231017675e-06,
+      "loss": 0.3183,
+      "step": 26727
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.203448971941056e-06,
+      "loss": 0.4592,
+      "step": 26728
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.202872220780345e-06,
+      "loss": 0.4368,
+      "step": 26729
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.202295469619632e-06,
+      "loss": 0.4891,
+      "step": 26730
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.201718718458922e-06,
+      "loss": 0.3465,
+      "step": 26731
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.20114196729821e-06,
+      "loss": 0.4222,
+      "step": 26732
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.200565216137498e-06,
+      "loss": 0.4629,
+      "step": 26733
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.199988464976786e-06,
+      "loss": 0.4258,
+      "step": 26734
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.199411713816075e-06,
+      "loss": 0.5159,
+      "step": 26735
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.198834962655363e-06,
+      "loss": 0.3734,
+      "step": 26736
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.198258211494651e-06,
+      "loss": 0.4015,
+      "step": 26737
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.19768146033394e-06,
+      "loss": 0.426,
+      "step": 26738
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.197104709173227e-06,
+      "loss": 0.355,
+      "step": 26739
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.196527958012516e-06,
+      "loss": 0.4154,
+      "step": 26740
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.195951206851804e-06,
+      "loss": 0.3111,
+      "step": 26741
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.195374455691092e-06,
+      "loss": 0.3747,
+      "step": 26742
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.194797704530381e-06,
+      "loss": 0.3136,
+      "step": 26743
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.194220953369668e-06,
+      "loss": 0.4436,
+      "step": 26744
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.193644202208957e-06,
+      "loss": 0.4148,
+      "step": 26745
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.193067451048246e-06,
+      "loss": 0.4046,
+      "step": 26746
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.192490699887534e-06,
+      "loss": 0.4545,
+      "step": 26747
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1919139487268224e-06,
+      "loss": 0.4291,
+      "step": 26748
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.191337197566111e-06,
+      "loss": 0.4135,
+      "step": 26749
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.190760446405399e-06,
+      "loss": 0.2973,
+      "step": 26750
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.190183695244687e-06,
+      "loss": 0.3344,
+      "step": 26751
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.189606944083976e-06,
+      "loss": 0.3951,
+      "step": 26752
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.189030192923263e-06,
+      "loss": 0.5351,
+      "step": 26753
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.188453441762552e-06,
+      "loss": 0.3154,
+      "step": 26754
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1878766906018405e-06,
+      "loss": 0.3858,
+      "step": 26755
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.187299939441128e-06,
+      "loss": 0.4266,
+      "step": 26756
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.186723188280417e-06,
+      "loss": 0.2803,
+      "step": 26757
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.186146437119706e-06,
+      "loss": 0.368,
+      "step": 26758
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.185569685958993e-06,
+      "loss": 0.4556,
+      "step": 26759
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.184992934798282e-06,
+      "loss": 0.5054,
+      "step": 26760
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.184416183637569e-06,
+      "loss": 0.4843,
+      "step": 26761
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1838394324768585e-06,
+      "loss": 0.3954,
+      "step": 26762
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.183262681316147e-06,
+      "loss": 0.4415,
+      "step": 26763
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.182685930155435e-06,
+      "loss": 0.4534,
+      "step": 26764
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.182109178994723e-06,
+      "loss": 0.4428,
+      "step": 26765
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.181532427834012e-06,
+      "loss": 0.5019,
+      "step": 26766
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1809556766732995e-06,
+      "loss": 0.5007,
+      "step": 26767
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.180378925512588e-06,
+      "loss": 0.3603,
+      "step": 26768
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1798021743518765e-06,
+      "loss": 0.3148,
+      "step": 26769
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.179225423191164e-06,
+      "loss": 0.4245,
+      "step": 26770
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.178648672030453e-06,
+      "loss": 0.4014,
+      "step": 26771
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.178071920869741e-06,
+      "loss": 0.3843,
+      "step": 26772
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.177495169709029e-06,
+      "loss": 0.5164,
+      "step": 26773
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1769184185483175e-06,
+      "loss": 0.5116,
+      "step": 26774
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.176341667387607e-06,
+      "loss": 0.35,
+      "step": 26775
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.175764916226894e-06,
+      "loss": 0.3504,
+      "step": 26776
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.175188165066183e-06,
+      "loss": 0.3813,
+      "step": 26777
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.174611413905471e-06,
+      "loss": 0.4826,
+      "step": 26778
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.174034662744759e-06,
+      "loss": 0.3238,
+      "step": 26779
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.173457911584048e-06,
+      "loss": 0.5568,
+      "step": 26780
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1728811604233355e-06,
+      "loss": 0.5592,
+      "step": 26781
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.172304409262624e-06,
+      "loss": 0.4151,
+      "step": 26782
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.171727658101913e-06,
+      "loss": 0.3611,
+      "step": 26783
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1711509069412e-06,
+      "loss": 0.3882,
+      "step": 26784
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.170574155780489e-06,
+      "loss": 0.42,
+      "step": 26785
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.169997404619777e-06,
+      "loss": 0.5243,
+      "step": 26786
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.169420653459065e-06,
+      "loss": 0.428,
+      "step": 26787
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1688439022983536e-06,
+      "loss": 0.5319,
+      "step": 26788
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.168267151137643e-06,
+      "loss": 0.2598,
+      "step": 26789
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.16769039997693e-06,
+      "loss": 0.3945,
+      "step": 26790
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.167113648816219e-06,
+      "loss": 0.4039,
+      "step": 26791
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.166536897655508e-06,
+      "loss": 0.3935,
+      "step": 26792
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.165960146494795e-06,
+      "loss": 0.5386,
+      "step": 26793
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.165383395334084e-06,
+      "loss": 0.4229,
+      "step": 26794
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.164806644173372e-06,
+      "loss": 0.401,
+      "step": 26795
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.16422989301266e-06,
+      "loss": 0.3449,
+      "step": 26796
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.163653141851949e-06,
+      "loss": 0.4111,
+      "step": 26797
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.163076390691236e-06,
+      "loss": 0.4303,
+      "step": 26798
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.162499639530525e-06,
+      "loss": 0.3802,
+      "step": 26799
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.161922888369813e-06,
+      "loss": 0.4227,
+      "step": 26800
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.161346137209101e-06,
+      "loss": 0.3882,
+      "step": 26801
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.16076938604839e-06,
+      "loss": 0.3063,
+      "step": 26802
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.160192634887678e-06,
+      "loss": 0.2191,
+      "step": 26803
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.159615883726966e-06,
+      "loss": 0.374,
+      "step": 26804
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.159039132566254e-06,
+      "loss": 0.4301,
+      "step": 26805
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.158462381405544e-06,
+      "loss": 0.3061,
+      "step": 26806
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1578856302448314e-06,
+      "loss": 0.4728,
+      "step": 26807
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.15730887908412e-06,
+      "loss": 0.2679,
+      "step": 26808
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1567321279234085e-06,
+      "loss": 0.3921,
+      "step": 26809
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.156155376762696e-06,
+      "loss": 0.3761,
+      "step": 26810
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.155578625601985e-06,
+      "loss": 0.3588,
+      "step": 26811
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.155001874441272e-06,
+      "loss": 0.3326,
+      "step": 26812
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.154425123280561e-06,
+      "loss": 0.2965,
+      "step": 26813
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1538483721198495e-06,
+      "loss": 0.3394,
+      "step": 26814
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.153271620959137e-06,
+      "loss": 0.489,
+      "step": 26815
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.152694869798426e-06,
+      "loss": 0.3389,
+      "step": 26816
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.152118118637714e-06,
+      "loss": 0.3329,
+      "step": 26817
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.151541367477002e-06,
+      "loss": 0.3776,
+      "step": 26818
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1509646163162904e-06,
+      "loss": 0.3758,
+      "step": 26819
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.15038786515558e-06,
+      "loss": 0.4089,
+      "step": 26820
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.149811113994867e-06,
+      "loss": 0.5103,
+      "step": 26821
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.149234362834156e-06,
+      "loss": 0.296,
+      "step": 26822
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1486576116734446e-06,
+      "loss": 0.4678,
+      "step": 26823
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.148080860512732e-06,
+      "loss": 0.275,
+      "step": 26824
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.147504109352021e-06,
+      "loss": 0.3915,
+      "step": 26825
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.146927358191309e-06,
+      "loss": 0.3749,
+      "step": 26826
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.146350607030597e-06,
+      "loss": 0.467,
+      "step": 26827
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1457738558698855e-06,
+      "loss": 0.4805,
+      "step": 26828
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.145197104709173e-06,
+      "loss": 0.3718,
+      "step": 26829
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.144620353548462e-06,
+      "loss": 0.4095,
+      "step": 26830
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.14404360238775e-06,
+      "loss": 0.3238,
+      "step": 26831
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.143466851227038e-06,
+      "loss": 0.3333,
+      "step": 26832
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1428901000663265e-06,
+      "loss": 0.5325,
+      "step": 26833
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.142313348905615e-06,
+      "loss": 0.4674,
+      "step": 26834
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.141736597744903e-06,
+      "loss": 0.4387,
+      "step": 26835
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.141159846584191e-06,
+      "loss": 0.4285,
+      "step": 26836
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.140583095423481e-06,
+      "loss": 0.5453,
+      "step": 26837
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.140006344262768e-06,
+      "loss": 0.5098,
+      "step": 26838
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.139429593102057e-06,
+      "loss": 0.4073,
+      "step": 26839
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.138852841941345e-06,
+      "loss": 0.3496,
+      "step": 26840
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.138276090780633e-06,
+      "loss": 0.3184,
+      "step": 26841
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.137699339619922e-06,
+      "loss": 0.3469,
+      "step": 26842
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.13712258845921e-06,
+      "loss": 0.3031,
+      "step": 26843
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.136545837298498e-06,
+      "loss": 0.3958,
+      "step": 26844
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.135969086137786e-06,
+      "loss": 0.3626,
+      "step": 26845
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.135392334977074e-06,
+      "loss": 0.3543,
+      "step": 26846
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1348155838163626e-06,
+      "loss": 0.5187,
+      "step": 26847
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.134238832655651e-06,
+      "loss": 0.4504,
+      "step": 26848
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.133662081494939e-06,
+      "loss": 0.4132,
+      "step": 26849
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.133085330334227e-06,
+      "loss": 0.5217,
+      "step": 26850
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.132508579173517e-06,
+      "loss": 0.3647,
+      "step": 26851
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1319318280128035e-06,
+      "loss": 0.4267,
+      "step": 26852
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.131355076852093e-06,
+      "loss": 0.4447,
+      "step": 26853
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1307783256913814e-06,
+      "loss": 0.3971,
+      "step": 26854
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.130201574530669e-06,
+      "loss": 0.3849,
+      "step": 26855
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.129624823369958e-06,
+      "loss": 0.2562,
+      "step": 26856
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.129048072209246e-06,
+      "loss": 0.5026,
+      "step": 26857
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.128471321048534e-06,
+      "loss": 0.4777,
+      "step": 26858
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.127894569887822e-06,
+      "loss": 0.4885,
+      "step": 26859
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.127317818727111e-06,
+      "loss": 0.4665,
+      "step": 26860
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.126741067566399e-06,
+      "loss": 0.4437,
+      "step": 26861
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.126164316405687e-06,
+      "loss": 0.4481,
+      "step": 26862
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.125587565244976e-06,
+      "loss": 0.3465,
+      "step": 26863
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.125010814084263e-06,
+      "loss": 0.2911,
+      "step": 26864
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.124434062923552e-06,
+      "loss": 0.5431,
+      "step": 26865
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.12385731176284e-06,
+      "loss": 0.4067,
+      "step": 26866
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.123280560602128e-06,
+      "loss": 0.4037,
+      "step": 26867
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1227038094414175e-06,
+      "loss": 0.4105,
+      "step": 26868
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.122127058280705e-06,
+      "loss": 0.4382,
+      "step": 26869
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.121550307119994e-06,
+      "loss": 0.3758,
+      "step": 26870
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.120973555959282e-06,
+      "loss": 0.3293,
+      "step": 26871
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.12039680479857e-06,
+      "loss": 0.3542,
+      "step": 26872
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1198200536378585e-06,
+      "loss": 0.5099,
+      "step": 26873
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.119243302477147e-06,
+      "loss": 0.3921,
+      "step": 26874
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.118666551316435e-06,
+      "loss": 0.5617,
+      "step": 26875
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.118089800155723e-06,
+      "loss": 0.5269,
+      "step": 26876
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.117513048995012e-06,
+      "loss": 0.257,
+      "step": 26877
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1169362978342994e-06,
+      "loss": 0.4696,
+      "step": 26878
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.116359546673588e-06,
+      "loss": 0.3701,
+      "step": 26879
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.115782795512877e-06,
+      "loss": 0.4988,
+      "step": 26880
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.115206044352164e-06,
+      "loss": 0.454,
+      "step": 26881
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1146292931914536e-06,
+      "loss": 0.3801,
+      "step": 26882
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.11405254203074e-06,
+      "loss": 0.4245,
+      "step": 26883
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.11347579087003e-06,
+      "loss": 0.3229,
+      "step": 26884
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.112899039709318e-06,
+      "loss": 0.4276,
+      "step": 26885
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.112322288548606e-06,
+      "loss": 0.5085,
+      "step": 26886
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1117455373878945e-06,
+      "loss": 0.5033,
+      "step": 26887
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.111168786227183e-06,
+      "loss": 0.5196,
+      "step": 26888
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.110592035066471e-06,
+      "loss": 0.3594,
+      "step": 26889
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.110015283905759e-06,
+      "loss": 0.4271,
+      "step": 26890
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.109438532745048e-06,
+      "loss": 0.4649,
+      "step": 26891
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1088617815843355e-06,
+      "loss": 0.4066,
+      "step": 26892
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.108285030423624e-06,
+      "loss": 0.3323,
+      "step": 26893
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1077082792629126e-06,
+      "loss": 0.4216,
+      "step": 26894
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.1071315281022e-06,
+      "loss": 0.449,
+      "step": 26895
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.106554776941489e-06,
+      "loss": 0.4674,
+      "step": 26896
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.105978025780778e-06,
+      "loss": 0.3603,
+      "step": 26897
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.105401274620065e-06,
+      "loss": 0.3953,
+      "step": 26898
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.104824523459354e-06,
+      "loss": 0.4165,
+      "step": 26899
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.104247772298642e-06,
+      "loss": 0.4036,
+      "step": 26900
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 5.103671021137931e-06,
+      "loss": 0.4197,
+      "step": 26901
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.103094269977219e-06,
+      "loss": 0.4288,
+      "step": 26902
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.102517518816507e-06,
+      "loss": 0.5483,
+      "step": 26903
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.101940767655795e-06,
+      "loss": 0.417,
+      "step": 26904
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.101364016495084e-06,
+      "loss": 0.377,
+      "step": 26905
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.1007872653343716e-06,
+      "loss": 0.3947,
+      "step": 26906
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.10021051417366e-06,
+      "loss": 0.3721,
+      "step": 26907
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.099633763012949e-06,
+      "loss": 0.3201,
+      "step": 26908
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.099057011852236e-06,
+      "loss": 0.4246,
+      "step": 26909
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.098480260691525e-06,
+      "loss": 0.4577,
+      "step": 26910
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.097903509530814e-06,
+      "loss": 0.3518,
+      "step": 26911
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.097326758370101e-06,
+      "loss": 0.435,
+      "step": 26912
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0967500072093904e-06,
+      "loss": 0.3763,
+      "step": 26913
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.096173256048679e-06,
+      "loss": 0.3905,
+      "step": 26914
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.095596504887967e-06,
+      "loss": 0.4494,
+      "step": 26915
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.095019753727255e-06,
+      "loss": 0.3747,
+      "step": 26916
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.094443002566543e-06,
+      "loss": 0.51,
+      "step": 26917
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.093866251405831e-06,
+      "loss": 0.3505,
+      "step": 26918
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.09328950024512e-06,
+      "loss": 0.3892,
+      "step": 26919
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.092712749084408e-06,
+      "loss": 0.4453,
+      "step": 26920
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.092135997923696e-06,
+      "loss": 0.3784,
+      "step": 26921
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.091559246762985e-06,
+      "loss": 0.3498,
+      "step": 26922
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.090982495602272e-06,
+      "loss": 0.4546,
+      "step": 26923
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.090405744441561e-06,
+      "loss": 0.4652,
+      "step": 26924
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0898289932808494e-06,
+      "loss": 0.4829,
+      "step": 26925
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.089252242120137e-06,
+      "loss": 0.4133,
+      "step": 26926
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.088675490959426e-06,
+      "loss": 0.3943,
+      "step": 26927
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.088098739798715e-06,
+      "loss": 0.3383,
+      "step": 26928
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.087521988638002e-06,
+      "loss": 0.4733,
+      "step": 26929
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.086945237477291e-06,
+      "loss": 0.4394,
+      "step": 26930
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.08636848631658e-06,
+      "loss": 0.3836,
+      "step": 26931
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0857917351558675e-06,
+      "loss": 0.3521,
+      "step": 26932
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.085214983995156e-06,
+      "loss": 0.3369,
+      "step": 26933
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.084638232834444e-06,
+      "loss": 0.5596,
+      "step": 26934
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.084061481673732e-06,
+      "loss": 0.4505,
+      "step": 26935
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.083484730513021e-06,
+      "loss": 0.3931,
+      "step": 26936
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0829079793523084e-06,
+      "loss": 0.3862,
+      "step": 26937
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.082331228191597e-06,
+      "loss": 0.4617,
+      "step": 26938
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0817544770308855e-06,
+      "loss": 0.5524,
+      "step": 26939
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.081177725870173e-06,
+      "loss": 0.3695,
+      "step": 26940
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.080600974709462e-06,
+      "loss": 0.3891,
+      "step": 26941
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.080024223548751e-06,
+      "loss": 0.3523,
+      "step": 26942
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.079447472388038e-06,
+      "loss": 0.5246,
+      "step": 26943
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.078870721227327e-06,
+      "loss": 0.433,
+      "step": 26944
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.078293970066616e-06,
+      "loss": 0.3728,
+      "step": 26945
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0777172189059035e-06,
+      "loss": 0.4854,
+      "step": 26946
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.077140467745192e-06,
+      "loss": 0.4461,
+      "step": 26947
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.076563716584481e-06,
+      "loss": 0.5514,
+      "step": 26948
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.075986965423768e-06,
+      "loss": 0.412,
+      "step": 26949
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.075410214263057e-06,
+      "loss": 0.4144,
+      "step": 26950
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0748334631023445e-06,
+      "loss": 0.4344,
+      "step": 26951
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.074256711941633e-06,
+      "loss": 0.4638,
+      "step": 26952
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0736799607809216e-06,
+      "loss": 0.4981,
+      "step": 26953
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.073103209620209e-06,
+      "loss": 0.4639,
+      "step": 26954
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.072526458459498e-06,
+      "loss": 0.4713,
+      "step": 26955
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.071949707298786e-06,
+      "loss": 0.343,
+      "step": 26956
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.071372956138074e-06,
+      "loss": 0.5215,
+      "step": 26957
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0707962049773625e-06,
+      "loss": 0.3833,
+      "step": 26958
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.070219453816652e-06,
+      "loss": 0.4562,
+      "step": 26959
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.069642702655939e-06,
+      "loss": 0.3674,
+      "step": 26960
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.069065951495228e-06,
+      "loss": 0.4631,
+      "step": 26961
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.068489200334517e-06,
+      "loss": 0.3626,
+      "step": 26962
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.067912449173804e-06,
+      "loss": 0.3052,
+      "step": 26963
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.067335698013093e-06,
+      "loss": 0.5219,
+      "step": 26964
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.066758946852381e-06,
+      "loss": 0.3623,
+      "step": 26965
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.066182195691669e-06,
+      "loss": 0.376,
+      "step": 26966
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.065605444530958e-06,
+      "loss": 0.3957,
+      "step": 26967
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.065028693370245e-06,
+      "loss": 0.4389,
+      "step": 26968
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.064451942209534e-06,
+      "loss": 0.4047,
+      "step": 26969
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.063875191048822e-06,
+      "loss": 0.4081,
+      "step": 26970
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.06329843988811e-06,
+      "loss": 0.4613,
+      "step": 26971
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.062721688727399e-06,
+      "loss": 0.4315,
+      "step": 26972
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.062144937566688e-06,
+      "loss": 0.3982,
+      "step": 26973
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.061568186405975e-06,
+      "loss": 0.4589,
+      "step": 26974
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.060991435245264e-06,
+      "loss": 0.2884,
+      "step": 26975
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.060414684084553e-06,
+      "loss": 0.424,
+      "step": 26976
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.05983793292384e-06,
+      "loss": 0.3766,
+      "step": 26977
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.059261181763129e-06,
+      "loss": 0.3371,
+      "step": 26978
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0586844306024175e-06,
+      "loss": 0.3981,
+      "step": 26979
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.058107679441705e-06,
+      "loss": 0.3027,
+      "step": 26980
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.057530928280994e-06,
+      "loss": 0.3956,
+      "step": 26981
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.056954177120282e-06,
+      "loss": 0.3449,
+      "step": 26982
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.05637742595957e-06,
+      "loss": 0.3601,
+      "step": 26983
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0558006747988584e-06,
+      "loss": 0.5482,
+      "step": 26984
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.055223923638147e-06,
+      "loss": 0.3848,
+      "step": 26985
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.054647172477435e-06,
+      "loss": 0.4505,
+      "step": 26986
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.054070421316723e-06,
+      "loss": 0.3659,
+      "step": 26987
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.053493670156011e-06,
+      "loss": 0.3926,
+      "step": 26988
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.052916918995299e-06,
+      "loss": 0.5425,
+      "step": 26989
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.052340167834589e-06,
+      "loss": 0.5504,
+      "step": 26990
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0517634166738765e-06,
+      "loss": 0.4315,
+      "step": 26991
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.051186665513165e-06,
+      "loss": 0.4525,
+      "step": 26992
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0506099143524535e-06,
+      "loss": 0.4032,
+      "step": 26993
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.050033163191741e-06,
+      "loss": 0.5182,
+      "step": 26994
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.04945641203103e-06,
+      "loss": 0.4614,
+      "step": 26995
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.048879660870318e-06,
+      "loss": 0.4347,
+      "step": 26996
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.048302909709606e-06,
+      "loss": 0.4464,
+      "step": 26997
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0477261585488945e-06,
+      "loss": 0.4158,
+      "step": 26998
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.047149407388183e-06,
+      "loss": 0.4414,
+      "step": 26999
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.046572656227471e-06,
+      "loss": 0.3576,
+      "step": 27000
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.045995905066759e-06,
+      "loss": 0.3948,
+      "step": 27001
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.045419153906048e-06,
+      "loss": 0.4862,
+      "step": 27002
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0448424027453355e-06,
+      "loss": 0.378,
+      "step": 27003
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.044265651584625e-06,
+      "loss": 0.3446,
+      "step": 27004
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.043688900423912e-06,
+      "loss": 0.3212,
+      "step": 27005
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.043112149263201e-06,
+      "loss": 0.3725,
+      "step": 27006
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.04253539810249e-06,
+      "loss": 0.3667,
+      "step": 27007
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.041958646941777e-06,
+      "loss": 0.2925,
+      "step": 27008
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.041381895781066e-06,
+      "loss": 0.4212,
+      "step": 27009
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.040805144620354e-06,
+      "loss": 0.4804,
+      "step": 27010
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.040228393459642e-06,
+      "loss": 0.3102,
+      "step": 27011
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0396516422989306e-06,
+      "loss": 0.3808,
+      "step": 27012
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.039074891138219e-06,
+      "loss": 0.4177,
+      "step": 27013
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.038498139977507e-06,
+      "loss": 0.4328,
+      "step": 27014
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.037921388816795e-06,
+      "loss": 0.4222,
+      "step": 27015
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.037344637656084e-06,
+      "loss": 0.4286,
+      "step": 27016
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0367678864953715e-06,
+      "loss": 0.3506,
+      "step": 27017
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.03619113533466e-06,
+      "loss": 0.441,
+      "step": 27018
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0356143841739494e-06,
+      "loss": 0.54,
+      "step": 27019
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.035037633013236e-06,
+      "loss": 0.4347,
+      "step": 27020
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.034460881852526e-06,
+      "loss": 0.3922,
+      "step": 27021
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.033884130691813e-06,
+      "loss": 0.3447,
+      "step": 27022
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.033307379531102e-06,
+      "loss": 0.3138,
+      "step": 27023
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.03273062837039e-06,
+      "loss": 0.3875,
+      "step": 27024
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.032153877209678e-06,
+      "loss": 0.457,
+      "step": 27025
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.031577126048967e-06,
+      "loss": 0.413,
+      "step": 27026
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.031000374888255e-06,
+      "loss": 0.4181,
+      "step": 27027
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.030423623727543e-06,
+      "loss": 0.5881,
+      "step": 27028
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.029846872566831e-06,
+      "loss": 0.4374,
+      "step": 27029
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.02927012140612e-06,
+      "loss": 0.4192,
+      "step": 27030
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.028693370245408e-06,
+      "loss": 0.3553,
+      "step": 27031
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.028116619084696e-06,
+      "loss": 0.4287,
+      "step": 27032
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0275398679239855e-06,
+      "loss": 0.52,
+      "step": 27033
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.026963116763272e-06,
+      "loss": 0.3756,
+      "step": 27034
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.026386365602562e-06,
+      "loss": 0.3738,
+      "step": 27035
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.02580961444185e-06,
+      "loss": 0.4306,
+      "step": 27036
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.025232863281138e-06,
+      "loss": 0.3187,
+      "step": 27037
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0246561121204265e-06,
+      "loss": 0.426,
+      "step": 27038
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.024079360959714e-06,
+      "loss": 0.3091,
+      "step": 27039
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.023502609799003e-06,
+      "loss": 0.4252,
+      "step": 27040
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.022925858638291e-06,
+      "loss": 0.4324,
+      "step": 27041
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.022349107477579e-06,
+      "loss": 0.402,
+      "step": 27042
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0217723563168674e-06,
+      "loss": 0.3136,
+      "step": 27043
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.021195605156156e-06,
+      "loss": 0.4898,
+      "step": 27044
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.020618853995444e-06,
+      "loss": 0.374,
+      "step": 27045
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.020042102834732e-06,
+      "loss": 0.4374,
+      "step": 27046
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.019465351674021e-06,
+      "loss": 0.4307,
+      "step": 27047
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.018888600513308e-06,
+      "loss": 0.5189,
+      "step": 27048
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.018311849352597e-06,
+      "loss": 0.3143,
+      "step": 27049
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.017735098191886e-06,
+      "loss": 0.4006,
+      "step": 27050
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.017158347031173e-06,
+      "loss": 0.5608,
+      "step": 27051
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0165815958704625e-06,
+      "loss": 0.4289,
+      "step": 27052
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.016004844709751e-06,
+      "loss": 0.3987,
+      "step": 27053
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.015428093549039e-06,
+      "loss": 0.5081,
+      "step": 27054
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.014851342388327e-06,
+      "loss": 0.441,
+      "step": 27055
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.014274591227615e-06,
+      "loss": 0.4007,
+      "step": 27056
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0136978400669035e-06,
+      "loss": 0.412,
+      "step": 27057
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.013121088906192e-06,
+      "loss": 0.5494,
+      "step": 27058
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.01254433774548e-06,
+      "loss": 0.377,
+      "step": 27059
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.011967586584768e-06,
+      "loss": 0.3343,
+      "step": 27060
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.011390835424057e-06,
+      "loss": 0.3934,
+      "step": 27061
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0108140842633445e-06,
+      "loss": 0.4112,
+      "step": 27062
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.010237333102633e-06,
+      "loss": 0.436,
+      "step": 27063
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.009660581941922e-06,
+      "loss": 0.4746,
+      "step": 27064
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.009083830781209e-06,
+      "loss": 0.481,
+      "step": 27065
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.008507079620499e-06,
+      "loss": 0.4686,
+      "step": 27066
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.007930328459787e-06,
+      "loss": 0.3494,
+      "step": 27067
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.007353577299075e-06,
+      "loss": 0.3777,
+      "step": 27068
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.006776826138363e-06,
+      "loss": 0.4486,
+      "step": 27069
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.006200074977652e-06,
+      "loss": 0.4418,
+      "step": 27070
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0056233238169396e-06,
+      "loss": 0.3822,
+      "step": 27071
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.005046572656228e-06,
+      "loss": 0.4085,
+      "step": 27072
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.004469821495516e-06,
+      "loss": 0.3189,
+      "step": 27073
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.003893070334804e-06,
+      "loss": 0.4831,
+      "step": 27074
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.003316319174093e-06,
+      "loss": 0.4578,
+      "step": 27075
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.0027395680133805e-06,
+      "loss": 0.4616,
+      "step": 27076
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.002162816852669e-06,
+      "loss": 0.3525,
+      "step": 27077
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.001586065691958e-06,
+      "loss": 0.4861,
+      "step": 27078
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.001009314531245e-06,
+      "loss": 0.4063,
+      "step": 27079
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 5.000432563370534e-06,
+      "loss": 0.4969,
+      "step": 27080
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.999855812209822e-06,
+      "loss": 0.5012,
+      "step": 27081
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.999279061049111e-06,
+      "loss": 0.3821,
+      "step": 27082
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.998702309888399e-06,
+      "loss": 0.3886,
+      "step": 27083
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.998125558727687e-06,
+      "loss": 0.5574,
+      "step": 27084
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.997548807566976e-06,
+      "loss": 0.4137,
+      "step": 27085
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.996972056406264e-06,
+      "loss": 0.4733,
+      "step": 27086
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.996395305245552e-06,
+      "loss": 0.4731,
+      "step": 27087
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.99581855408484e-06,
+      "loss": 0.4245,
+      "step": 27088
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.995241802924129e-06,
+      "loss": 0.4297,
+      "step": 27089
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9946650517634174e-06,
+      "loss": 0.4349,
+      "step": 27090
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.994088300602705e-06,
+      "loss": 0.4434,
+      "step": 27091
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.993511549441994e-06,
+      "loss": 0.4451,
+      "step": 27092
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.992934798281282e-06,
+      "loss": 0.38,
+      "step": 27093
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.99235804712057e-06,
+      "loss": 0.4267,
+      "step": 27094
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.991781295959858e-06,
+      "loss": 0.4937,
+      "step": 27095
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.991204544799147e-06,
+      "loss": 0.3939,
+      "step": 27096
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9906277936384355e-06,
+      "loss": 0.4208,
+      "step": 27097
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.990051042477723e-06,
+      "loss": 0.458,
+      "step": 27098
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.989474291317012e-06,
+      "loss": 0.5472,
+      "step": 27099
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9888975401563e-06,
+      "loss": 0.3506,
+      "step": 27100
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.988320788995588e-06,
+      "loss": 0.3957,
+      "step": 27101
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.987744037834876e-06,
+      "loss": 0.5754,
+      "step": 27102
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.987167286674165e-06,
+      "loss": 0.4033,
+      "step": 27103
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9865905355134535e-06,
+      "loss": 0.4657,
+      "step": 27104
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.986013784352741e-06,
+      "loss": 0.5112,
+      "step": 27105
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.98543703319203e-06,
+      "loss": 0.4165,
+      "step": 27106
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.984860282031318e-06,
+      "loss": 0.4501,
+      "step": 27107
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.984283530870606e-06,
+      "loss": 0.4189,
+      "step": 27108
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9837067797098945e-06,
+      "loss": 0.3439,
+      "step": 27109
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.983130028549183e-06,
+      "loss": 0.4499,
+      "step": 27110
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.982553277388471e-06,
+      "loss": 0.3709,
+      "step": 27111
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.981976526227759e-06,
+      "loss": 0.3887,
+      "step": 27112
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.981399775067048e-06,
+      "loss": 0.3255,
+      "step": 27113
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.980823023906336e-06,
+      "loss": 0.4419,
+      "step": 27114
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.980246272745624e-06,
+      "loss": 0.4049,
+      "step": 27115
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9796695215849125e-06,
+      "loss": 0.38,
+      "step": 27116
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.979092770424201e-06,
+      "loss": 0.3931,
+      "step": 27117
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.978516019263489e-06,
+      "loss": 0.4766,
+      "step": 27118
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.977939268102778e-06,
+      "loss": 0.4482,
+      "step": 27119
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.977362516942066e-06,
+      "loss": 0.4322,
+      "step": 27120
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.976785765781354e-06,
+      "loss": 0.469,
+      "step": 27121
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.976209014620642e-06,
+      "loss": 0.4642,
+      "step": 27122
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9756322634599305e-06,
+      "loss": 0.4795,
+      "step": 27123
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.975055512299219e-06,
+      "loss": 0.3801,
+      "step": 27124
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.974478761138507e-06,
+      "loss": 0.4283,
+      "step": 27125
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.973902009977796e-06,
+      "loss": 0.4478,
+      "step": 27126
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.973325258817084e-06,
+      "loss": 0.406,
+      "step": 27127
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.972748507656372e-06,
+      "loss": 0.3734,
+      "step": 27128
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.97217175649566e-06,
+      "loss": 0.3166,
+      "step": 27129
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9715950053349485e-06,
+      "loss": 0.4025,
+      "step": 27130
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.971018254174237e-06,
+      "loss": 0.4494,
+      "step": 27131
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.970441503013525e-06,
+      "loss": 0.4921,
+      "step": 27132
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.969864751852813e-06,
+      "loss": 0.4113,
+      "step": 27133
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.969288000692102e-06,
+      "loss": 0.2979,
+      "step": 27134
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.96871124953139e-06,
+      "loss": 0.4556,
+      "step": 27135
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.968134498370679e-06,
+      "loss": 0.3445,
+      "step": 27136
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.967557747209967e-06,
+      "loss": 0.4042,
+      "step": 27137
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.966980996049255e-06,
+      "loss": 0.4271,
+      "step": 27138
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.966404244888543e-06,
+      "loss": 0.312,
+      "step": 27139
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.965827493727831e-06,
+      "loss": 0.4376,
+      "step": 27140
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.96525074256712e-06,
+      "loss": 0.3858,
+      "step": 27141
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9646739914064075e-06,
+      "loss": 0.4223,
+      "step": 27142
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.964097240245697e-06,
+      "loss": 0.3822,
+      "step": 27143
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.963520489084985e-06,
+      "loss": 0.3708,
+      "step": 27144
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.962943737924273e-06,
+      "loss": 0.3712,
+      "step": 27145
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.962366986763562e-06,
+      "loss": 0.322,
+      "step": 27146
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.961790235602849e-06,
+      "loss": 0.5119,
+      "step": 27147
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.961213484442138e-06,
+      "loss": 0.5381,
+      "step": 27148
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9606367332814256e-06,
+      "loss": 0.4579,
+      "step": 27149
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.960059982120715e-06,
+      "loss": 0.4748,
+      "step": 27150
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.959483230960003e-06,
+      "loss": 0.4758,
+      "step": 27151
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.958906479799291e-06,
+      "loss": 0.3542,
+      "step": 27152
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.95832972863858e-06,
+      "loss": 0.3972,
+      "step": 27153
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.957752977477867e-06,
+      "loss": 0.504,
+      "step": 27154
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.957176226317156e-06,
+      "loss": 0.3593,
+      "step": 27155
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.956599475156444e-06,
+      "loss": 0.3468,
+      "step": 27156
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.956022723995733e-06,
+      "loss": 0.4362,
+      "step": 27157
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.955445972835021e-06,
+      "loss": 0.3672,
+      "step": 27158
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.954869221674309e-06,
+      "loss": 0.4328,
+      "step": 27159
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.954292470513598e-06,
+      "loss": 0.3789,
+      "step": 27160
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.953715719352885e-06,
+      "loss": 0.3515,
+      "step": 27161
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.953138968192174e-06,
+      "loss": 0.4407,
+      "step": 27162
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9525622170314625e-06,
+      "loss": 0.3767,
+      "step": 27163
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.95198546587075e-06,
+      "loss": 0.4724,
+      "step": 27164
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.951408714710039e-06,
+      "loss": 0.4054,
+      "step": 27165
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.950831963549327e-06,
+      "loss": 0.4305,
+      "step": 27166
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.950255212388616e-06,
+      "loss": 0.3946,
+      "step": 27167
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9496784612279035e-06,
+      "loss": 0.4886,
+      "step": 27168
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.949101710067192e-06,
+      "loss": 0.38,
+      "step": 27169
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9485249589064805e-06,
+      "loss": 0.4002,
+      "step": 27170
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.947948207745768e-06,
+      "loss": 0.3946,
+      "step": 27171
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.947371456585057e-06,
+      "loss": 0.4679,
+      "step": 27172
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.946794705424344e-06,
+      "loss": 0.3559,
+      "step": 27173
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.946217954263634e-06,
+      "loss": 0.4612,
+      "step": 27174
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9456412031029215e-06,
+      "loss": 0.3856,
+      "step": 27175
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.94506445194221e-06,
+      "loss": 0.421,
+      "step": 27176
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9444877007814985e-06,
+      "loss": 0.38,
+      "step": 27177
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.943910949620786e-06,
+      "loss": 0.4312,
+      "step": 27178
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.943334198460075e-06,
+      "loss": 0.3505,
+      "step": 27179
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.942757447299363e-06,
+      "loss": 0.3571,
+      "step": 27180
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.942180696138652e-06,
+      "loss": 0.503,
+      "step": 27181
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9416039449779395e-06,
+      "loss": 0.3714,
+      "step": 27182
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.941027193817228e-06,
+      "loss": 0.4207,
+      "step": 27183
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9404504426565166e-06,
+      "loss": 0.4483,
+      "step": 27184
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.939873691495804e-06,
+      "loss": 0.3495,
+      "step": 27185
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.939296940335093e-06,
+      "loss": 0.4918,
+      "step": 27186
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.938720189174381e-06,
+      "loss": 0.3606,
+      "step": 27187
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.93814343801367e-06,
+      "loss": 0.3666,
+      "step": 27188
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9375666868529575e-06,
+      "loss": 0.4431,
+      "step": 27189
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.936989935692246e-06,
+      "loss": 0.3408,
+      "step": 27190
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.936413184531535e-06,
+      "loss": 0.4887,
+      "step": 27191
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.935836433370822e-06,
+      "loss": 0.4198,
+      "step": 27192
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.935259682210111e-06,
+      "loss": 0.4053,
+      "step": 27193
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.934682931049399e-06,
+      "loss": 0.5418,
+      "step": 27194
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.934106179888687e-06,
+      "loss": 0.4712,
+      "step": 27195
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9335294287279756e-06,
+      "loss": 0.4305,
+      "step": 27196
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.932952677567264e-06,
+      "loss": 0.3315,
+      "step": 27197
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.932375926406553e-06,
+      "loss": 0.5261,
+      "step": 27198
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.93179917524584e-06,
+      "loss": 0.4032,
+      "step": 27199
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.931222424085129e-06,
+      "loss": 0.5209,
+      "step": 27200
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.930645672924417e-06,
+      "loss": 0.4073,
+      "step": 27201
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.930068921763705e-06,
+      "loss": 0.3716,
+      "step": 27202
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.929492170602994e-06,
+      "loss": 0.4623,
+      "step": 27203
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.928915419442282e-06,
+      "loss": 0.5041,
+      "step": 27204
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.928338668281571e-06,
+      "loss": 0.3787,
+      "step": 27205
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.927761917120858e-06,
+      "loss": 0.5617,
+      "step": 27206
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.927185165960147e-06,
+      "loss": 0.3659,
+      "step": 27207
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.926608414799435e-06,
+      "loss": 0.4157,
+      "step": 27208
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.926031663638723e-06,
+      "loss": 0.4017,
+      "step": 27209
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.925454912478012e-06,
+      "loss": 0.4708,
+      "step": 27210
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9248781613173e-06,
+      "loss": 0.5045,
+      "step": 27211
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.924301410156589e-06,
+      "loss": 0.383,
+      "step": 27212
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.923724658995876e-06,
+      "loss": 0.3684,
+      "step": 27213
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.923147907835165e-06,
+      "loss": 0.5167,
+      "step": 27214
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9225711566744534e-06,
+      "loss": 0.4891,
+      "step": 27215
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.921994405513741e-06,
+      "loss": 0.3857,
+      "step": 27216
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.92141765435303e-06,
+      "loss": 0.3815,
+      "step": 27217
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.920840903192318e-06,
+      "loss": 0.4109,
+      "step": 27218
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.920264152031607e-06,
+      "loss": 0.4135,
+      "step": 27219
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.919687400870894e-06,
+      "loss": 0.2952,
+      "step": 27220
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.919110649710183e-06,
+      "loss": 0.4108,
+      "step": 27221
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9185338985494715e-06,
+      "loss": 0.3917,
+      "step": 27222
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.917957147388759e-06,
+      "loss": 0.3824,
+      "step": 27223
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.917380396228048e-06,
+      "loss": 0.4283,
+      "step": 27224
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.916803645067336e-06,
+      "loss": 0.4085,
+      "step": 27225
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.916226893906624e-06,
+      "loss": 0.4263,
+      "step": 27226
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9156501427459124e-06,
+      "loss": 0.4713,
+      "step": 27227
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.915073391585201e-06,
+      "loss": 0.199,
+      "step": 27228
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9144966404244895e-06,
+      "loss": 0.3089,
+      "step": 27229
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.913919889263777e-06,
+      "loss": 0.4575,
+      "step": 27230
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.913343138103066e-06,
+      "loss": 0.3698,
+      "step": 27231
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.912766386942354e-06,
+      "loss": 0.453,
+      "step": 27232
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.912189635781642e-06,
+      "loss": 0.312,
+      "step": 27233
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9116128846209305e-06,
+      "loss": 0.3002,
+      "step": 27234
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.911036133460219e-06,
+      "loss": 0.4202,
+      "step": 27235
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9104593822995075e-06,
+      "loss": 0.4385,
+      "step": 27236
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.909882631138795e-06,
+      "loss": 0.3587,
+      "step": 27237
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.909305879978084e-06,
+      "loss": 0.3294,
+      "step": 27238
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.908729128817372e-06,
+      "loss": 0.399,
+      "step": 27239
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.90815237765666e-06,
+      "loss": 0.3972,
+      "step": 27240
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.907575626495949e-06,
+      "loss": 0.3346,
+      "step": 27241
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.906998875335237e-06,
+      "loss": 0.4593,
+      "step": 27242
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.9064221241745256e-06,
+      "loss": 0.3337,
+      "step": 27243
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.905845373013813e-06,
+      "loss": 0.3271,
+      "step": 27244
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.905268621853102e-06,
+      "loss": 0.3611,
+      "step": 27245
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.90469187069239e-06,
+      "loss": 0.415,
+      "step": 27246
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.904115119531678e-06,
+      "loss": 0.3833,
+      "step": 27247
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.903538368370967e-06,
+      "loss": 0.4186,
+      "step": 27248
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.902961617210255e-06,
+      "loss": 0.3375,
+      "step": 27249
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.902384866049544e-06,
+      "loss": 0.4127,
+      "step": 27250
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.901808114888831e-06,
+      "loss": 0.4849,
+      "step": 27251
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.90123136372812e-06,
+      "loss": 0.4074,
+      "step": 27252
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.900654612567408e-06,
+      "loss": 0.4038,
+      "step": 27253
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.900077861406696e-06,
+      "loss": 0.47,
+      "step": 27254
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.8995011102459846e-06,
+      "loss": 0.4769,
+      "step": 27255
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.898924359085273e-06,
+      "loss": 0.428,
+      "step": 27256
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.898347607924561e-06,
+      "loss": 0.4074,
+      "step": 27257
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.89777085676385e-06,
+      "loss": 0.4187,
+      "step": 27258
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 4.897194105603138e-06,
+      "loss": 0.402,
+      "step": 27259
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.896617354442426e-06,
+      "loss": 0.3986,
+      "step": 27260
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.896040603281714e-06,
+      "loss": 0.4597,
+      "step": 27261
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.895463852121003e-06,
+      "loss": 0.4119,
+      "step": 27262
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.894887100960291e-06,
+      "loss": 0.4945,
+      "step": 27263
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.894310349799579e-06,
+      "loss": 0.474,
+      "step": 27264
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.893733598638868e-06,
+      "loss": 0.3273,
+      "step": 27265
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.893156847478156e-06,
+      "loss": 0.2706,
+      "step": 27266
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.892580096317444e-06,
+      "loss": 0.5388,
+      "step": 27267
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.892003345156732e-06,
+      "loss": 0.4941,
+      "step": 27268
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.891426593996021e-06,
+      "loss": 0.5477,
+      "step": 27269
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.890849842835309e-06,
+      "loss": 0.3527,
+      "step": 27270
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.890273091674597e-06,
+      "loss": 0.4315,
+      "step": 27271
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.889696340513886e-06,
+      "loss": 0.4219,
+      "step": 27272
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.889119589353174e-06,
+      "loss": 0.4598,
+      "step": 27273
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8885428381924624e-06,
+      "loss": 0.335,
+      "step": 27274
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.887966087031751e-06,
+      "loss": 0.4074,
+      "step": 27275
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.887389335871039e-06,
+      "loss": 0.3652,
+      "step": 27276
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.886812584710327e-06,
+      "loss": 0.4498,
+      "step": 27277
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.886235833549615e-06,
+      "loss": 0.4515,
+      "step": 27278
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.885659082388904e-06,
+      "loss": 0.3507,
+      "step": 27279
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.885082331228192e-06,
+      "loss": 0.4356,
+      "step": 27280
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8845055800674805e-06,
+      "loss": 0.3978,
+      "step": 27281
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.883928828906769e-06,
+      "loss": 0.3815,
+      "step": 27282
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.883352077746057e-06,
+      "loss": 0.4135,
+      "step": 27283
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.882775326585345e-06,
+      "loss": 0.4488,
+      "step": 27284
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.882198575424634e-06,
+      "loss": 0.3044,
+      "step": 27285
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8816218242639214e-06,
+      "loss": 0.407,
+      "step": 27286
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.88104507310321e-06,
+      "loss": 0.3695,
+      "step": 27287
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8804683219424985e-06,
+      "loss": 0.4327,
+      "step": 27288
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.879891570781787e-06,
+      "loss": 0.4292,
+      "step": 27289
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.879314819621075e-06,
+      "loss": 0.478,
+      "step": 27290
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.878738068460363e-06,
+      "loss": 0.3493,
+      "step": 27291
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.878161317299652e-06,
+      "loss": 0.3197,
+      "step": 27292
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8775845661389395e-06,
+      "loss": 0.4483,
+      "step": 27293
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.877007814978228e-06,
+      "loss": 0.4325,
+      "step": 27294
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.876431063817516e-06,
+      "loss": 0.4055,
+      "step": 27295
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.875854312656805e-06,
+      "loss": 0.4203,
+      "step": 27296
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.875277561496093e-06,
+      "loss": 0.4356,
+      "step": 27297
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.874700810335381e-06,
+      "loss": 0.4662,
+      "step": 27298
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.87412405917467e-06,
+      "loss": 0.4332,
+      "step": 27299
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8735473080139575e-06,
+      "loss": 0.5014,
+      "step": 27300
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.872970556853246e-06,
+      "loss": 0.3669,
+      "step": 27301
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8723938056925346e-06,
+      "loss": 0.4558,
+      "step": 27302
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.871817054531823e-06,
+      "loss": 0.4233,
+      "step": 27303
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.871240303371111e-06,
+      "loss": 0.4008,
+      "step": 27304
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.870663552210399e-06,
+      "loss": 0.4011,
+      "step": 27305
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.870086801049688e-06,
+      "loss": 0.4471,
+      "step": 27306
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8695100498889755e-06,
+      "loss": 0.3835,
+      "step": 27307
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.868933298728264e-06,
+      "loss": 0.4145,
+      "step": 27308
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.868356547567553e-06,
+      "loss": 0.3624,
+      "step": 27309
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.867779796406841e-06,
+      "loss": 0.384,
+      "step": 27310
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.867203045246129e-06,
+      "loss": 0.4181,
+      "step": 27311
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.866626294085417e-06,
+      "loss": 0.4751,
+      "step": 27312
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.866049542924706e-06,
+      "loss": 0.4409,
+      "step": 27313
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8654727917639936e-06,
+      "loss": 0.4708,
+      "step": 27314
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.864896040603282e-06,
+      "loss": 0.4239,
+      "step": 27315
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.864319289442571e-06,
+      "loss": 0.3885,
+      "step": 27316
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.863742538281858e-06,
+      "loss": 0.3464,
+      "step": 27317
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.863165787121147e-06,
+      "loss": 0.4498,
+      "step": 27318
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.862589035960435e-06,
+      "loss": 0.4114,
+      "step": 27319
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.862012284799724e-06,
+      "loss": 0.401,
+      "step": 27320
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.861435533639012e-06,
+      "loss": 0.4122,
+      "step": 27321
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8608587824783e-06,
+      "loss": 0.4036,
+      "step": 27322
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.860282031317589e-06,
+      "loss": 0.4408,
+      "step": 27323
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.859705280156876e-06,
+      "loss": 0.41,
+      "step": 27324
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.859128528996165e-06,
+      "loss": 0.4425,
+      "step": 27325
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.858551777835453e-06,
+      "loss": 0.3489,
+      "step": 27326
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.857975026674742e-06,
+      "loss": 0.3976,
+      "step": 27327
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.85739827551403e-06,
+      "loss": 0.4627,
+      "step": 27328
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.856821524353318e-06,
+      "loss": 0.3532,
+      "step": 27329
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.856244773192607e-06,
+      "loss": 0.368,
+      "step": 27330
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.855668022031894e-06,
+      "loss": 0.3536,
+      "step": 27331
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.855091270871183e-06,
+      "loss": 0.4076,
+      "step": 27332
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8545145197104714e-06,
+      "loss": 0.4625,
+      "step": 27333
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.85393776854976e-06,
+      "loss": 0.4112,
+      "step": 27334
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.853361017389048e-06,
+      "loss": 0.5578,
+      "step": 27335
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.852784266228336e-06,
+      "loss": 0.286,
+      "step": 27336
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.852207515067625e-06,
+      "loss": 0.431,
+      "step": 27337
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.851630763906912e-06,
+      "loss": 0.4111,
+      "step": 27338
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.851054012746201e-06,
+      "loss": 0.3914,
+      "step": 27339
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8504772615854895e-06,
+      "loss": 0.462,
+      "step": 27340
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.849900510424778e-06,
+      "loss": 0.3365,
+      "step": 27341
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.849323759264066e-06,
+      "loss": 0.3553,
+      "step": 27342
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.848747008103354e-06,
+      "loss": 0.4234,
+      "step": 27343
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.848170256942643e-06,
+      "loss": 0.4363,
+      "step": 27344
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8475935057819304e-06,
+      "loss": 0.3742,
+      "step": 27345
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.847016754621219e-06,
+      "loss": 0.3894,
+      "step": 27346
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8464400034605075e-06,
+      "loss": 0.4532,
+      "step": 27347
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.845863252299795e-06,
+      "loss": 0.4594,
+      "step": 27348
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.845286501139084e-06,
+      "loss": 0.4665,
+      "step": 27349
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.844709749978372e-06,
+      "loss": 0.5188,
+      "step": 27350
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.844132998817661e-06,
+      "loss": 0.356,
+      "step": 27351
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8435562476569485e-06,
+      "loss": 0.4196,
+      "step": 27352
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.842979496496237e-06,
+      "loss": 0.3798,
+      "step": 27353
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8424027453355255e-06,
+      "loss": 0.461,
+      "step": 27354
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.841825994174813e-06,
+      "loss": 0.4641,
+      "step": 27355
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.841249243014102e-06,
+      "loss": 0.4939,
+      "step": 27356
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.84067249185339e-06,
+      "loss": 0.5015,
+      "step": 27357
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.840095740692679e-06,
+      "loss": 0.5026,
+      "step": 27358
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8395189895319665e-06,
+      "loss": 0.3373,
+      "step": 27359
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.838942238371255e-06,
+      "loss": 0.4069,
+      "step": 27360
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8383654872105436e-06,
+      "loss": 0.438,
+      "step": 27361
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.837788736049831e-06,
+      "loss": 0.4694,
+      "step": 27362
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.837211984889121e-06,
+      "loss": 0.4668,
+      "step": 27363
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.836635233728408e-06,
+      "loss": 0.4396,
+      "step": 27364
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.836058482567697e-06,
+      "loss": 0.3241,
+      "step": 27365
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8354817314069845e-06,
+      "loss": 0.4134,
+      "step": 27366
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.834904980246273e-06,
+      "loss": 0.3771,
+      "step": 27367
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.834328229085562e-06,
+      "loss": 0.45,
+      "step": 27368
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.833751477924849e-06,
+      "loss": 0.3429,
+      "step": 27369
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.833174726764138e-06,
+      "loss": 0.3361,
+      "step": 27370
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.832597975603426e-06,
+      "loss": 0.3947,
+      "step": 27371
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.832021224442715e-06,
+      "loss": 0.3889,
+      "step": 27372
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8314444732820026e-06,
+      "loss": 0.4159,
+      "step": 27373
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.830867722121291e-06,
+      "loss": 0.4816,
+      "step": 27374
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.83029097096058e-06,
+      "loss": 0.4415,
+      "step": 27375
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.829714219799867e-06,
+      "loss": 0.4449,
+      "step": 27376
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.829137468639156e-06,
+      "loss": 0.4051,
+      "step": 27377
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.828560717478444e-06,
+      "loss": 0.441,
+      "step": 27378
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.827983966317732e-06,
+      "loss": 0.3181,
+      "step": 27379
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8274072151570214e-06,
+      "loss": 0.3955,
+      "step": 27380
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.826830463996309e-06,
+      "loss": 0.5126,
+      "step": 27381
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.826253712835598e-06,
+      "loss": 0.3729,
+      "step": 27382
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.825676961674885e-06,
+      "loss": 0.4484,
+      "step": 27383
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.825100210514174e-06,
+      "loss": 0.3151,
+      "step": 27384
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.824523459353462e-06,
+      "loss": 0.3716,
+      "step": 27385
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.82394670819275e-06,
+      "loss": 0.5476,
+      "step": 27386
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8233699570320395e-06,
+      "loss": 0.499,
+      "step": 27387
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.822793205871327e-06,
+      "loss": 0.4709,
+      "step": 27388
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.822216454710616e-06,
+      "loss": 0.2928,
+      "step": 27389
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.821639703549903e-06,
+      "loss": 0.3842,
+      "step": 27390
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.821062952389192e-06,
+      "loss": 0.4671,
+      "step": 27391
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8204862012284804e-06,
+      "loss": 0.411,
+      "step": 27392
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.819909450067768e-06,
+      "loss": 0.431,
+      "step": 27393
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8193326989070575e-06,
+      "loss": 0.3744,
+      "step": 27394
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.818755947746345e-06,
+      "loss": 0.4286,
+      "step": 27395
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.818179196585634e-06,
+      "loss": 0.4638,
+      "step": 27396
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.817602445424922e-06,
+      "loss": 0.517,
+      "step": 27397
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.81702569426421e-06,
+      "loss": 0.4433,
+      "step": 27398
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8164489431034985e-06,
+      "loss": 0.4332,
+      "step": 27399
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.815872191942786e-06,
+      "loss": 0.4544,
+      "step": 27400
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8152954407820755e-06,
+      "loss": 0.3658,
+      "step": 27401
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.814718689621363e-06,
+      "loss": 0.4903,
+      "step": 27402
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.814141938460652e-06,
+      "loss": 0.4741,
+      "step": 27403
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.81356518729994e-06,
+      "loss": 0.4174,
+      "step": 27404
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.812988436139228e-06,
+      "loss": 0.3992,
+      "step": 27405
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8124116849785165e-06,
+      "loss": 0.4067,
+      "step": 27406
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.811834933817805e-06,
+      "loss": 0.3832,
+      "step": 27407
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.811258182657093e-06,
+      "loss": 0.3082,
+      "step": 27408
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.810681431496381e-06,
+      "loss": 0.4862,
+      "step": 27409
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.810104680335669e-06,
+      "loss": 0.4449,
+      "step": 27410
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.809527929174958e-06,
+      "loss": 0.5446,
+      "step": 27411
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.808951178014246e-06,
+      "loss": 0.3704,
+      "step": 27412
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8083744268535345e-06,
+      "loss": 0.4104,
+      "step": 27413
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.807797675692823e-06,
+      "loss": 0.607,
+      "step": 27414
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.807220924532111e-06,
+      "loss": 0.4115,
+      "step": 27415
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.806644173371399e-06,
+      "loss": 0.5123,
+      "step": 27416
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.806067422210687e-06,
+      "loss": 0.4695,
+      "step": 27417
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.805490671049976e-06,
+      "loss": 0.467,
+      "step": 27418
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.804913919889264e-06,
+      "loss": 0.3974,
+      "step": 27419
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.8043371687285526e-06,
+      "loss": 0.4532,
+      "step": 27420
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.803760417567841e-06,
+      "loss": 0.4237,
+      "step": 27421
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.803183666407129e-06,
+      "loss": 0.3908,
+      "step": 27422
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.802606915246417e-06,
+      "loss": 0.4634,
+      "step": 27423
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.802030164085706e-06,
+      "loss": 0.3879,
+      "step": 27424
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.801453412924994e-06,
+      "loss": 0.4533,
+      "step": 27425
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.800876661764282e-06,
+      "loss": 0.3267,
+      "step": 27426
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.800299910603571e-06,
+      "loss": 0.4971,
+      "step": 27427
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.799723159442859e-06,
+      "loss": 0.4329,
+      "step": 27428
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.799146408282147e-06,
+      "loss": 0.4992,
+      "step": 27429
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.798569657121435e-06,
+      "loss": 0.3188,
+      "step": 27430
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.797992905960724e-06,
+      "loss": 0.407,
+      "step": 27431
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.797416154800012e-06,
+      "loss": 0.4953,
+      "step": 27432
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.7968394036393e-06,
+      "loss": 0.4057,
+      "step": 27433
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.796262652478589e-06,
+      "loss": 0.4296,
+      "step": 27434
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.795685901317877e-06,
+      "loss": 0.3772,
+      "step": 27435
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.795109150157165e-06,
+      "loss": 0.3945,
+      "step": 27436
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.794532398996453e-06,
+      "loss": 0.4202,
+      "step": 27437
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 4.793955647835742e-06,
+      "loss": 0.4921,
+      "step": 27438
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.79337889667503e-06,
+      "loss": 0.3311,
+      "step": 27439
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.792802145514318e-06,
+      "loss": 0.4209,
+      "step": 27440
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.792225394353607e-06,
+      "loss": 0.4027,
+      "step": 27441
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.791648643192895e-06,
+      "loss": 0.2877,
+      "step": 27442
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.791071892032183e-06,
+      "loss": 0.4836,
+      "step": 27443
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.790495140871471e-06,
+      "loss": 0.4157,
+      "step": 27444
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.78991838971076e-06,
+      "loss": 0.3888,
+      "step": 27445
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.789341638550048e-06,
+      "loss": 0.3929,
+      "step": 27446
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.788764887389336e-06,
+      "loss": 0.3836,
+      "step": 27447
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.788188136228625e-06,
+      "loss": 0.3318,
+      "step": 27448
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.787611385067913e-06,
+      "loss": 0.4457,
+      "step": 27449
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.787034633907201e-06,
+      "loss": 0.4427,
+      "step": 27450
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7864578827464894e-06,
+      "loss": 0.3988,
+      "step": 27451
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.785881131585778e-06,
+      "loss": 0.4262,
+      "step": 27452
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.785304380425066e-06,
+      "loss": 0.4404,
+      "step": 27453
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.784727629264354e-06,
+      "loss": 0.3617,
+      "step": 27454
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.784150878103643e-06,
+      "loss": 0.4313,
+      "step": 27455
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.783574126942931e-06,
+      "loss": 0.4156,
+      "step": 27456
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.782997375782219e-06,
+      "loss": 0.3641,
+      "step": 27457
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7824206246215075e-06,
+      "loss": 0.5053,
+      "step": 27458
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.781843873460796e-06,
+      "loss": 0.5035,
+      "step": 27459
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.781267122300084e-06,
+      "loss": 0.521,
+      "step": 27460
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.780690371139372e-06,
+      "loss": 0.3946,
+      "step": 27461
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.780113619978661e-06,
+      "loss": 0.4331,
+      "step": 27462
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.779536868817949e-06,
+      "loss": 0.4019,
+      "step": 27463
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.778960117657237e-06,
+      "loss": 0.4631,
+      "step": 27464
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7783833664965255e-06,
+      "loss": 0.3223,
+      "step": 27465
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.777806615335814e-06,
+      "loss": 0.4217,
+      "step": 27466
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.777229864175102e-06,
+      "loss": 0.3792,
+      "step": 27467
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.77665311301439e-06,
+      "loss": 0.4808,
+      "step": 27468
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.776076361853679e-06,
+      "loss": 0.413,
+      "step": 27469
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7754996106929665e-06,
+      "loss": 0.3635,
+      "step": 27470
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.774922859532255e-06,
+      "loss": 0.4576,
+      "step": 27471
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7743461083715435e-06,
+      "loss": 0.4053,
+      "step": 27472
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.773769357210832e-06,
+      "loss": 0.442,
+      "step": 27473
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.77319260605012e-06,
+      "loss": 0.4294,
+      "step": 27474
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.772615854889408e-06,
+      "loss": 0.4076,
+      "step": 27475
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.772039103728697e-06,
+      "loss": 0.3533,
+      "step": 27476
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7714623525679845e-06,
+      "loss": 0.4146,
+      "step": 27477
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.770885601407273e-06,
+      "loss": 0.4473,
+      "step": 27478
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7703088502465616e-06,
+      "loss": 0.3878,
+      "step": 27479
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.76973209908585e-06,
+      "loss": 0.4128,
+      "step": 27480
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.769155347925138e-06,
+      "loss": 0.3409,
+      "step": 27481
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.768578596764426e-06,
+      "loss": 0.3394,
+      "step": 27482
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.768001845603715e-06,
+      "loss": 0.4131,
+      "step": 27483
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7674250944430025e-06,
+      "loss": 0.3804,
+      "step": 27484
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.766848343282292e-06,
+      "loss": 0.3791,
+      "step": 27485
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.76627159212158e-06,
+      "loss": 0.3608,
+      "step": 27486
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.765694840960868e-06,
+      "loss": 0.3588,
+      "step": 27487
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.765118089800156e-06,
+      "loss": 0.4107,
+      "step": 27488
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.764541338639444e-06,
+      "loss": 0.5162,
+      "step": 27489
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.763964587478733e-06,
+      "loss": 0.4124,
+      "step": 27490
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7633878363180206e-06,
+      "loss": 0.4424,
+      "step": 27491
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.762811085157309e-06,
+      "loss": 0.5033,
+      "step": 27492
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.762234333996598e-06,
+      "loss": 0.5002,
+      "step": 27493
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.761657582835886e-06,
+      "loss": 0.463,
+      "step": 27494
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.761080831675174e-06,
+      "loss": 0.5275,
+      "step": 27495
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.760504080514462e-06,
+      "loss": 0.3256,
+      "step": 27496
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.759927329353751e-06,
+      "loss": 0.5406,
+      "step": 27497
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.759350578193039e-06,
+      "loss": 0.3244,
+      "step": 27498
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.758773827032327e-06,
+      "loss": 0.3261,
+      "step": 27499
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.758197075871616e-06,
+      "loss": 0.3916,
+      "step": 27500
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.757620324710903e-06,
+      "loss": 0.4464,
+      "step": 27501
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.757043573550193e-06,
+      "loss": 0.3973,
+      "step": 27502
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.75646682238948e-06,
+      "loss": 0.4807,
+      "step": 27503
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.755890071228769e-06,
+      "loss": 0.4553,
+      "step": 27504
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.755313320068057e-06,
+      "loss": 0.384,
+      "step": 27505
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.754736568907345e-06,
+      "loss": 0.3934,
+      "step": 27506
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.754159817746634e-06,
+      "loss": 0.6507,
+      "step": 27507
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.753583066585921e-06,
+      "loss": 0.38,
+      "step": 27508
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.753006315425211e-06,
+      "loss": 0.3338,
+      "step": 27509
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7524295642644984e-06,
+      "loss": 0.4654,
+      "step": 27510
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.751852813103787e-06,
+      "loss": 0.3976,
+      "step": 27511
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.751276061943075e-06,
+      "loss": 0.4536,
+      "step": 27512
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.750699310782363e-06,
+      "loss": 0.4576,
+      "step": 27513
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.750122559621652e-06,
+      "loss": 0.4329,
+      "step": 27514
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.749545808460939e-06,
+      "loss": 0.3935,
+      "step": 27515
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.748969057300229e-06,
+      "loss": 0.5965,
+      "step": 27516
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7483923061395165e-06,
+      "loss": 0.3597,
+      "step": 27517
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.747815554978805e-06,
+      "loss": 0.525,
+      "step": 27518
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7472388038180935e-06,
+      "loss": 0.3895,
+      "step": 27519
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.746662052657381e-06,
+      "loss": 0.468,
+      "step": 27520
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.74608530149667e-06,
+      "loss": 0.498,
+      "step": 27521
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7455085503359574e-06,
+      "loss": 0.4808,
+      "step": 27522
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.744931799175246e-06,
+      "loss": 0.4174,
+      "step": 27523
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7443550480145345e-06,
+      "loss": 0.3968,
+      "step": 27524
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.743778296853823e-06,
+      "loss": 0.3111,
+      "step": 27525
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7432015456931116e-06,
+      "loss": 0.3997,
+      "step": 27526
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.742624794532399e-06,
+      "loss": 0.3847,
+      "step": 27527
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.742048043371688e-06,
+      "loss": 0.3933,
+      "step": 27528
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.741471292210976e-06,
+      "loss": 0.235,
+      "step": 27529
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.740894541050264e-06,
+      "loss": 0.2856,
+      "step": 27530
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7403177898895525e-06,
+      "loss": 0.4846,
+      "step": 27531
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.73974103872884e-06,
+      "loss": 0.4896,
+      "step": 27532
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.73916428756813e-06,
+      "loss": 0.4357,
+      "step": 27533
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.738587536407417e-06,
+      "loss": 0.471,
+      "step": 27534
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.738010785246706e-06,
+      "loss": 0.4975,
+      "step": 27535
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.737434034085994e-06,
+      "loss": 0.445,
+      "step": 27536
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.736857282925282e-06,
+      "loss": 0.3901,
+      "step": 27537
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7362805317645706e-06,
+      "loss": 0.4151,
+      "step": 27538
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.735703780603858e-06,
+      "loss": 0.3898,
+      "step": 27539
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.735127029443148e-06,
+      "loss": 0.3782,
+      "step": 27540
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.734550278282435e-06,
+      "loss": 0.4749,
+      "step": 27541
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.733973527121724e-06,
+      "loss": 0.3226,
+      "step": 27542
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.733396775961012e-06,
+      "loss": 0.3389,
+      "step": 27543
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7328200248003e-06,
+      "loss": 0.3679,
+      "step": 27544
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.732243273639589e-06,
+      "loss": 0.3018,
+      "step": 27545
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.731666522478877e-06,
+      "loss": 0.4206,
+      "step": 27546
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.731089771318166e-06,
+      "loss": 0.332,
+      "step": 27547
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.730513020157453e-06,
+      "loss": 0.5603,
+      "step": 27548
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.729936268996742e-06,
+      "loss": 0.3613,
+      "step": 27549
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.72935951783603e-06,
+      "loss": 0.3936,
+      "step": 27550
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.728782766675318e-06,
+      "loss": 0.4058,
+      "step": 27551
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.728206015514607e-06,
+      "loss": 0.4698,
+      "step": 27552
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.727629264353895e-06,
+      "loss": 0.4441,
+      "step": 27553
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.727052513193183e-06,
+      "loss": 0.491,
+      "step": 27554
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.726475762032471e-06,
+      "loss": 0.3261,
+      "step": 27555
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.72589901087176e-06,
+      "loss": 0.379,
+      "step": 27556
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7253222597110484e-06,
+      "loss": 0.3304,
+      "step": 27557
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.724745508550336e-06,
+      "loss": 0.4121,
+      "step": 27558
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.724168757389625e-06,
+      "loss": 0.3308,
+      "step": 27559
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.723592006228913e-06,
+      "loss": 0.4371,
+      "step": 27560
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.723015255068201e-06,
+      "loss": 0.5001,
+      "step": 27561
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.722438503907489e-06,
+      "loss": 0.4659,
+      "step": 27562
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.721861752746778e-06,
+      "loss": 0.4377,
+      "step": 27563
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7212850015860665e-06,
+      "loss": 0.3488,
+      "step": 27564
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.720708250425354e-06,
+      "loss": 0.4411,
+      "step": 27565
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.720131499264643e-06,
+      "loss": 0.4285,
+      "step": 27566
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.719554748103931e-06,
+      "loss": 0.47,
+      "step": 27567
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.718977996943219e-06,
+      "loss": 0.3947,
+      "step": 27568
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7184012457825074e-06,
+      "loss": 0.3835,
+      "step": 27569
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.717824494621796e-06,
+      "loss": 0.476,
+      "step": 27570
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7172477434610845e-06,
+      "loss": 0.535,
+      "step": 27571
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.716670992300372e-06,
+      "loss": 0.4126,
+      "step": 27572
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.716094241139661e-06,
+      "loss": 0.439,
+      "step": 27573
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.715517489978949e-06,
+      "loss": 0.5767,
+      "step": 27574
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.714940738818237e-06,
+      "loss": 0.3941,
+      "step": 27575
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7143639876575255e-06,
+      "loss": 0.4547,
+      "step": 27576
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.713787236496814e-06,
+      "loss": 0.2611,
+      "step": 27577
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7132104853361025e-06,
+      "loss": 0.4736,
+      "step": 27578
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.71263373417539e-06,
+      "loss": 0.6327,
+      "step": 27579
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.712056983014679e-06,
+      "loss": 0.366,
+      "step": 27580
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.711480231853967e-06,
+      "loss": 0.3668,
+      "step": 27581
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.710903480693255e-06,
+      "loss": 0.3315,
+      "step": 27582
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7103267295325435e-06,
+      "loss": 0.4387,
+      "step": 27583
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.709749978371832e-06,
+      "loss": 0.3843,
+      "step": 27584
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7091732272111206e-06,
+      "loss": 0.3411,
+      "step": 27585
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.708596476050408e-06,
+      "loss": 0.4674,
+      "step": 27586
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.708019724889697e-06,
+      "loss": 0.424,
+      "step": 27587
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.707442973728985e-06,
+      "loss": 0.4219,
+      "step": 27588
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.706866222568273e-06,
+      "loss": 0.4191,
+      "step": 27589
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7062894714075615e-06,
+      "loss": 0.3706,
+      "step": 27590
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.70571272024685e-06,
+      "loss": 0.4944,
+      "step": 27591
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.705135969086138e-06,
+      "loss": 0.4685,
+      "step": 27592
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.704559217925426e-06,
+      "loss": 0.2658,
+      "step": 27593
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.703982466764715e-06,
+      "loss": 0.4088,
+      "step": 27594
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.703405715604003e-06,
+      "loss": 0.4764,
+      "step": 27595
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.702828964443291e-06,
+      "loss": 0.3455,
+      "step": 27596
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.7022522132825796e-06,
+      "loss": 0.4651,
+      "step": 27597
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.701675462121868e-06,
+      "loss": 0.3725,
+      "step": 27598
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.701098710961156e-06,
+      "loss": 0.3277,
+      "step": 27599
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.700521959800444e-06,
+      "loss": 0.434,
+      "step": 27600
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.699945208639733e-06,
+      "loss": 0.5095,
+      "step": 27601
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.699368457479021e-06,
+      "loss": 0.3823,
+      "step": 27602
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.698791706318309e-06,
+      "loss": 0.3816,
+      "step": 27603
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.698214955157598e-06,
+      "loss": 0.4691,
+      "step": 27604
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.697638203996886e-06,
+      "loss": 0.4215,
+      "step": 27605
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.697061452836174e-06,
+      "loss": 0.3096,
+      "step": 27606
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.696484701675463e-06,
+      "loss": 0.4063,
+      "step": 27607
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.695907950514751e-06,
+      "loss": 0.3921,
+      "step": 27608
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.695331199354039e-06,
+      "loss": 0.3695,
+      "step": 27609
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.694754448193327e-06,
+      "loss": 0.3351,
+      "step": 27610
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.694177697032616e-06,
+      "loss": 0.4382,
+      "step": 27611
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.693600945871904e-06,
+      "loss": 0.5237,
+      "step": 27612
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.693024194711192e-06,
+      "loss": 0.4192,
+      "step": 27613
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.69244744355048e-06,
+      "loss": 0.4628,
+      "step": 27614
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.691870692389769e-06,
+      "loss": 0.3002,
+      "step": 27615
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 4.6912939412290574e-06,
+      "loss": 0.4007,
+      "step": 27616
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.690717190068345e-06,
+      "loss": 0.3622,
+      "step": 27617
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.690140438907634e-06,
+      "loss": 0.3802,
+      "step": 27618
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.689563687746922e-06,
+      "loss": 0.3718,
+      "step": 27619
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.68898693658621e-06,
+      "loss": 0.4033,
+      "step": 27620
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.688410185425498e-06,
+      "loss": 0.3685,
+      "step": 27621
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.687833434264787e-06,
+      "loss": 0.3663,
+      "step": 27622
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.687256683104075e-06,
+      "loss": 0.4207,
+      "step": 27623
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.686679931943364e-06,
+      "loss": 0.3764,
+      "step": 27624
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.686103180782652e-06,
+      "loss": 0.4918,
+      "step": 27625
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.68552642962194e-06,
+      "loss": 0.4725,
+      "step": 27626
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.684949678461228e-06,
+      "loss": 0.4469,
+      "step": 27627
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6843729273005164e-06,
+      "loss": 0.3653,
+      "step": 27628
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.683796176139805e-06,
+      "loss": 0.3853,
+      "step": 27629
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.683219424979093e-06,
+      "loss": 0.3828,
+      "step": 27630
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.682642673818382e-06,
+      "loss": 0.406,
+      "step": 27631
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.68206592265767e-06,
+      "loss": 0.4081,
+      "step": 27632
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.681489171496958e-06,
+      "loss": 0.4526,
+      "step": 27633
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.680912420336246e-06,
+      "loss": 0.3235,
+      "step": 27634
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6803356691755345e-06,
+      "loss": 0.4828,
+      "step": 27635
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.679758918014823e-06,
+      "loss": 0.328,
+      "step": 27636
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.679182166854111e-06,
+      "loss": 0.6143,
+      "step": 27637
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6786054156934e-06,
+      "loss": 0.401,
+      "step": 27638
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.678028664532688e-06,
+      "loss": 0.3733,
+      "step": 27639
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.677451913371976e-06,
+      "loss": 0.5212,
+      "step": 27640
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.676875162211265e-06,
+      "loss": 0.4174,
+      "step": 27641
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6762984110505525e-06,
+      "loss": 0.4503,
+      "step": 27642
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.675721659889841e-06,
+      "loss": 0.3995,
+      "step": 27643
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.675144908729129e-06,
+      "loss": 0.3723,
+      "step": 27644
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.674568157568417e-06,
+      "loss": 0.5022,
+      "step": 27645
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.673991406407706e-06,
+      "loss": 0.3576,
+      "step": 27646
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.673414655246994e-06,
+      "loss": 0.4662,
+      "step": 27647
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.672837904086283e-06,
+      "loss": 0.5401,
+      "step": 27648
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6722611529255705e-06,
+      "loss": 0.4355,
+      "step": 27649
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.671684401764859e-06,
+      "loss": 0.3767,
+      "step": 27650
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.671107650604147e-06,
+      "loss": 0.4496,
+      "step": 27651
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.670530899443435e-06,
+      "loss": 0.3445,
+      "step": 27652
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.669954148282724e-06,
+      "loss": 0.2779,
+      "step": 27653
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6693773971220115e-06,
+      "loss": 0.5398,
+      "step": 27654
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.668800645961301e-06,
+      "loss": 0.3681,
+      "step": 27655
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6682238948005885e-06,
+      "loss": 0.5172,
+      "step": 27656
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.667647143639877e-06,
+      "loss": 0.42,
+      "step": 27657
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.667070392479166e-06,
+      "loss": 0.4761,
+      "step": 27658
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.666493641318453e-06,
+      "loss": 0.372,
+      "step": 27659
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.665916890157742e-06,
+      "loss": 0.2577,
+      "step": 27660
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6653401389970295e-06,
+      "loss": 0.3693,
+      "step": 27661
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.664763387836319e-06,
+      "loss": 0.4768,
+      "step": 27662
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.664186636675607e-06,
+      "loss": 0.3726,
+      "step": 27663
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.663609885514895e-06,
+      "loss": 0.3691,
+      "step": 27664
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.663033134354184e-06,
+      "loss": 0.4066,
+      "step": 27665
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.662456383193471e-06,
+      "loss": 0.5172,
+      "step": 27666
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.66187963203276e-06,
+      "loss": 0.3653,
+      "step": 27667
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.661302880872048e-06,
+      "loss": 0.2526,
+      "step": 27668
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.660726129711337e-06,
+      "loss": 0.4637,
+      "step": 27669
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.660149378550625e-06,
+      "loss": 0.3303,
+      "step": 27670
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.659572627389913e-06,
+      "loss": 0.3311,
+      "step": 27671
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.658995876229202e-06,
+      "loss": 0.3668,
+      "step": 27672
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.658419125068489e-06,
+      "loss": 0.4957,
+      "step": 27673
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.657842373907778e-06,
+      "loss": 0.3547,
+      "step": 27674
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6572656227470664e-06,
+      "loss": 0.3911,
+      "step": 27675
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.656688871586354e-06,
+      "loss": 0.3493,
+      "step": 27676
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.656112120425643e-06,
+      "loss": 0.5095,
+      "step": 27677
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.655535369264931e-06,
+      "loss": 0.3993,
+      "step": 27678
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.65495861810422e-06,
+      "loss": 0.3926,
+      "step": 27679
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.654381866943507e-06,
+      "loss": 0.3663,
+      "step": 27680
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.653805115782796e-06,
+      "loss": 0.3974,
+      "step": 27681
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6532283646220845e-06,
+      "loss": 0.4697,
+      "step": 27682
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.652651613461372e-06,
+      "loss": 0.4353,
+      "step": 27683
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.652074862300661e-06,
+      "loss": 0.3028,
+      "step": 27684
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.651498111139949e-06,
+      "loss": 0.4128,
+      "step": 27685
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.650921359979238e-06,
+      "loss": 0.4107,
+      "step": 27686
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.650344608818525e-06,
+      "loss": 0.34,
+      "step": 27687
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.649767857657814e-06,
+      "loss": 0.3973,
+      "step": 27688
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6491911064971025e-06,
+      "loss": 0.4841,
+      "step": 27689
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.64861435533639e-06,
+      "loss": 0.4707,
+      "step": 27690
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.648037604175679e-06,
+      "loss": 0.4678,
+      "step": 27691
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.647460853014967e-06,
+      "loss": 0.4432,
+      "step": 27692
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.646884101854256e-06,
+      "loss": 0.4532,
+      "step": 27693
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6463073506935435e-06,
+      "loss": 0.4399,
+      "step": 27694
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.645730599532832e-06,
+      "loss": 0.3706,
+      "step": 27695
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6451538483721205e-06,
+      "loss": 0.4488,
+      "step": 27696
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.644577097211408e-06,
+      "loss": 0.4189,
+      "step": 27697
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.644000346050697e-06,
+      "loss": 0.4018,
+      "step": 27698
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.643423594889985e-06,
+      "loss": 0.5141,
+      "step": 27699
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.642846843729274e-06,
+      "loss": 0.4291,
+      "step": 27700
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6422700925685615e-06,
+      "loss": 0.3964,
+      "step": 27701
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.64169334140785e-06,
+      "loss": 0.3106,
+      "step": 27702
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6411165902471385e-06,
+      "loss": 0.5422,
+      "step": 27703
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.640539839086426e-06,
+      "loss": 0.5,
+      "step": 27704
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.639963087925715e-06,
+      "loss": 0.3721,
+      "step": 27705
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.639386336765003e-06,
+      "loss": 0.4675,
+      "step": 27706
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.638809585604291e-06,
+      "loss": 0.5076,
+      "step": 27707
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6382328344435795e-06,
+      "loss": 0.4561,
+      "step": 27708
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.637656083282868e-06,
+      "loss": 0.3886,
+      "step": 27709
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.637079332122157e-06,
+      "loss": 0.2872,
+      "step": 27710
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.636502580961444e-06,
+      "loss": 0.507,
+      "step": 27711
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.635925829800733e-06,
+      "loss": 0.444,
+      "step": 27712
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.635349078640021e-06,
+      "loss": 0.5026,
+      "step": 27713
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.634772327479309e-06,
+      "loss": 0.4611,
+      "step": 27714
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6341955763185975e-06,
+      "loss": 0.4659,
+      "step": 27715
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.633618825157886e-06,
+      "loss": 0.3957,
+      "step": 27716
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.633042073997175e-06,
+      "loss": 0.4309,
+      "step": 27717
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.632465322836462e-06,
+      "loss": 0.4994,
+      "step": 27718
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.631888571675751e-06,
+      "loss": 0.4644,
+      "step": 27719
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.631311820515039e-06,
+      "loss": 0.4156,
+      "step": 27720
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.630735069354327e-06,
+      "loss": 0.4169,
+      "step": 27721
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6301583181936156e-06,
+      "loss": 0.459,
+      "step": 27722
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.629581567032904e-06,
+      "loss": 0.3675,
+      "step": 27723
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.629004815872193e-06,
+      "loss": 0.3352,
+      "step": 27724
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62842806471148e-06,
+      "loss": 0.4245,
+      "step": 27725
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.627851313550769e-06,
+      "loss": 0.3818,
+      "step": 27726
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.627274562390057e-06,
+      "loss": 0.3896,
+      "step": 27727
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.626697811229345e-06,
+      "loss": 0.3744,
+      "step": 27728
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6261210600686345e-06,
+      "loss": 0.582,
+      "step": 27729
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.625544308907922e-06,
+      "loss": 0.3656,
+      "step": 27730
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.624967557747211e-06,
+      "loss": 0.3886,
+      "step": 27731
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.624390806586498e-06,
+      "loss": 0.3826,
+      "step": 27732
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.623814055425787e-06,
+      "loss": 0.4476,
+      "step": 27733
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.623237304265075e-06,
+      "loss": 0.5228,
+      "step": 27734
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.622660553104363e-06,
+      "loss": 0.4099,
+      "step": 27735
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.622083801943652e-06,
+      "loss": 0.4485,
+      "step": 27736
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.62150705078294e-06,
+      "loss": 0.4035,
+      "step": 27737
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.620930299622228e-06,
+      "loss": 0.4204,
+      "step": 27738
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.620353548461516e-06,
+      "loss": 0.4087,
+      "step": 27739
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.619776797300805e-06,
+      "loss": 0.3967,
+      "step": 27740
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6192000461400935e-06,
+      "loss": 0.3163,
+      "step": 27741
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.618623294979381e-06,
+      "loss": 0.2637,
+      "step": 27742
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.61804654381867e-06,
+      "loss": 0.4642,
+      "step": 27743
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.617469792657958e-06,
+      "loss": 0.4187,
+      "step": 27744
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.616893041497246e-06,
+      "loss": 0.3192,
+      "step": 27745
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.616316290336535e-06,
+      "loss": 0.3749,
+      "step": 27746
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.615739539175823e-06,
+      "loss": 0.2435,
+      "step": 27747
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6151627880151115e-06,
+      "loss": 0.408,
+      "step": 27748
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.614586036854399e-06,
+      "loss": 0.3313,
+      "step": 27749
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.614009285693688e-06,
+      "loss": 0.5232,
+      "step": 27750
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.613432534532976e-06,
+      "loss": 0.3404,
+      "step": 27751
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.612855783372264e-06,
+      "loss": 0.4337,
+      "step": 27752
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.612279032211553e-06,
+      "loss": 0.3507,
+      "step": 27753
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.611702281050841e-06,
+      "loss": 0.4363,
+      "step": 27754
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6111255298901295e-06,
+      "loss": 0.5465,
+      "step": 27755
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.610548778729417e-06,
+      "loss": 0.5035,
+      "step": 27756
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.609972027568706e-06,
+      "loss": 0.3587,
+      "step": 27757
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.609395276407994e-06,
+      "loss": 0.4871,
+      "step": 27758
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.608818525247282e-06,
+      "loss": 0.3534,
+      "step": 27759
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.608241774086571e-06,
+      "loss": 0.3647,
+      "step": 27760
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.607665022925859e-06,
+      "loss": 0.4077,
+      "step": 27761
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6070882717651475e-06,
+      "loss": 0.3205,
+      "step": 27762
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.606511520604436e-06,
+      "loss": 0.4553,
+      "step": 27763
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.605934769443724e-06,
+      "loss": 0.3762,
+      "step": 27764
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.605358018283012e-06,
+      "loss": 0.6092,
+      "step": 27765
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6047812671223e-06,
+      "loss": 0.3698,
+      "step": 27766
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6042045159615885e-06,
+      "loss": 0.4675,
+      "step": 27767
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.603627764800877e-06,
+      "loss": 0.4524,
+      "step": 27768
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6030510136401656e-06,
+      "loss": 0.5297,
+      "step": 27769
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.602474262479454e-06,
+      "loss": 0.3889,
+      "step": 27770
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.601897511318742e-06,
+      "loss": 0.3726,
+      "step": 27771
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.60132076015803e-06,
+      "loss": 0.4767,
+      "step": 27772
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.600744008997318e-06,
+      "loss": 0.3452,
+      "step": 27773
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.6001672578366065e-06,
+      "loss": 0.4467,
+      "step": 27774
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.599590506675895e-06,
+      "loss": 0.5131,
+      "step": 27775
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.599013755515183e-06,
+      "loss": 0.4266,
+      "step": 27776
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.598437004354472e-06,
+      "loss": 0.5384,
+      "step": 27777
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.59786025319376e-06,
+      "loss": 0.3927,
+      "step": 27778
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.597283502033048e-06,
+      "loss": 0.4869,
+      "step": 27779
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.596706750872337e-06,
+      "loss": 0.3134,
+      "step": 27780
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.5961299997116246e-06,
+      "loss": 0.444,
+      "step": 27781
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.595553248550913e-06,
+      "loss": 0.4311,
+      "step": 27782
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.594976497390201e-06,
+      "loss": 0.3147,
+      "step": 27783
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.59439974622949e-06,
+      "loss": 0.3402,
+      "step": 27784
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.593822995068778e-06,
+      "loss": 0.5138,
+      "step": 27785
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.593246243908066e-06,
+      "loss": 0.4258,
+      "step": 27786
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.592669492747355e-06,
+      "loss": 0.3664,
+      "step": 27787
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.592092741586643e-06,
+      "loss": 0.485,
+      "step": 27788
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.591515990425931e-06,
+      "loss": 0.3672,
+      "step": 27789
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.59093923926522e-06,
+      "loss": 0.4363,
+      "step": 27790
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.590362488104508e-06,
+      "loss": 0.3526,
+      "step": 27791
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.589785736943796e-06,
+      "loss": 0.3691,
+      "step": 27792
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.589208985783084e-06,
+      "loss": 0.3104,
+      "step": 27793
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.588632234622373e-06,
+      "loss": 0.3994,
+      "step": 27794
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 4.588055483461661e-06,
+      "loss": 0.3649,
+      "step": 27795
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.587478732300949e-06,
+      "loss": 0.3574,
+      "step": 27796
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.586901981140238e-06,
+      "loss": 0.4641,
+      "step": 27797
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.586325229979525e-06,
+      "loss": 0.4899,
+      "step": 27798
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.585748478818814e-06,
+      "loss": 0.3727,
+      "step": 27799
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5851717276581024e-06,
+      "loss": 0.4165,
+      "step": 27800
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.584594976497391e-06,
+      "loss": 0.4069,
+      "step": 27801
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.584018225336679e-06,
+      "loss": 0.4005,
+      "step": 27802
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.583441474175967e-06,
+      "loss": 0.5465,
+      "step": 27803
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.582864723015256e-06,
+      "loss": 0.4929,
+      "step": 27804
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.582287971854543e-06,
+      "loss": 0.4897,
+      "step": 27805
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.581711220693832e-06,
+      "loss": 0.4722,
+      "step": 27806
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5811344695331205e-06,
+      "loss": 0.3477,
+      "step": 27807
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.580557718372409e-06,
+      "loss": 0.3088,
+      "step": 27808
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.579980967211697e-06,
+      "loss": 0.4401,
+      "step": 27809
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.579404216050985e-06,
+      "loss": 0.4697,
+      "step": 27810
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.578827464890274e-06,
+      "loss": 0.2779,
+      "step": 27811
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5782507137295614e-06,
+      "loss": 0.4061,
+      "step": 27812
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.57767396256885e-06,
+      "loss": 0.4673,
+      "step": 27813
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5770972114081385e-06,
+      "loss": 0.4801,
+      "step": 27814
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.576520460247427e-06,
+      "loss": 0.4754,
+      "step": 27815
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.575943709086715e-06,
+      "loss": 0.3053,
+      "step": 27816
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.575366957926003e-06,
+      "loss": 0.5017,
+      "step": 27817
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.574790206765292e-06,
+      "loss": 0.3792,
+      "step": 27818
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5742134556045795e-06,
+      "loss": 0.4455,
+      "step": 27819
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.573636704443868e-06,
+      "loss": 0.4525,
+      "step": 27820
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5730599532831565e-06,
+      "loss": 0.4062,
+      "step": 27821
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.572483202122445e-06,
+      "loss": 0.4307,
+      "step": 27822
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.571906450961733e-06,
+      "loss": 0.4013,
+      "step": 27823
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.571329699801021e-06,
+      "loss": 0.3898,
+      "step": 27824
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.57075294864031e-06,
+      "loss": 0.4335,
+      "step": 27825
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5701761974795975e-06,
+      "loss": 0.2742,
+      "step": 27826
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.569599446318886e-06,
+      "loss": 0.4984,
+      "step": 27827
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5690226951581746e-06,
+      "loss": 0.3919,
+      "step": 27828
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.568445943997462e-06,
+      "loss": 0.3189,
+      "step": 27829
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.567869192836751e-06,
+      "loss": 0.2995,
+      "step": 27830
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.567292441676039e-06,
+      "loss": 0.4559,
+      "step": 27831
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.566715690515328e-06,
+      "loss": 0.3988,
+      "step": 27832
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5661389393546155e-06,
+      "loss": 0.31,
+      "step": 27833
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.565562188193904e-06,
+      "loss": 0.3474,
+      "step": 27834
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.564985437033193e-06,
+      "loss": 0.3476,
+      "step": 27835
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.56440868587248e-06,
+      "loss": 0.4977,
+      "step": 27836
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.563831934711769e-06,
+      "loss": 0.4178,
+      "step": 27837
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.563255183551057e-06,
+      "loss": 0.3724,
+      "step": 27838
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.562678432390346e-06,
+      "loss": 0.4392,
+      "step": 27839
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5621016812296336e-06,
+      "loss": 0.4059,
+      "step": 27840
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.561524930068922e-06,
+      "loss": 0.4093,
+      "step": 27841
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.560948178908211e-06,
+      "loss": 0.4691,
+      "step": 27842
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.560371427747498e-06,
+      "loss": 0.3636,
+      "step": 27843
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.559794676586787e-06,
+      "loss": 0.4835,
+      "step": 27844
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.559217925426075e-06,
+      "loss": 0.4345,
+      "step": 27845
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.558641174265364e-06,
+      "loss": 0.3759,
+      "step": 27846
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.558064423104652e-06,
+      "loss": 0.4157,
+      "step": 27847
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.55748767194394e-06,
+      "loss": 0.4147,
+      "step": 27848
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.556910920783229e-06,
+      "loss": 0.3608,
+      "step": 27849
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.556334169622516e-06,
+      "loss": 0.3943,
+      "step": 27850
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.555757418461805e-06,
+      "loss": 0.4286,
+      "step": 27851
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.555180667301093e-06,
+      "loss": 0.3345,
+      "step": 27852
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.554603916140382e-06,
+      "loss": 0.4342,
+      "step": 27853
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.55402716497967e-06,
+      "loss": 0.4455,
+      "step": 27854
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.553450413818958e-06,
+      "loss": 0.3427,
+      "step": 27855
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.552873662658247e-06,
+      "loss": 0.4914,
+      "step": 27856
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.552296911497534e-06,
+      "loss": 0.3415,
+      "step": 27857
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.551720160336823e-06,
+      "loss": 0.4176,
+      "step": 27858
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5511434091761114e-06,
+      "loss": 0.458,
+      "step": 27859
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.550566658015399e-06,
+      "loss": 0.475,
+      "step": 27860
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.549989906854688e-06,
+      "loss": 0.4344,
+      "step": 27861
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.549413155693976e-06,
+      "loss": 0.3366,
+      "step": 27862
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.548836404533265e-06,
+      "loss": 0.4121,
+      "step": 27863
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.548259653372552e-06,
+      "loss": 0.4277,
+      "step": 27864
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.547682902211841e-06,
+      "loss": 0.4171,
+      "step": 27865
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5471061510511295e-06,
+      "loss": 0.4464,
+      "step": 27866
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.546529399890417e-06,
+      "loss": 0.4942,
+      "step": 27867
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5459526487297065e-06,
+      "loss": 0.4568,
+      "step": 27868
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.545375897568994e-06,
+      "loss": 0.3863,
+      "step": 27869
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.544799146408283e-06,
+      "loss": 0.4794,
+      "step": 27870
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5442223952475704e-06,
+      "loss": 0.487,
+      "step": 27871
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.543645644086859e-06,
+      "loss": 0.4224,
+      "step": 27872
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5430688929261475e-06,
+      "loss": 0.6271,
+      "step": 27873
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.542492141765435e-06,
+      "loss": 0.476,
+      "step": 27874
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5419153906047246e-06,
+      "loss": 0.4482,
+      "step": 27875
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.541338639444012e-06,
+      "loss": 0.4385,
+      "step": 27876
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.540761888283301e-06,
+      "loss": 0.4486,
+      "step": 27877
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5401851371225885e-06,
+      "loss": 0.4394,
+      "step": 27878
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.539608385961877e-06,
+      "loss": 0.493,
+      "step": 27879
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5390316348011655e-06,
+      "loss": 0.407,
+      "step": 27880
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.538454883640453e-06,
+      "loss": 0.421,
+      "step": 27881
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.537878132479743e-06,
+      "loss": 0.3352,
+      "step": 27882
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.53730138131903e-06,
+      "loss": 0.3937,
+      "step": 27883
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.536724630158319e-06,
+      "loss": 0.3266,
+      "step": 27884
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.536147878997607e-06,
+      "loss": 0.4072,
+      "step": 27885
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.535571127836895e-06,
+      "loss": 0.4195,
+      "step": 27886
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5349943766761836e-06,
+      "loss": 0.4803,
+      "step": 27887
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.534417625515471e-06,
+      "loss": 0.3344,
+      "step": 27888
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.53384087435476e-06,
+      "loss": 0.3288,
+      "step": 27889
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.533264123194048e-06,
+      "loss": 0.3675,
+      "step": 27890
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.532687372033336e-06,
+      "loss": 0.476,
+      "step": 27891
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.532110620872625e-06,
+      "loss": 0.2712,
+      "step": 27892
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.531533869711913e-06,
+      "loss": 0.468,
+      "step": 27893
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.530957118551202e-06,
+      "loss": 0.3707,
+      "step": 27894
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.530380367390489e-06,
+      "loss": 0.4789,
+      "step": 27895
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.529803616229778e-06,
+      "loss": 0.2943,
+      "step": 27896
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.529226865069066e-06,
+      "loss": 0.2662,
+      "step": 27897
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.528650113908354e-06,
+      "loss": 0.3951,
+      "step": 27898
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.528073362747643e-06,
+      "loss": 0.3847,
+      "step": 27899
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.527496611586931e-06,
+      "loss": 0.4077,
+      "step": 27900
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.52691986042622e-06,
+      "loss": 0.337,
+      "step": 27901
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.526343109265508e-06,
+      "loss": 0.4013,
+      "step": 27902
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.525766358104796e-06,
+      "loss": 0.5178,
+      "step": 27903
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.525189606944084e-06,
+      "loss": 0.399,
+      "step": 27904
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.524612855783372e-06,
+      "loss": 0.5104,
+      "step": 27905
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5240361046226614e-06,
+      "loss": 0.3805,
+      "step": 27906
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.523459353461949e-06,
+      "loss": 0.359,
+      "step": 27907
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.522882602301238e-06,
+      "loss": 0.4245,
+      "step": 27908
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.522305851140526e-06,
+      "loss": 0.4712,
+      "step": 27909
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.521729099979814e-06,
+      "loss": 0.3853,
+      "step": 27910
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.521152348819102e-06,
+      "loss": 0.3811,
+      "step": 27911
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.52057559765839e-06,
+      "loss": 0.3272,
+      "step": 27912
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5199988464976795e-06,
+      "loss": 0.5548,
+      "step": 27913
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.519422095336967e-06,
+      "loss": 0.387,
+      "step": 27914
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.518845344176256e-06,
+      "loss": 0.5286,
+      "step": 27915
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.518268593015544e-06,
+      "loss": 0.3992,
+      "step": 27916
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.517691841854832e-06,
+      "loss": 0.3223,
+      "step": 27917
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5171150906941204e-06,
+      "loss": 0.326,
+      "step": 27918
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.516538339533409e-06,
+      "loss": 0.3014,
+      "step": 27919
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.515961588372697e-06,
+      "loss": 0.3956,
+      "step": 27920
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.515384837211985e-06,
+      "loss": 0.2928,
+      "step": 27921
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.514808086051273e-06,
+      "loss": 0.3943,
+      "step": 27922
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.514231334890562e-06,
+      "loss": 0.4202,
+      "step": 27923
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.51365458372985e-06,
+      "loss": 0.3608,
+      "step": 27924
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5130778325691385e-06,
+      "loss": 0.4704,
+      "step": 27925
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.512501081408427e-06,
+      "loss": 0.4095,
+      "step": 27926
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.511924330247715e-06,
+      "loss": 0.531,
+      "step": 27927
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.511347579087003e-06,
+      "loss": 0.3286,
+      "step": 27928
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.510770827926292e-06,
+      "loss": 0.3216,
+      "step": 27929
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.51019407676558e-06,
+      "loss": 0.3481,
+      "step": 27930
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.509617325604868e-06,
+      "loss": 0.4316,
+      "step": 27931
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5090405744441565e-06,
+      "loss": 0.4629,
+      "step": 27932
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.508463823283445e-06,
+      "loss": 0.3545,
+      "step": 27933
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.507887072122733e-06,
+      "loss": 0.4595,
+      "step": 27934
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.507310320962021e-06,
+      "loss": 0.4729,
+      "step": 27935
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.50673356980131e-06,
+      "loss": 0.4449,
+      "step": 27936
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.506156818640598e-06,
+      "loss": 0.3816,
+      "step": 27937
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.505580067479886e-06,
+      "loss": 0.4672,
+      "step": 27938
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5050033163191745e-06,
+      "loss": 0.4181,
+      "step": 27939
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.504426565158463e-06,
+      "loss": 0.3829,
+      "step": 27940
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.503849813997751e-06,
+      "loss": 0.4347,
+      "step": 27941
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.503273062837039e-06,
+      "loss": 0.3723,
+      "step": 27942
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.502696311676328e-06,
+      "loss": 0.4074,
+      "step": 27943
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.502119560515616e-06,
+      "loss": 0.5019,
+      "step": 27944
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.501542809354904e-06,
+      "loss": 0.4197,
+      "step": 27945
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.5009660581941926e-06,
+      "loss": 0.3543,
+      "step": 27946
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.500389307033481e-06,
+      "loss": 0.3933,
+      "step": 27947
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.499812555872769e-06,
+      "loss": 0.5593,
+      "step": 27948
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.499235804712057e-06,
+      "loss": 0.3736,
+      "step": 27949
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.498659053551346e-06,
+      "loss": 0.368,
+      "step": 27950
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.4980823023906335e-06,
+      "loss": 0.4045,
+      "step": 27951
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.497505551229922e-06,
+      "loss": 0.3301,
+      "step": 27952
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.496928800069211e-06,
+      "loss": 0.4033,
+      "step": 27953
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.496352048908499e-06,
+      "loss": 0.5588,
+      "step": 27954
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.495775297747787e-06,
+      "loss": 0.4538,
+      "step": 27955
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.495198546587075e-06,
+      "loss": 0.3076,
+      "step": 27956
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.494621795426364e-06,
+      "loss": 0.4576,
+      "step": 27957
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.4940450442656516e-06,
+      "loss": 0.482,
+      "step": 27958
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.49346829310494e-06,
+      "loss": 0.4925,
+      "step": 27959
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.492891541944229e-06,
+      "loss": 0.4414,
+      "step": 27960
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.492314790783517e-06,
+      "loss": 0.3064,
+      "step": 27961
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.491738039622805e-06,
+      "loss": 0.4187,
+      "step": 27962
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.491161288462093e-06,
+      "loss": 0.388,
+      "step": 27963
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.490584537301382e-06,
+      "loss": 0.5284,
+      "step": 27964
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.49000778614067e-06,
+      "loss": 0.3172,
+      "step": 27965
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.489431034979958e-06,
+      "loss": 0.3412,
+      "step": 27966
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.488854283819247e-06,
+      "loss": 0.6549,
+      "step": 27967
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.488277532658535e-06,
+      "loss": 0.3607,
+      "step": 27968
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.487700781497823e-06,
+      "loss": 0.4891,
+      "step": 27969
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.487124030337111e-06,
+      "loss": 0.3965,
+      "step": 27970
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.4865472791764e-06,
+      "loss": 0.4653,
+      "step": 27971
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.485970528015688e-06,
+      "loss": 0.3915,
+      "step": 27972
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.485393776854976e-06,
+      "loss": 0.5607,
+      "step": 27973
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 4.484817025694265e-06,
+      "loss": 0.3354,
+      "step": 27974
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.484240274533553e-06,
+      "loss": 0.5001,
+      "step": 27975
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.483663523372841e-06,
+      "loss": 0.3282,
+      "step": 27976
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4830867722121294e-06,
+      "loss": 0.4403,
+      "step": 27977
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.482510021051418e-06,
+      "loss": 0.4633,
+      "step": 27978
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.481933269890706e-06,
+      "loss": 0.4082,
+      "step": 27979
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.481356518729994e-06,
+      "loss": 0.3716,
+      "step": 27980
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.480779767569283e-06,
+      "loss": 0.3677,
+      "step": 27981
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.48020301640857e-06,
+      "loss": 0.4397,
+      "step": 27982
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.479626265247859e-06,
+      "loss": 0.3711,
+      "step": 27983
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4790495140871475e-06,
+      "loss": 0.4133,
+      "step": 27984
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.478472762926436e-06,
+      "loss": 0.4245,
+      "step": 27985
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.477896011765724e-06,
+      "loss": 0.3948,
+      "step": 27986
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.477319260605012e-06,
+      "loss": 0.4832,
+      "step": 27987
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.476742509444301e-06,
+      "loss": 0.4575,
+      "step": 27988
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4761657582835884e-06,
+      "loss": 0.3769,
+      "step": 27989
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.475589007122878e-06,
+      "loss": 0.3951,
+      "step": 27990
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4750122559621655e-06,
+      "loss": 0.3256,
+      "step": 27991
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.474435504801454e-06,
+      "loss": 0.464,
+      "step": 27992
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.473858753640742e-06,
+      "loss": 0.4028,
+      "step": 27993
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.47328200248003e-06,
+      "loss": 0.5134,
+      "step": 27994
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.472705251319319e-06,
+      "loss": 0.4008,
+      "step": 27995
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4721285001586065e-06,
+      "loss": 0.4892,
+      "step": 27996
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.471551748997896e-06,
+      "loss": 0.4513,
+      "step": 27997
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4709749978371835e-06,
+      "loss": 0.4424,
+      "step": 27998
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.470398246676472e-06,
+      "loss": 0.4633,
+      "step": 27999
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.46982149551576e-06,
+      "loss": 0.4749,
+      "step": 28000
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.469244744355048e-06,
+      "loss": 0.4318,
+      "step": 28001
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.468667993194337e-06,
+      "loss": 0.3928,
+      "step": 28002
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4680912420336245e-06,
+      "loss": 0.4002,
+      "step": 28003
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.467514490872913e-06,
+      "loss": 0.4736,
+      "step": 28004
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4669377397122016e-06,
+      "loss": 0.3803,
+      "step": 28005
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.46636098855149e-06,
+      "loss": 0.5072,
+      "step": 28006
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.465784237390779e-06,
+      "loss": 0.4,
+      "step": 28007
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.465207486230066e-06,
+      "loss": 0.4293,
+      "step": 28008
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.464630735069355e-06,
+      "loss": 0.4787,
+      "step": 28009
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4640539839086425e-06,
+      "loss": 0.437,
+      "step": 28010
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.463477232747931e-06,
+      "loss": 0.4133,
+      "step": 28011
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.46290048158722e-06,
+      "loss": 0.3858,
+      "step": 28012
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.462323730426507e-06,
+      "loss": 0.4441,
+      "step": 28013
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.461746979265797e-06,
+      "loss": 0.2965,
+      "step": 28014
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.461170228105084e-06,
+      "loss": 0.4323,
+      "step": 28015
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.460593476944373e-06,
+      "loss": 0.417,
+      "step": 28016
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4600167257836606e-06,
+      "loss": 0.3302,
+      "step": 28017
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.459439974622949e-06,
+      "loss": 0.5174,
+      "step": 28018
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.458863223462238e-06,
+      "loss": 0.5242,
+      "step": 28019
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.458286472301525e-06,
+      "loss": 0.4433,
+      "step": 28020
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.457709721140815e-06,
+      "loss": 0.4157,
+      "step": 28021
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.457132969980102e-06,
+      "loss": 0.4164,
+      "step": 28022
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.456556218819391e-06,
+      "loss": 0.4813,
+      "step": 28023
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4559794676586794e-06,
+      "loss": 0.3316,
+      "step": 28024
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.455402716497967e-06,
+      "loss": 0.3891,
+      "step": 28025
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.454825965337256e-06,
+      "loss": 0.4785,
+      "step": 28026
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.454249214176543e-06,
+      "loss": 0.4877,
+      "step": 28027
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.453672463015833e-06,
+      "loss": 0.3774,
+      "step": 28028
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.45309571185512e-06,
+      "loss": 0.3291,
+      "step": 28029
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.452518960694409e-06,
+      "loss": 0.4193,
+      "step": 28030
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4519422095336975e-06,
+      "loss": 0.473,
+      "step": 28031
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.451365458372985e-06,
+      "loss": 0.2917,
+      "step": 28032
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.450788707212274e-06,
+      "loss": 0.4145,
+      "step": 28033
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.450211956051561e-06,
+      "loss": 0.4244,
+      "step": 28034
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.44963520489085e-06,
+      "loss": 0.4792,
+      "step": 28035
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4490584537301384e-06,
+      "loss": 0.4438,
+      "step": 28036
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.448481702569427e-06,
+      "loss": 0.3948,
+      "step": 28037
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4479049514087155e-06,
+      "loss": 0.3981,
+      "step": 28038
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.447328200248003e-06,
+      "loss": 0.4114,
+      "step": 28039
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.446751449087292e-06,
+      "loss": 0.4247,
+      "step": 28040
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.44617469792658e-06,
+      "loss": 0.5177,
+      "step": 28041
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.445597946765868e-06,
+      "loss": 0.4036,
+      "step": 28042
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4450211956051565e-06,
+      "loss": 0.4719,
+      "step": 28043
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.444444444444444e-06,
+      "loss": 0.3109,
+      "step": 28044
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4438676932837335e-06,
+      "loss": 0.4842,
+      "step": 28045
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.443290942123021e-06,
+      "loss": 0.4027,
+      "step": 28046
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.44271419096231e-06,
+      "loss": 0.3989,
+      "step": 28047
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.442137439801598e-06,
+      "loss": 0.4995,
+      "step": 28048
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.441560688640886e-06,
+      "loss": 0.4421,
+      "step": 28049
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4409839374801745e-06,
+      "loss": 0.4641,
+      "step": 28050
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.440407186319463e-06,
+      "loss": 0.3601,
+      "step": 28051
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4398304351587516e-06,
+      "loss": 0.4223,
+      "step": 28052
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.439253683998039e-06,
+      "loss": 0.444,
+      "step": 28053
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.438676932837328e-06,
+      "loss": 0.4386,
+      "step": 28054
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.438100181676616e-06,
+      "loss": 0.4719,
+      "step": 28055
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.437523430515904e-06,
+      "loss": 0.4414,
+      "step": 28056
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4369466793551925e-06,
+      "loss": 0.2974,
+      "step": 28057
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.436369928194481e-06,
+      "loss": 0.4082,
+      "step": 28058
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.43579317703377e-06,
+      "loss": 0.4065,
+      "step": 28059
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.435216425873057e-06,
+      "loss": 0.3521,
+      "step": 28060
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.434639674712346e-06,
+      "loss": 0.3396,
+      "step": 28061
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.434062923551634e-06,
+      "loss": 0.4101,
+      "step": 28062
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.433486172390922e-06,
+      "loss": 0.5161,
+      "step": 28063
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4329094212302106e-06,
+      "loss": 0.4514,
+      "step": 28064
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.432332670069499e-06,
+      "loss": 0.4392,
+      "step": 28065
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.431755918908788e-06,
+      "loss": 0.4622,
+      "step": 28066
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.431179167748075e-06,
+      "loss": 0.4584,
+      "step": 28067
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.430602416587364e-06,
+      "loss": 0.29,
+      "step": 28068
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.430025665426652e-06,
+      "loss": 0.4417,
+      "step": 28069
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.42944891426594e-06,
+      "loss": 0.319,
+      "step": 28070
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.428872163105229e-06,
+      "loss": 0.3276,
+      "step": 28071
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.428295411944517e-06,
+      "loss": 0.5797,
+      "step": 28072
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.427718660783805e-06,
+      "loss": 0.4315,
+      "step": 28073
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.427141909623093e-06,
+      "loss": 0.2901,
+      "step": 28074
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.426565158462382e-06,
+      "loss": 0.3817,
+      "step": 28075
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.42598840730167e-06,
+      "loss": 0.4934,
+      "step": 28076
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.425411656140958e-06,
+      "loss": 0.3994,
+      "step": 28077
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.424834904980247e-06,
+      "loss": 0.4804,
+      "step": 28078
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.424258153819535e-06,
+      "loss": 0.4076,
+      "step": 28079
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.423681402658823e-06,
+      "loss": 0.3938,
+      "step": 28080
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.423104651498111e-06,
+      "loss": 0.4149,
+      "step": 28081
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4225279003374e-06,
+      "loss": 0.3957,
+      "step": 28082
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4219511491766884e-06,
+      "loss": 0.432,
+      "step": 28083
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.421374398015976e-06,
+      "loss": 0.36,
+      "step": 28084
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.420797646855265e-06,
+      "loss": 0.4705,
+      "step": 28085
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.420220895694553e-06,
+      "loss": 0.2353,
+      "step": 28086
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.419644144533841e-06,
+      "loss": 0.4843,
+      "step": 28087
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.419067393373129e-06,
+      "loss": 0.3742,
+      "step": 28088
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.418490642212418e-06,
+      "loss": 0.3302,
+      "step": 28089
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4179138910517065e-06,
+      "loss": 0.3277,
+      "step": 28090
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.417337139890994e-06,
+      "loss": 0.4545,
+      "step": 28091
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.416760388730283e-06,
+      "loss": 0.3141,
+      "step": 28092
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.416183637569571e-06,
+      "loss": 0.4772,
+      "step": 28093
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.415606886408859e-06,
+      "loss": 0.3611,
+      "step": 28094
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4150301352481474e-06,
+      "loss": 0.3218,
+      "step": 28095
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.414453384087436e-06,
+      "loss": 0.4395,
+      "step": 28096
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4138766329267245e-06,
+      "loss": 0.314,
+      "step": 28097
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.413299881766012e-06,
+      "loss": 0.325,
+      "step": 28098
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.412723130605301e-06,
+      "loss": 0.4534,
+      "step": 28099
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.412146379444589e-06,
+      "loss": 0.3695,
+      "step": 28100
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.411569628283877e-06,
+      "loss": 0.4864,
+      "step": 28101
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4109928771231655e-06,
+      "loss": 0.4921,
+      "step": 28102
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.410416125962454e-06,
+      "loss": 0.5103,
+      "step": 28103
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.409839374801742e-06,
+      "loss": 0.3694,
+      "step": 28104
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.40926262364103e-06,
+      "loss": 0.3167,
+      "step": 28105
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.408685872480319e-06,
+      "loss": 0.4853,
+      "step": 28106
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.408109121319607e-06,
+      "loss": 0.36,
+      "step": 28107
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.407532370158895e-06,
+      "loss": 0.4199,
+      "step": 28108
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4069556189981835e-06,
+      "loss": 0.49,
+      "step": 28109
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.406378867837472e-06,
+      "loss": 0.4545,
+      "step": 28110
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.40580211667676e-06,
+      "loss": 0.4466,
+      "step": 28111
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.405225365516049e-06,
+      "loss": 0.4969,
+      "step": 28112
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.404648614355337e-06,
+      "loss": 0.5595,
+      "step": 28113
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.404071863194625e-06,
+      "loss": 0.252,
+      "step": 28114
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.403495112033913e-06,
+      "loss": 0.4143,
+      "step": 28115
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.4029183608732015e-06,
+      "loss": 0.4772,
+      "step": 28116
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.40234160971249e-06,
+      "loss": 0.4758,
+      "step": 28117
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.401764858551778e-06,
+      "loss": 0.3756,
+      "step": 28118
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.401188107391067e-06,
+      "loss": 0.3459,
+      "step": 28119
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.400611356230355e-06,
+      "loss": 0.446,
+      "step": 28120
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.400034605069643e-06,
+      "loss": 0.4335,
+      "step": 28121
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.399457853908931e-06,
+      "loss": 0.3822,
+      "step": 28122
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.3988811027482196e-06,
+      "loss": 0.3635,
+      "step": 28123
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.398304351587508e-06,
+      "loss": 0.3657,
+      "step": 28124
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.397727600426796e-06,
+      "loss": 0.4059,
+      "step": 28125
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.397150849266084e-06,
+      "loss": 0.3608,
+      "step": 28126
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.396574098105373e-06,
+      "loss": 0.3279,
+      "step": 28127
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.395997346944661e-06,
+      "loss": 0.4781,
+      "step": 28128
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.39542059578395e-06,
+      "loss": 0.4983,
+      "step": 28129
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.394843844623238e-06,
+      "loss": 0.3073,
+      "step": 28130
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.394267093462526e-06,
+      "loss": 0.45,
+      "step": 28131
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.393690342301814e-06,
+      "loss": 0.4725,
+      "step": 28132
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.393113591141102e-06,
+      "loss": 0.3441,
+      "step": 28133
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.392536839980391e-06,
+      "loss": 0.4651,
+      "step": 28134
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.3919600888196786e-06,
+      "loss": 0.3555,
+      "step": 28135
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.391383337658968e-06,
+      "loss": 0.3943,
+      "step": 28136
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.390806586498256e-06,
+      "loss": 0.4705,
+      "step": 28137
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.390229835337544e-06,
+      "loss": 0.4692,
+      "step": 28138
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.389653084176832e-06,
+      "loss": 0.4264,
+      "step": 28139
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.38907633301612e-06,
+      "loss": 0.4436,
+      "step": 28140
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.388499581855409e-06,
+      "loss": 0.4419,
+      "step": 28141
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.387922830694697e-06,
+      "loss": 0.4783,
+      "step": 28142
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.387346079533986e-06,
+      "loss": 0.552,
+      "step": 28143
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.386769328373274e-06,
+      "loss": 0.3213,
+      "step": 28144
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.386192577212562e-06,
+      "loss": 0.3604,
+      "step": 28145
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.385615826051851e-06,
+      "loss": 0.3913,
+      "step": 28146
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.385039074891138e-06,
+      "loss": 0.5064,
+      "step": 28147
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.384462323730427e-06,
+      "loss": 0.4247,
+      "step": 28148
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.383885572569715e-06,
+      "loss": 0.3743,
+      "step": 28149
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.383308821409004e-06,
+      "loss": 0.4486,
+      "step": 28150
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.382732070248292e-06,
+      "loss": 0.4477,
+      "step": 28151
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.38215531908758e-06,
+      "loss": 0.5907,
+      "step": 28152
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 4.381578567926869e-06,
+      "loss": 0.3273,
+      "step": 28153
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3810018167661564e-06,
+      "loss": 0.5128,
+      "step": 28154
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.380425065605445e-06,
+      "loss": 0.4045,
+      "step": 28155
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.379848314444733e-06,
+      "loss": 0.498,
+      "step": 28156
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.379271563284021e-06,
+      "loss": 0.3117,
+      "step": 28157
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.37869481212331e-06,
+      "loss": 0.3801,
+      "step": 28158
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.378118060962598e-06,
+      "loss": 0.4216,
+      "step": 28159
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.377541309801887e-06,
+      "loss": 0.3731,
+      "step": 28160
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3769645586411745e-06,
+      "loss": 0.3619,
+      "step": 28161
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.376387807480463e-06,
+      "loss": 0.3737,
+      "step": 28162
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3758110563197515e-06,
+      "loss": 0.3878,
+      "step": 28163
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.375234305159039e-06,
+      "loss": 0.4383,
+      "step": 28164
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.374657553998328e-06,
+      "loss": 0.4633,
+      "step": 28165
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3740808028376154e-06,
+      "loss": 0.4818,
+      "step": 28166
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.373504051676905e-06,
+      "loss": 0.4128,
+      "step": 28167
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3729273005161925e-06,
+      "loss": 0.3945,
+      "step": 28168
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.372350549355481e-06,
+      "loss": 0.3527,
+      "step": 28169
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3717737981947696e-06,
+      "loss": 0.444,
+      "step": 28170
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.371197047034057e-06,
+      "loss": 0.3761,
+      "step": 28171
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.370620295873346e-06,
+      "loss": 0.4957,
+      "step": 28172
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.370043544712634e-06,
+      "loss": 0.3844,
+      "step": 28173
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.369466793551923e-06,
+      "loss": 0.4277,
+      "step": 28174
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3688900423912105e-06,
+      "loss": 0.3594,
+      "step": 28175
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.368313291230499e-06,
+      "loss": 0.3831,
+      "step": 28176
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.367736540069788e-06,
+      "loss": 0.4666,
+      "step": 28177
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.367159788909075e-06,
+      "loss": 0.4226,
+      "step": 28178
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.366583037748364e-06,
+      "loss": 0.415,
+      "step": 28179
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.366006286587652e-06,
+      "loss": 0.3955,
+      "step": 28180
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.365429535426941e-06,
+      "loss": 0.3764,
+      "step": 28181
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3648527842662285e-06,
+      "loss": 0.428,
+      "step": 28182
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.364276033105517e-06,
+      "loss": 0.447,
+      "step": 28183
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.363699281944806e-06,
+      "loss": 0.4187,
+      "step": 28184
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.363122530784093e-06,
+      "loss": 0.4156,
+      "step": 28185
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.362545779623382e-06,
+      "loss": 0.3547,
+      "step": 28186
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.36196902846267e-06,
+      "loss": 0.4337,
+      "step": 28187
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.361392277301958e-06,
+      "loss": 0.3822,
+      "step": 28188
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.360815526141247e-06,
+      "loss": 0.411,
+      "step": 28189
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.360238774980535e-06,
+      "loss": 0.3834,
+      "step": 28190
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.359662023819824e-06,
+      "loss": 0.3919,
+      "step": 28191
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.359085272659111e-06,
+      "loss": 0.3758,
+      "step": 28192
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3585085214984e-06,
+      "loss": 0.3966,
+      "step": 28193
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.357931770337688e-06,
+      "loss": 0.4347,
+      "step": 28194
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.357355019176976e-06,
+      "loss": 0.3862,
+      "step": 28195
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.356778268016265e-06,
+      "loss": 0.332,
+      "step": 28196
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.356201516855553e-06,
+      "loss": 0.3426,
+      "step": 28197
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.355624765694842e-06,
+      "loss": 0.5041,
+      "step": 28198
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.355048014534129e-06,
+      "loss": 0.5525,
+      "step": 28199
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.354471263373418e-06,
+      "loss": 0.3374,
+      "step": 28200
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3538945122127064e-06,
+      "loss": 0.3985,
+      "step": 28201
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.353317761051994e-06,
+      "loss": 0.4183,
+      "step": 28202
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.352741009891283e-06,
+      "loss": 0.4316,
+      "step": 28203
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.352164258730571e-06,
+      "loss": 0.4786,
+      "step": 28204
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.35158750756986e-06,
+      "loss": 0.3685,
+      "step": 28205
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.351010756409147e-06,
+      "loss": 0.4082,
+      "step": 28206
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.350434005248436e-06,
+      "loss": 0.3877,
+      "step": 28207
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3498572540877245e-06,
+      "loss": 0.4271,
+      "step": 28208
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.349280502927012e-06,
+      "loss": 0.4591,
+      "step": 28209
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.348703751766301e-06,
+      "loss": 0.3655,
+      "step": 28210
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.348127000605589e-06,
+      "loss": 0.3353,
+      "step": 28211
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.347550249444878e-06,
+      "loss": 0.4113,
+      "step": 28212
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.346973498284165e-06,
+      "loss": 0.3075,
+      "step": 28213
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.346396747123454e-06,
+      "loss": 0.2371,
+      "step": 28214
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3458199959627425e-06,
+      "loss": 0.454,
+      "step": 28215
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.34524324480203e-06,
+      "loss": 0.4227,
+      "step": 28216
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.344666493641319e-06,
+      "loss": 0.4837,
+      "step": 28217
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.344089742480607e-06,
+      "loss": 0.3035,
+      "step": 28218
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.343512991319895e-06,
+      "loss": 0.5222,
+      "step": 28219
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3429362401591835e-06,
+      "loss": 0.4732,
+      "step": 28220
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.342359488998472e-06,
+      "loss": 0.432,
+      "step": 28221
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3417827378377605e-06,
+      "loss": 0.4747,
+      "step": 28222
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.341205986677048e-06,
+      "loss": 0.377,
+      "step": 28223
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.340629235516337e-06,
+      "loss": 0.4444,
+      "step": 28224
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.340052484355625e-06,
+      "loss": 0.5168,
+      "step": 28225
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.339475733194913e-06,
+      "loss": 0.3148,
+      "step": 28226
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3388989820342015e-06,
+      "loss": 0.3589,
+      "step": 28227
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.33832223087349e-06,
+      "loss": 0.514,
+      "step": 28228
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3377454797127785e-06,
+      "loss": 0.5112,
+      "step": 28229
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.337168728552066e-06,
+      "loss": 0.3097,
+      "step": 28230
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.336591977391355e-06,
+      "loss": 0.4262,
+      "step": 28231
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.336015226230643e-06,
+      "loss": 0.3907,
+      "step": 28232
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.335438475069931e-06,
+      "loss": 0.3575,
+      "step": 28233
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.33486172390922e-06,
+      "loss": 0.3869,
+      "step": 28234
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.334284972748508e-06,
+      "loss": 0.4335,
+      "step": 28235
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.333708221587797e-06,
+      "loss": 0.4688,
+      "step": 28236
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.333131470427084e-06,
+      "loss": 0.4504,
+      "step": 28237
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.332554719266373e-06,
+      "loss": 0.3808,
+      "step": 28238
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.331977968105661e-06,
+      "loss": 0.27,
+      "step": 28239
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.331401216944949e-06,
+      "loss": 0.3074,
+      "step": 28240
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.330824465784238e-06,
+      "loss": 0.4337,
+      "step": 28241
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.330247714623526e-06,
+      "loss": 0.345,
+      "step": 28242
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.329670963462815e-06,
+      "loss": 0.5829,
+      "step": 28243
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.329094212302102e-06,
+      "loss": 0.5149,
+      "step": 28244
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.328517461141391e-06,
+      "loss": 0.4583,
+      "step": 28245
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.327940709980679e-06,
+      "loss": 0.398,
+      "step": 28246
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.327363958819967e-06,
+      "loss": 0.3469,
+      "step": 28247
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.326787207659256e-06,
+      "loss": 0.4576,
+      "step": 28248
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.326210456498544e-06,
+      "loss": 0.5322,
+      "step": 28249
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.325633705337833e-06,
+      "loss": 0.3905,
+      "step": 28250
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.325056954177121e-06,
+      "loss": 0.341,
+      "step": 28251
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.324480203016409e-06,
+      "loss": 0.2546,
+      "step": 28252
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.323903451855697e-06,
+      "loss": 0.4152,
+      "step": 28253
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.323326700694985e-06,
+      "loss": 0.4656,
+      "step": 28254
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.322749949534274e-06,
+      "loss": 0.4478,
+      "step": 28255
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.322173198373562e-06,
+      "loss": 0.4645,
+      "step": 28256
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.32159644721285e-06,
+      "loss": 0.491,
+      "step": 28257
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.321019696052139e-06,
+      "loss": 0.4196,
+      "step": 28258
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.320442944891427e-06,
+      "loss": 0.3529,
+      "step": 28259
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.319866193730715e-06,
+      "loss": 0.4757,
+      "step": 28260
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.319289442570003e-06,
+      "loss": 0.4485,
+      "step": 28261
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.318712691409292e-06,
+      "loss": 0.3374,
+      "step": 28262
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.31813594024858e-06,
+      "loss": 0.3586,
+      "step": 28263
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.317559189087868e-06,
+      "loss": 0.4105,
+      "step": 28264
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.316982437927157e-06,
+      "loss": 0.4223,
+      "step": 28265
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.316405686766445e-06,
+      "loss": 0.4671,
+      "step": 28266
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3158289356057335e-06,
+      "loss": 0.2641,
+      "step": 28267
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.315252184445022e-06,
+      "loss": 0.3382,
+      "step": 28268
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.31467543328431e-06,
+      "loss": 0.4055,
+      "step": 28269
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.314098682123598e-06,
+      "loss": 0.4345,
+      "step": 28270
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.313521930962886e-06,
+      "loss": 0.3736,
+      "step": 28271
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.312945179802175e-06,
+      "loss": 0.3146,
+      "step": 28272
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.312368428641463e-06,
+      "loss": 0.416,
+      "step": 28273
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3117916774807515e-06,
+      "loss": 0.3387,
+      "step": 28274
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.31121492632004e-06,
+      "loss": 0.351,
+      "step": 28275
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.310638175159328e-06,
+      "loss": 0.3796,
+      "step": 28276
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.310061423998616e-06,
+      "loss": 0.3072,
+      "step": 28277
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.309484672837904e-06,
+      "loss": 0.4824,
+      "step": 28278
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3089079216771924e-06,
+      "loss": 0.3458,
+      "step": 28279
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.308331170516481e-06,
+      "loss": 0.2791,
+      "step": 28280
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3077544193557695e-06,
+      "loss": 0.4389,
+      "step": 28281
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.307177668195058e-06,
+      "loss": 0.3281,
+      "step": 28282
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.306600917034346e-06,
+      "loss": 0.347,
+      "step": 28283
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.306024165873634e-06,
+      "loss": 0.4123,
+      "step": 28284
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.305447414712923e-06,
+      "loss": 0.476,
+      "step": 28285
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3048706635522105e-06,
+      "loss": 0.5466,
+      "step": 28286
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.304293912391499e-06,
+      "loss": 0.3832,
+      "step": 28287
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.303717161230787e-06,
+      "loss": 0.4534,
+      "step": 28288
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.303140410070076e-06,
+      "loss": 0.5191,
+      "step": 28289
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.302563658909364e-06,
+      "loss": 0.3814,
+      "step": 28290
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.301986907748652e-06,
+      "loss": 0.4697,
+      "step": 28291
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.301410156587941e-06,
+      "loss": 0.5961,
+      "step": 28292
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.3008334054272285e-06,
+      "loss": 0.4928,
+      "step": 28293
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.300256654266517e-06,
+      "loss": 0.4184,
+      "step": 28294
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.299679903105805e-06,
+      "loss": 0.3676,
+      "step": 28295
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.299103151945094e-06,
+      "loss": 0.3262,
+      "step": 28296
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.298526400784382e-06,
+      "loss": 0.4489,
+      "step": 28297
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.29794964962367e-06,
+      "loss": 0.5749,
+      "step": 28298
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.297372898462959e-06,
+      "loss": 0.331,
+      "step": 28299
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.2967961473022465e-06,
+      "loss": 0.4218,
+      "step": 28300
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.296219396141535e-06,
+      "loss": 0.5545,
+      "step": 28301
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.295642644980824e-06,
+      "loss": 0.4104,
+      "step": 28302
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.295065893820112e-06,
+      "loss": 0.557,
+      "step": 28303
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.2944891426594e-06,
+      "loss": 0.4556,
+      "step": 28304
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.293912391498688e-06,
+      "loss": 0.384,
+      "step": 28305
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.293335640337977e-06,
+      "loss": 0.2068,
+      "step": 28306
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.2927588891772646e-06,
+      "loss": 0.3604,
+      "step": 28307
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.292182138016553e-06,
+      "loss": 0.4137,
+      "step": 28308
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.291605386855842e-06,
+      "loss": 0.4318,
+      "step": 28309
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.291028635695129e-06,
+      "loss": 0.4572,
+      "step": 28310
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.290451884534418e-06,
+      "loss": 0.4852,
+      "step": 28311
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.289875133373706e-06,
+      "loss": 0.4561,
+      "step": 28312
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.289298382212995e-06,
+      "loss": 0.3644,
+      "step": 28313
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.288721631052283e-06,
+      "loss": 0.4409,
+      "step": 28314
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.288144879891571e-06,
+      "loss": 0.4784,
+      "step": 28315
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.28756812873086e-06,
+      "loss": 0.2832,
+      "step": 28316
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.286991377570147e-06,
+      "loss": 0.3697,
+      "step": 28317
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.286414626409436e-06,
+      "loss": 0.3656,
+      "step": 28318
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.285837875248724e-06,
+      "loss": 0.5571,
+      "step": 28319
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.285261124088013e-06,
+      "loss": 0.3906,
+      "step": 28320
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.284684372927301e-06,
+      "loss": 0.4405,
+      "step": 28321
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.284107621766589e-06,
+      "loss": 0.2531,
+      "step": 28322
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.283530870605878e-06,
+      "loss": 0.4167,
+      "step": 28323
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.282954119445165e-06,
+      "loss": 0.3544,
+      "step": 28324
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.282377368284454e-06,
+      "loss": 0.4699,
+      "step": 28325
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.2818006171237424e-06,
+      "loss": 0.5604,
+      "step": 28326
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.281223865963031e-06,
+      "loss": 0.3402,
+      "step": 28327
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.280647114802319e-06,
+      "loss": 0.2968,
+      "step": 28328
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.280070363641607e-06,
+      "loss": 0.3171,
+      "step": 28329
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.279493612480896e-06,
+      "loss": 0.4071,
+      "step": 28330
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 4.278916861320183e-06,
+      "loss": 0.3342,
+      "step": 28331
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.278340110159472e-06,
+      "loss": 0.4402,
+      "step": 28332
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2777633589987605e-06,
+      "loss": 0.2718,
+      "step": 28333
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.277186607838049e-06,
+      "loss": 0.3733,
+      "step": 28334
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.276609856677337e-06,
+      "loss": 0.462,
+      "step": 28335
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.276033105516625e-06,
+      "loss": 0.4307,
+      "step": 28336
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.275456354355914e-06,
+      "loss": 0.3675,
+      "step": 28337
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2748796031952014e-06,
+      "loss": 0.3246,
+      "step": 28338
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.27430285203449e-06,
+      "loss": 0.4551,
+      "step": 28339
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2737261008737785e-06,
+      "loss": 0.439,
+      "step": 28340
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.273149349713066e-06,
+      "loss": 0.3368,
+      "step": 28341
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.272572598552355e-06,
+      "loss": 0.3851,
+      "step": 28342
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.271995847391643e-06,
+      "loss": 0.3375,
+      "step": 28343
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.271419096230932e-06,
+      "loss": 0.2933,
+      "step": 28344
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2708423450702195e-06,
+      "loss": 0.5966,
+      "step": 28345
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.270265593909508e-06,
+      "loss": 0.3941,
+      "step": 28346
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2696888427487965e-06,
+      "loss": 0.2664,
+      "step": 28347
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.269112091588084e-06,
+      "loss": 0.3598,
+      "step": 28348
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.268535340427373e-06,
+      "loss": 0.3936,
+      "step": 28349
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.267958589266661e-06,
+      "loss": 0.3247,
+      "step": 28350
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.26738183810595e-06,
+      "loss": 0.3927,
+      "step": 28351
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2668050869452375e-06,
+      "loss": 0.3863,
+      "step": 28352
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.266228335784526e-06,
+      "loss": 0.3886,
+      "step": 28353
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2656515846238146e-06,
+      "loss": 0.3421,
+      "step": 28354
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.265074833463102e-06,
+      "loss": 0.3693,
+      "step": 28355
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.264498082302391e-06,
+      "loss": 0.5603,
+      "step": 28356
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.263921331141679e-06,
+      "loss": 0.3397,
+      "step": 28357
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.263344579980968e-06,
+      "loss": 0.405,
+      "step": 28358
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2627678288202555e-06,
+      "loss": 0.3547,
+      "step": 28359
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.262191077659544e-06,
+      "loss": 0.4377,
+      "step": 28360
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.261614326498833e-06,
+      "loss": 0.3562,
+      "step": 28361
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.26103757533812e-06,
+      "loss": 0.4519,
+      "step": 28362
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.26046082417741e-06,
+      "loss": 0.4539,
+      "step": 28363
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.259884073016697e-06,
+      "loss": 0.4693,
+      "step": 28364
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.259307321855986e-06,
+      "loss": 0.3917,
+      "step": 28365
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2587305706952736e-06,
+      "loss": 0.3586,
+      "step": 28366
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.258153819534562e-06,
+      "loss": 0.417,
+      "step": 28367
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.257577068373851e-06,
+      "loss": 0.418,
+      "step": 28368
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.257000317213138e-06,
+      "loss": 0.4458,
+      "step": 28369
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.256423566052427e-06,
+      "loss": 0.3501,
+      "step": 28370
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.255846814891715e-06,
+      "loss": 0.4218,
+      "step": 28371
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.255270063731003e-06,
+      "loss": 0.483,
+      "step": 28372
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2546933125702924e-06,
+      "loss": 0.389,
+      "step": 28373
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.25411656140958e-06,
+      "loss": 0.5211,
+      "step": 28374
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.253539810248869e-06,
+      "loss": 0.5024,
+      "step": 28375
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.252963059088156e-06,
+      "loss": 0.4403,
+      "step": 28376
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.252386307927445e-06,
+      "loss": 0.3835,
+      "step": 28377
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.251809556766733e-06,
+      "loss": 0.4095,
+      "step": 28378
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.251232805606021e-06,
+      "loss": 0.4009,
+      "step": 28379
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2506560544453105e-06,
+      "loss": 0.3879,
+      "step": 28380
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.250079303284598e-06,
+      "loss": 0.4306,
+      "step": 28381
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.249502552123887e-06,
+      "loss": 0.4418,
+      "step": 28382
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.248925800963174e-06,
+      "loss": 0.3358,
+      "step": 28383
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.248349049802463e-06,
+      "loss": 0.3299,
+      "step": 28384
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2477722986417514e-06,
+      "loss": 0.4582,
+      "step": 28385
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.247195547481039e-06,
+      "loss": 0.4218,
+      "step": 28386
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2466187963203285e-06,
+      "loss": 0.4265,
+      "step": 28387
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.246042045159616e-06,
+      "loss": 0.3261,
+      "step": 28388
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.245465293998905e-06,
+      "loss": 0.4474,
+      "step": 28389
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.244888542838193e-06,
+      "loss": 0.4141,
+      "step": 28390
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.244311791677481e-06,
+      "loss": 0.4394,
+      "step": 28391
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2437350405167695e-06,
+      "loss": 0.4328,
+      "step": 28392
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.243158289356057e-06,
+      "loss": 0.3778,
+      "step": 28393
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2425815381953465e-06,
+      "loss": 0.2878,
+      "step": 28394
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.242004787034634e-06,
+      "loss": 0.4345,
+      "step": 28395
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.241428035873923e-06,
+      "loss": 0.4477,
+      "step": 28396
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.240851284713211e-06,
+      "loss": 0.3376,
+      "step": 28397
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.240274533552499e-06,
+      "loss": 0.533,
+      "step": 28398
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2396977823917875e-06,
+      "loss": 0.3946,
+      "step": 28399
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.239121031231075e-06,
+      "loss": 0.3412,
+      "step": 28400
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.238544280070364e-06,
+      "loss": 0.4396,
+      "step": 28401
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.237967528909652e-06,
+      "loss": 0.3672,
+      "step": 28402
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.23739077774894e-06,
+      "loss": 0.3791,
+      "step": 28403
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.236814026588229e-06,
+      "loss": 0.5383,
+      "step": 28404
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.236237275427517e-06,
+      "loss": 0.3368,
+      "step": 28405
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2356605242668055e-06,
+      "loss": 0.4323,
+      "step": 28406
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.235083773106094e-06,
+      "loss": 0.3893,
+      "step": 28407
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.234507021945382e-06,
+      "loss": 0.3635,
+      "step": 28408
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.23393027078467e-06,
+      "loss": 0.3729,
+      "step": 28409
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.233353519623958e-06,
+      "loss": 0.4762,
+      "step": 28410
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.232776768463247e-06,
+      "loss": 0.2882,
+      "step": 28411
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.232200017302535e-06,
+      "loss": 0.4222,
+      "step": 28412
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2316232661418236e-06,
+      "loss": 0.3333,
+      "step": 28413
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.231046514981112e-06,
+      "loss": 0.5025,
+      "step": 28414
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2304697638204e-06,
+      "loss": 0.3456,
+      "step": 28415
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.229893012659688e-06,
+      "loss": 0.3503,
+      "step": 28416
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.229316261498976e-06,
+      "loss": 0.2926,
+      "step": 28417
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.228739510338265e-06,
+      "loss": 0.4136,
+      "step": 28418
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.228162759177553e-06,
+      "loss": 0.4557,
+      "step": 28419
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.227586008016842e-06,
+      "loss": 0.5248,
+      "step": 28420
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.22700925685613e-06,
+      "loss": 0.326,
+      "step": 28421
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.226432505695418e-06,
+      "loss": 0.4622,
+      "step": 28422
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.225855754534706e-06,
+      "loss": 0.3963,
+      "step": 28423
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.225279003373995e-06,
+      "loss": 0.4552,
+      "step": 28424
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.224702252213283e-06,
+      "loss": 0.435,
+      "step": 28425
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.224125501052571e-06,
+      "loss": 0.4661,
+      "step": 28426
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.22354874989186e-06,
+      "loss": 0.4037,
+      "step": 28427
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.222971998731148e-06,
+      "loss": 0.403,
+      "step": 28428
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.222395247570436e-06,
+      "loss": 0.339,
+      "step": 28429
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.221818496409724e-06,
+      "loss": 0.3829,
+      "step": 28430
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.221241745249013e-06,
+      "loss": 0.3834,
+      "step": 28431
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.220664994088301e-06,
+      "loss": 0.3985,
+      "step": 28432
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.220088242927589e-06,
+      "loss": 0.2806,
+      "step": 28433
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.219511491766878e-06,
+      "loss": 0.3239,
+      "step": 28434
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.218934740606166e-06,
+      "loss": 0.4381,
+      "step": 28435
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.218357989445454e-06,
+      "loss": 0.4975,
+      "step": 28436
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.217781238284742e-06,
+      "loss": 0.3594,
+      "step": 28437
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.217204487124031e-06,
+      "loss": 0.4258,
+      "step": 28438
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.216627735963319e-06,
+      "loss": 0.4044,
+      "step": 28439
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.216050984802607e-06,
+      "loss": 0.4796,
+      "step": 28440
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.215474233641896e-06,
+      "loss": 0.4171,
+      "step": 28441
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.214897482481184e-06,
+      "loss": 0.4344,
+      "step": 28442
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.214320731320472e-06,
+      "loss": 0.5832,
+      "step": 28443
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2137439801597604e-06,
+      "loss": 0.3776,
+      "step": 28444
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.213167228999049e-06,
+      "loss": 0.4835,
+      "step": 28445
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.212590477838337e-06,
+      "loss": 0.3914,
+      "step": 28446
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.212013726677625e-06,
+      "loss": 0.4825,
+      "step": 28447
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.211436975516914e-06,
+      "loss": 0.3937,
+      "step": 28448
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.210860224356202e-06,
+      "loss": 0.3331,
+      "step": 28449
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.21028347319549e-06,
+      "loss": 0.474,
+      "step": 28450
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2097067220347785e-06,
+      "loss": 0.3114,
+      "step": 28451
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.209129970874067e-06,
+      "loss": 0.3759,
+      "step": 28452
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.208553219713355e-06,
+      "loss": 0.4054,
+      "step": 28453
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.207976468552643e-06,
+      "loss": 0.4303,
+      "step": 28454
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.207399717391932e-06,
+      "loss": 0.3854,
+      "step": 28455
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.20682296623122e-06,
+      "loss": 0.3131,
+      "step": 28456
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.206246215070508e-06,
+      "loss": 0.3894,
+      "step": 28457
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2056694639097965e-06,
+      "loss": 0.3416,
+      "step": 28458
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.205092712749085e-06,
+      "loss": 0.3829,
+      "step": 28459
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.204515961588373e-06,
+      "loss": 0.4486,
+      "step": 28460
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.203939210427661e-06,
+      "loss": 0.3499,
+      "step": 28461
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.20336245926695e-06,
+      "loss": 0.4388,
+      "step": 28462
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2027857081062375e-06,
+      "loss": 0.4165,
+      "step": 28463
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.202208956945526e-06,
+      "loss": 0.3518,
+      "step": 28464
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.2016322057848145e-06,
+      "loss": 0.3612,
+      "step": 28465
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.201055454624103e-06,
+      "loss": 0.5096,
+      "step": 28466
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.200478703463391e-06,
+      "loss": 0.4046,
+      "step": 28467
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.199901952302679e-06,
+      "loss": 0.31,
+      "step": 28468
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.199325201141968e-06,
+      "loss": 0.3978,
+      "step": 28469
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1987484499812555e-06,
+      "loss": 0.5149,
+      "step": 28470
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.198171698820544e-06,
+      "loss": 0.4192,
+      "step": 28471
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1975949476598326e-06,
+      "loss": 0.4463,
+      "step": 28472
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.197018196499121e-06,
+      "loss": 0.4067,
+      "step": 28473
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.196441445338409e-06,
+      "loss": 0.3182,
+      "step": 28474
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.195864694177697e-06,
+      "loss": 0.371,
+      "step": 28475
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.195287943016986e-06,
+      "loss": 0.3182,
+      "step": 28476
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1947111918562735e-06,
+      "loss": 0.4899,
+      "step": 28477
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.194134440695562e-06,
+      "loss": 0.47,
+      "step": 28478
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.193557689534851e-06,
+      "loss": 0.3956,
+      "step": 28479
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.192980938374139e-06,
+      "loss": 0.4875,
+      "step": 28480
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.192404187213427e-06,
+      "loss": 0.3777,
+      "step": 28481
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.191827436052715e-06,
+      "loss": 0.2508,
+      "step": 28482
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.191250684892004e-06,
+      "loss": 0.4521,
+      "step": 28483
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1906739337312916e-06,
+      "loss": 0.4537,
+      "step": 28484
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.19009718257058e-06,
+      "loss": 0.2527,
+      "step": 28485
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.189520431409869e-06,
+      "loss": 0.3796,
+      "step": 28486
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.188943680249157e-06,
+      "loss": 0.3396,
+      "step": 28487
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.188366929088445e-06,
+      "loss": 0.3365,
+      "step": 28488
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.187790177927733e-06,
+      "loss": 0.4098,
+      "step": 28489
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.187213426767022e-06,
+      "loss": 0.3964,
+      "step": 28490
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.18663667560631e-06,
+      "loss": 0.3696,
+      "step": 28491
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.186059924445598e-06,
+      "loss": 0.4208,
+      "step": 28492
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.185483173284887e-06,
+      "loss": 0.3096,
+      "step": 28493
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.184906422124174e-06,
+      "loss": 0.324,
+      "step": 28494
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.184329670963464e-06,
+      "loss": 0.3352,
+      "step": 28495
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.183752919802751e-06,
+      "loss": 0.5111,
+      "step": 28496
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.18317616864204e-06,
+      "loss": 0.3702,
+      "step": 28497
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.182599417481328e-06,
+      "loss": 0.4831,
+      "step": 28498
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.182022666320616e-06,
+      "loss": 0.4419,
+      "step": 28499
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.181445915159905e-06,
+      "loss": 0.3401,
+      "step": 28500
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.180869163999192e-06,
+      "loss": 0.357,
+      "step": 28501
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.180292412838482e-06,
+      "loss": 0.3477,
+      "step": 28502
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1797156616777694e-06,
+      "loss": 0.404,
+      "step": 28503
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.179138910517058e-06,
+      "loss": 0.4729,
+      "step": 28504
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.178562159356346e-06,
+      "loss": 0.4453,
+      "step": 28505
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.177985408195634e-06,
+      "loss": 0.4441,
+      "step": 28506
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.177408657034923e-06,
+      "loss": 0.4587,
+      "step": 28507
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.17683190587421e-06,
+      "loss": 0.3635,
+      "step": 28508
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1762551547135e-06,
+      "loss": 0.36,
+      "step": 28509
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 4.1756784035527875e-06,
+      "loss": 0.3005,
+      "step": 28510
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.175101652392076e-06,
+      "loss": 0.5002,
+      "step": 28511
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1745249012313645e-06,
+      "loss": 0.234,
+      "step": 28512
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.173948150070652e-06,
+      "loss": 0.4146,
+      "step": 28513
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.173371398909941e-06,
+      "loss": 0.4195,
+      "step": 28514
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1727946477492284e-06,
+      "loss": 0.3617,
+      "step": 28515
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.172217896588517e-06,
+      "loss": 0.3706,
+      "step": 28516
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1716411454278055e-06,
+      "loss": 0.5282,
+      "step": 28517
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.171064394267094e-06,
+      "loss": 0.4133,
+      "step": 28518
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1704876431063826e-06,
+      "loss": 0.3604,
+      "step": 28519
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.16991089194567e-06,
+      "loss": 0.3842,
+      "step": 28520
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.169334140784959e-06,
+      "loss": 0.4063,
+      "step": 28521
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1687573896242465e-06,
+      "loss": 0.4707,
+      "step": 28522
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.168180638463535e-06,
+      "loss": 0.4878,
+      "step": 28523
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1676038873028235e-06,
+      "loss": 0.5104,
+      "step": 28524
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.167027136142111e-06,
+      "loss": 0.4122,
+      "step": 28525
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.166450384981401e-06,
+      "loss": 0.4318,
+      "step": 28526
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.165873633820688e-06,
+      "loss": 0.4632,
+      "step": 28527
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.165296882659977e-06,
+      "loss": 0.4238,
+      "step": 28528
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.164720131499265e-06,
+      "loss": 0.3216,
+      "step": 28529
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.164143380338553e-06,
+      "loss": 0.3837,
+      "step": 28530
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1635666291778416e-06,
+      "loss": 0.6191,
+      "step": 28531
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.162989878017129e-06,
+      "loss": 0.4595,
+      "step": 28532
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.162413126856419e-06,
+      "loss": 0.4248,
+      "step": 28533
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.161836375695706e-06,
+      "loss": 0.4237,
+      "step": 28534
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.161259624534995e-06,
+      "loss": 0.4144,
+      "step": 28535
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.160682873374283e-06,
+      "loss": 0.4204,
+      "step": 28536
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.160106122213571e-06,
+      "loss": 0.3678,
+      "step": 28537
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.15952937105286e-06,
+      "loss": 0.3624,
+      "step": 28538
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.158952619892147e-06,
+      "loss": 0.3834,
+      "step": 28539
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.158375868731437e-06,
+      "loss": 0.4397,
+      "step": 28540
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.157799117570724e-06,
+      "loss": 0.4995,
+      "step": 28541
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.157222366410013e-06,
+      "loss": 0.4794,
+      "step": 28542
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.156645615249301e-06,
+      "loss": 0.454,
+      "step": 28543
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.156068864088589e-06,
+      "loss": 0.38,
+      "step": 28544
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.155492112927878e-06,
+      "loss": 0.434,
+      "step": 28545
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.154915361767166e-06,
+      "loss": 0.3273,
+      "step": 28546
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.154338610606455e-06,
+      "loss": 0.3516,
+      "step": 28547
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.153761859445742e-06,
+      "loss": 0.463,
+      "step": 28548
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.153185108285031e-06,
+      "loss": 0.4827,
+      "step": 28549
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1526083571243194e-06,
+      "loss": 0.3183,
+      "step": 28550
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.152031605963607e-06,
+      "loss": 0.3998,
+      "step": 28551
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.151454854802896e-06,
+      "loss": 0.3418,
+      "step": 28552
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.150878103642184e-06,
+      "loss": 0.3474,
+      "step": 28553
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.150301352481472e-06,
+      "loss": 0.533,
+      "step": 28554
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.14972460132076e-06,
+      "loss": 0.4805,
+      "step": 28555
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.149147850160049e-06,
+      "loss": 0.3658,
+      "step": 28556
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1485710989993375e-06,
+      "loss": 0.4858,
+      "step": 28557
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.147994347838625e-06,
+      "loss": 0.3981,
+      "step": 28558
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.147417596677914e-06,
+      "loss": 0.4991,
+      "step": 28559
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.146840845517202e-06,
+      "loss": 0.3931,
+      "step": 28560
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.14626409435649e-06,
+      "loss": 0.4627,
+      "step": 28561
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1456873431957784e-06,
+      "loss": 0.4663,
+      "step": 28562
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.145110592035067e-06,
+      "loss": 0.4133,
+      "step": 28563
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1445338408743555e-06,
+      "loss": 0.3039,
+      "step": 28564
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.143957089713643e-06,
+      "loss": 0.5968,
+      "step": 28565
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.143380338552932e-06,
+      "loss": 0.3435,
+      "step": 28566
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.14280358739222e-06,
+      "loss": 0.356,
+      "step": 28567
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.142226836231508e-06,
+      "loss": 0.4157,
+      "step": 28568
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1416500850707965e-06,
+      "loss": 0.4429,
+      "step": 28569
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.141073333910085e-06,
+      "loss": 0.3275,
+      "step": 28570
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1404965827493735e-06,
+      "loss": 0.3385,
+      "step": 28571
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.139919831588661e-06,
+      "loss": 0.4736,
+      "step": 28572
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.13934308042795e-06,
+      "loss": 0.4285,
+      "step": 28573
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.138766329267238e-06,
+      "loss": 0.4486,
+      "step": 28574
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.138189578106526e-06,
+      "loss": 0.5521,
+      "step": 28575
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1376128269458145e-06,
+      "loss": 0.4138,
+      "step": 28576
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.137036075785103e-06,
+      "loss": 0.4284,
+      "step": 28577
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1364593246243916e-06,
+      "loss": 0.5066,
+      "step": 28578
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.135882573463679e-06,
+      "loss": 0.385,
+      "step": 28579
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.135305822302968e-06,
+      "loss": 0.4383,
+      "step": 28580
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.134729071142256e-06,
+      "loss": 0.4121,
+      "step": 28581
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.134152319981544e-06,
+      "loss": 0.3468,
+      "step": 28582
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1335755688208325e-06,
+      "loss": 0.3936,
+      "step": 28583
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.132998817660121e-06,
+      "loss": 0.4339,
+      "step": 28584
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.132422066499409e-06,
+      "loss": 0.4216,
+      "step": 28585
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.131845315338697e-06,
+      "loss": 0.4818,
+      "step": 28586
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.131268564177986e-06,
+      "loss": 0.3559,
+      "step": 28587
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.130691813017274e-06,
+      "loss": 0.3909,
+      "step": 28588
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.130115061856562e-06,
+      "loss": 0.3657,
+      "step": 28589
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1295383106958506e-06,
+      "loss": 0.4133,
+      "step": 28590
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.128961559535139e-06,
+      "loss": 0.408,
+      "step": 28591
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.128384808374427e-06,
+      "loss": 0.365,
+      "step": 28592
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.127808057213715e-06,
+      "loss": 0.3213,
+      "step": 28593
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.127231306053004e-06,
+      "loss": 0.5076,
+      "step": 28594
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.126654554892292e-06,
+      "loss": 0.4035,
+      "step": 28595
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.12607780373158e-06,
+      "loss": 0.4817,
+      "step": 28596
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.125501052570869e-06,
+      "loss": 0.3535,
+      "step": 28597
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.124924301410157e-06,
+      "loss": 0.4116,
+      "step": 28598
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.124347550249445e-06,
+      "loss": 0.3569,
+      "step": 28599
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.123770799088733e-06,
+      "loss": 0.4701,
+      "step": 28600
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.123194047928022e-06,
+      "loss": 0.521,
+      "step": 28601
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.12261729676731e-06,
+      "loss": 0.4224,
+      "step": 28602
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.122040545606598e-06,
+      "loss": 0.4752,
+      "step": 28603
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.121463794445887e-06,
+      "loss": 0.5122,
+      "step": 28604
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.120887043285175e-06,
+      "loss": 0.4929,
+      "step": 28605
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.120310292124463e-06,
+      "loss": 0.4011,
+      "step": 28606
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.119733540963751e-06,
+      "loss": 0.3592,
+      "step": 28607
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.11915678980304e-06,
+      "loss": 0.31,
+      "step": 28608
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1185800386423284e-06,
+      "loss": 0.4874,
+      "step": 28609
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.118003287481616e-06,
+      "loss": 0.521,
+      "step": 28610
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.117426536320905e-06,
+      "loss": 0.3718,
+      "step": 28611
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.116849785160193e-06,
+      "loss": 0.3697,
+      "step": 28612
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.116273033999481e-06,
+      "loss": 0.3404,
+      "step": 28613
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.115696282838769e-06,
+      "loss": 0.3985,
+      "step": 28614
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.115119531678058e-06,
+      "loss": 0.4244,
+      "step": 28615
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.114542780517346e-06,
+      "loss": 0.4377,
+      "step": 28616
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.113966029356635e-06,
+      "loss": 0.3244,
+      "step": 28617
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.113389278195923e-06,
+      "loss": 0.3411,
+      "step": 28618
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.112812527035211e-06,
+      "loss": 0.4052,
+      "step": 28619
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.112235775874499e-06,
+      "loss": 0.4037,
+      "step": 28620
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1116590247137874e-06,
+      "loss": 0.37,
+      "step": 28621
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.111082273553076e-06,
+      "loss": 0.3777,
+      "step": 28622
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.110505522392364e-06,
+      "loss": 0.4913,
+      "step": 28623
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.109928771231653e-06,
+      "loss": 0.3709,
+      "step": 28624
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.109352020070941e-06,
+      "loss": 0.4897,
+      "step": 28625
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.108775268910229e-06,
+      "loss": 0.3395,
+      "step": 28626
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.108198517749517e-06,
+      "loss": 0.4396,
+      "step": 28627
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1076217665888055e-06,
+      "loss": 0.4744,
+      "step": 28628
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.107045015428094e-06,
+      "loss": 0.356,
+      "step": 28629
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.106468264267382e-06,
+      "loss": 0.4919,
+      "step": 28630
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.105891513106671e-06,
+      "loss": 0.369,
+      "step": 28631
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.105314761945959e-06,
+      "loss": 0.4196,
+      "step": 28632
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.104738010785247e-06,
+      "loss": 0.3426,
+      "step": 28633
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.104161259624536e-06,
+      "loss": 0.416,
+      "step": 28634
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1035845084638235e-06,
+      "loss": 0.4578,
+      "step": 28635
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.103007757303112e-06,
+      "loss": 0.3047,
+      "step": 28636
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.1024310061424e-06,
+      "loss": 0.3675,
+      "step": 28637
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.101854254981688e-06,
+      "loss": 0.4822,
+      "step": 28638
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.101277503820977e-06,
+      "loss": 0.3888,
+      "step": 28639
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.100700752660265e-06,
+      "loss": 0.3998,
+      "step": 28640
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.100124001499554e-06,
+      "loss": 0.3679,
+      "step": 28641
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0995472503388415e-06,
+      "loss": 0.3805,
+      "step": 28642
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.09897049917813e-06,
+      "loss": 0.4353,
+      "step": 28643
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.098393748017418e-06,
+      "loss": 0.3432,
+      "step": 28644
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.097816996856706e-06,
+      "loss": 0.3778,
+      "step": 28645
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.097240245695995e-06,
+      "loss": 0.3756,
+      "step": 28646
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0966634945352825e-06,
+      "loss": 0.3498,
+      "step": 28647
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.096086743374572e-06,
+      "loss": 0.3496,
+      "step": 28648
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0955099922138596e-06,
+      "loss": 0.4363,
+      "step": 28649
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.094933241053148e-06,
+      "loss": 0.3964,
+      "step": 28650
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.094356489892437e-06,
+      "loss": 0.4238,
+      "step": 28651
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.093779738731724e-06,
+      "loss": 0.4551,
+      "step": 28652
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.093202987571013e-06,
+      "loss": 0.373,
+      "step": 28653
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0926262364103005e-06,
+      "loss": 0.4572,
+      "step": 28654
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.09204948524959e-06,
+      "loss": 0.4442,
+      "step": 28655
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.091472734088878e-06,
+      "loss": 0.4444,
+      "step": 28656
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.090895982928166e-06,
+      "loss": 0.327,
+      "step": 28657
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.090319231767455e-06,
+      "loss": 0.3758,
+      "step": 28658
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.089742480606742e-06,
+      "loss": 0.4075,
+      "step": 28659
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.089165729446031e-06,
+      "loss": 0.3799,
+      "step": 28660
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0885889782853186e-06,
+      "loss": 0.3536,
+      "step": 28661
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.088012227124608e-06,
+      "loss": 0.4169,
+      "step": 28662
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.087435475963896e-06,
+      "loss": 0.4162,
+      "step": 28663
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.086858724803184e-06,
+      "loss": 0.385,
+      "step": 28664
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.086281973642473e-06,
+      "loss": 0.3534,
+      "step": 28665
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.08570522248176e-06,
+      "loss": 0.414,
+      "step": 28666
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.085128471321049e-06,
+      "loss": 0.4497,
+      "step": 28667
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0845517201603374e-06,
+      "loss": 0.4024,
+      "step": 28668
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.083974968999625e-06,
+      "loss": 0.3441,
+      "step": 28669
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.083398217838914e-06,
+      "loss": 0.535,
+      "step": 28670
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.082821466678202e-06,
+      "loss": 0.2483,
+      "step": 28671
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.082244715517491e-06,
+      "loss": 0.4553,
+      "step": 28672
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.081667964356778e-06,
+      "loss": 0.3094,
+      "step": 28673
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.081091213196067e-06,
+      "loss": 0.4793,
+      "step": 28674
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0805144620353555e-06,
+      "loss": 0.3617,
+      "step": 28675
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.079937710874643e-06,
+      "loss": 0.418,
+      "step": 28676
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.079360959713932e-06,
+      "loss": 0.4213,
+      "step": 28677
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.078784208553219e-06,
+      "loss": 0.3467,
+      "step": 28678
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.078207457392509e-06,
+      "loss": 0.4225,
+      "step": 28679
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0776307062317964e-06,
+      "loss": 0.3749,
+      "step": 28680
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.077053955071085e-06,
+      "loss": 0.4222,
+      "step": 28681
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0764772039103735e-06,
+      "loss": 0.3916,
+      "step": 28682
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.075900452749661e-06,
+      "loss": 0.4903,
+      "step": 28683
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.07532370158895e-06,
+      "loss": 0.2967,
+      "step": 28684
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.074746950428238e-06,
+      "loss": 0.3463,
+      "step": 28685
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.074170199267527e-06,
+      "loss": 0.4774,
+      "step": 28686
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0735934481068145e-06,
+      "loss": 0.3454,
+      "step": 28687
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.073016696946103e-06,
+      "loss": 0.412,
+      "step": 28688
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 4.0724399457853915e-06,
+      "loss": 0.377,
+      "step": 28689
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.071863194624679e-06,
+      "loss": 0.4059,
+      "step": 28690
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.071286443463968e-06,
+      "loss": 0.4288,
+      "step": 28691
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.070709692303256e-06,
+      "loss": 0.454,
+      "step": 28692
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.070132941142545e-06,
+      "loss": 0.3225,
+      "step": 28693
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0695561899818325e-06,
+      "loss": 0.3817,
+      "step": 28694
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.068979438821121e-06,
+      "loss": 0.3189,
+      "step": 28695
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0684026876604096e-06,
+      "loss": 0.467,
+      "step": 28696
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.067825936499697e-06,
+      "loss": 0.4876,
+      "step": 28697
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.067249185338986e-06,
+      "loss": 0.4369,
+      "step": 28698
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.066672434178274e-06,
+      "loss": 0.4035,
+      "step": 28699
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.066095683017562e-06,
+      "loss": 0.4327,
+      "step": 28700
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0655189318568505e-06,
+      "loss": 0.3446,
+      "step": 28701
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.064942180696139e-06,
+      "loss": 0.4381,
+      "step": 28702
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.064365429535428e-06,
+      "loss": 0.4303,
+      "step": 28703
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.063788678374715e-06,
+      "loss": 0.3919,
+      "step": 28704
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.063211927214004e-06,
+      "loss": 0.4601,
+      "step": 28705
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.062635176053292e-06,
+      "loss": 0.3573,
+      "step": 28706
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.06205842489258e-06,
+      "loss": 0.4378,
+      "step": 28707
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0614816737318686e-06,
+      "loss": 0.4882,
+      "step": 28708
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.060904922571157e-06,
+      "loss": 0.384,
+      "step": 28709
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.060328171410446e-06,
+      "loss": 0.4242,
+      "step": 28710
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.059751420249733e-06,
+      "loss": 0.4381,
+      "step": 28711
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.059174669089022e-06,
+      "loss": 0.4498,
+      "step": 28712
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.05859791792831e-06,
+      "loss": 0.455,
+      "step": 28713
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.058021166767598e-06,
+      "loss": 0.4385,
+      "step": 28714
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.057444415606887e-06,
+      "loss": 0.3729,
+      "step": 28715
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.056867664446175e-06,
+      "loss": 0.4287,
+      "step": 28716
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.056290913285464e-06,
+      "loss": 0.3494,
+      "step": 28717
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.055714162124751e-06,
+      "loss": 0.368,
+      "step": 28718
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.05513741096404e-06,
+      "loss": 0.4335,
+      "step": 28719
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.054560659803328e-06,
+      "loss": 0.3067,
+      "step": 28720
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.053983908642616e-06,
+      "loss": 0.3106,
+      "step": 28721
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.053407157481905e-06,
+      "loss": 0.4419,
+      "step": 28722
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.052830406321193e-06,
+      "loss": 0.5009,
+      "step": 28723
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.052253655160482e-06,
+      "loss": 0.5028,
+      "step": 28724
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.051676903999769e-06,
+      "loss": 0.3227,
+      "step": 28725
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.051100152839058e-06,
+      "loss": 0.2918,
+      "step": 28726
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0505234016783464e-06,
+      "loss": 0.5061,
+      "step": 28727
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.049946650517634e-06,
+      "loss": 0.4978,
+      "step": 28728
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.049369899356923e-06,
+      "loss": 0.4606,
+      "step": 28729
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.048793148196211e-06,
+      "loss": 0.4077,
+      "step": 28730
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0482163970355e-06,
+      "loss": 0.4537,
+      "step": 28731
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.047639645874787e-06,
+      "loss": 0.3731,
+      "step": 28732
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.047062894714076e-06,
+      "loss": 0.4586,
+      "step": 28733
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0464861435533645e-06,
+      "loss": 0.4036,
+      "step": 28734
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.045909392392652e-06,
+      "loss": 0.4324,
+      "step": 28735
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.045332641231941e-06,
+      "loss": 0.4337,
+      "step": 28736
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.044755890071229e-06,
+      "loss": 0.2824,
+      "step": 28737
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.044179138910517e-06,
+      "loss": 0.4087,
+      "step": 28738
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0436023877498054e-06,
+      "loss": 0.4399,
+      "step": 28739
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.043025636589094e-06,
+      "loss": 0.3952,
+      "step": 28740
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0424488854283825e-06,
+      "loss": 0.4775,
+      "step": 28741
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.04187213426767e-06,
+      "loss": 0.3789,
+      "step": 28742
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.041295383106959e-06,
+      "loss": 0.4571,
+      "step": 28743
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.040718631946247e-06,
+      "loss": 0.323,
+      "step": 28744
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.040141880785535e-06,
+      "loss": 0.4119,
+      "step": 28745
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.039565129624824e-06,
+      "loss": 0.3638,
+      "step": 28746
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.038988378464112e-06,
+      "loss": 0.5232,
+      "step": 28747
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0384116273034005e-06,
+      "loss": 0.4245,
+      "step": 28748
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.037834876142688e-06,
+      "loss": 0.3565,
+      "step": 28749
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.037258124981977e-06,
+      "loss": 0.3102,
+      "step": 28750
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.036681373821265e-06,
+      "loss": 0.2912,
+      "step": 28751
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.036104622660553e-06,
+      "loss": 0.3869,
+      "step": 28752
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.035527871499842e-06,
+      "loss": 0.4429,
+      "step": 28753
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.03495112033913e-06,
+      "loss": 0.4247,
+      "step": 28754
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0343743691784186e-06,
+      "loss": 0.4042,
+      "step": 28755
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.033797618017707e-06,
+      "loss": 0.4151,
+      "step": 28756
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.033220866856995e-06,
+      "loss": 0.4657,
+      "step": 28757
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.032644115696283e-06,
+      "loss": 0.4853,
+      "step": 28758
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.032067364535571e-06,
+      "loss": 0.4189,
+      "step": 28759
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0314906133748595e-06,
+      "loss": 0.4545,
+      "step": 28760
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.030913862214148e-06,
+      "loss": 0.4732,
+      "step": 28761
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.030337111053437e-06,
+      "loss": 0.5799,
+      "step": 28762
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.029760359892725e-06,
+      "loss": 0.4867,
+      "step": 28763
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.029183608732013e-06,
+      "loss": 0.335,
+      "step": 28764
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.028606857571301e-06,
+      "loss": 0.3118,
+      "step": 28765
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.028030106410589e-06,
+      "loss": 0.4493,
+      "step": 28766
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0274533552498775e-06,
+      "loss": 0.4222,
+      "step": 28767
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.026876604089166e-06,
+      "loss": 0.5046,
+      "step": 28768
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.026299852928454e-06,
+      "loss": 0.4162,
+      "step": 28769
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.025723101767743e-06,
+      "loss": 0.4088,
+      "step": 28770
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.025146350607031e-06,
+      "loss": 0.444,
+      "step": 28771
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.024569599446319e-06,
+      "loss": 0.3668,
+      "step": 28772
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.023992848285608e-06,
+      "loss": 0.4221,
+      "step": 28773
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.023416097124896e-06,
+      "loss": 0.3534,
+      "step": 28774
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.022839345964184e-06,
+      "loss": 0.3836,
+      "step": 28775
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.022262594803472e-06,
+      "loss": 0.4754,
+      "step": 28776
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.021685843642761e-06,
+      "loss": 0.4668,
+      "step": 28777
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.021109092482049e-06,
+      "loss": 0.5646,
+      "step": 28778
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.020532341321337e-06,
+      "loss": 0.4719,
+      "step": 28779
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.019955590160626e-06,
+      "loss": 0.4308,
+      "step": 28780
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.019378838999914e-06,
+      "loss": 0.3922,
+      "step": 28781
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.018802087839202e-06,
+      "loss": 0.3611,
+      "step": 28782
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.01822533667849e-06,
+      "loss": 0.4802,
+      "step": 28783
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.017648585517779e-06,
+      "loss": 0.4618,
+      "step": 28784
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.017071834357067e-06,
+      "loss": 0.4211,
+      "step": 28785
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.016495083196355e-06,
+      "loss": 0.2965,
+      "step": 28786
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.015918332035644e-06,
+      "loss": 0.4575,
+      "step": 28787
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.015341580874932e-06,
+      "loss": 0.4254,
+      "step": 28788
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.01476482971422e-06,
+      "loss": 0.4972,
+      "step": 28789
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.014188078553509e-06,
+      "loss": 0.3638,
+      "step": 28790
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.013611327392796e-06,
+      "loss": 0.392,
+      "step": 28791
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.013034576232085e-06,
+      "loss": 0.4776,
+      "step": 28792
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0124578250713735e-06,
+      "loss": 0.4031,
+      "step": 28793
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.011881073910662e-06,
+      "loss": 0.3729,
+      "step": 28794
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.01130432274995e-06,
+      "loss": 0.449,
+      "step": 28795
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.010727571589238e-06,
+      "loss": 0.4696,
+      "step": 28796
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.010150820428527e-06,
+      "loss": 0.5178,
+      "step": 28797
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.009574069267814e-06,
+      "loss": 0.3231,
+      "step": 28798
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.008997318107103e-06,
+      "loss": 0.533,
+      "step": 28799
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.008420566946391e-06,
+      "loss": 0.3508,
+      "step": 28800
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.00784381578568e-06,
+      "loss": 0.53,
+      "step": 28801
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.007267064624968e-06,
+      "loss": 0.3759,
+      "step": 28802
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.006690313464256e-06,
+      "loss": 0.4441,
+      "step": 28803
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.006113562303545e-06,
+      "loss": 0.4814,
+      "step": 28804
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0055368111428325e-06,
+      "loss": 0.4341,
+      "step": 28805
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.004960059982121e-06,
+      "loss": 0.3552,
+      "step": 28806
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0043833088214095e-06,
+      "loss": 0.3165,
+      "step": 28807
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.003806557660698e-06,
+      "loss": 0.4202,
+      "step": 28808
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.003229806499986e-06,
+      "loss": 0.321,
+      "step": 28809
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.002653055339274e-06,
+      "loss": 0.3399,
+      "step": 28810
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.002076304178563e-06,
+      "loss": 0.3954,
+      "step": 28811
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0014995530178505e-06,
+      "loss": 0.3835,
+      "step": 28812
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.000922801857139e-06,
+      "loss": 0.4049,
+      "step": 28813
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 4.0003460506964275e-06,
+      "loss": 0.4119,
+      "step": 28814
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.999769299535716e-06,
+      "loss": 0.3529,
+      "step": 28815
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.999192548375004e-06,
+      "loss": 0.4746,
+      "step": 28816
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.998615797214292e-06,
+      "loss": 0.3675,
+      "step": 28817
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.998039046053581e-06,
+      "loss": 0.3478,
+      "step": 28818
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9974622948928685e-06,
+      "loss": 0.4777,
+      "step": 28819
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.996885543732157e-06,
+      "loss": 0.4043,
+      "step": 28820
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.996308792571446e-06,
+      "loss": 0.3161,
+      "step": 28821
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.995732041410733e-06,
+      "loss": 0.3634,
+      "step": 28822
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.995155290250022e-06,
+      "loss": 0.4471,
+      "step": 28823
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.99457853908931e-06,
+      "loss": 0.3432,
+      "step": 28824
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.994001787928599e-06,
+      "loss": 0.4324,
+      "step": 28825
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9934250367678865e-06,
+      "loss": 0.3849,
+      "step": 28826
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.992848285607175e-06,
+      "loss": 0.4393,
+      "step": 28827
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.992271534446464e-06,
+      "loss": 0.4051,
+      "step": 28828
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.991694783285751e-06,
+      "loss": 0.4296,
+      "step": 28829
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.99111803212504e-06,
+      "loss": 0.3342,
+      "step": 28830
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.990541280964328e-06,
+      "loss": 0.4509,
+      "step": 28831
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.989964529803617e-06,
+      "loss": 0.3868,
+      "step": 28832
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9893877786429046e-06,
+      "loss": 0.4216,
+      "step": 28833
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.988811027482193e-06,
+      "loss": 0.4868,
+      "step": 28834
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.988234276321482e-06,
+      "loss": 0.4907,
+      "step": 28835
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.987657525160769e-06,
+      "loss": 0.4104,
+      "step": 28836
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.987080774000058e-06,
+      "loss": 0.477,
+      "step": 28837
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.986504022839346e-06,
+      "loss": 0.4433,
+      "step": 28838
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.985927271678635e-06,
+      "loss": 0.3366,
+      "step": 28839
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.985350520517923e-06,
+      "loss": 0.4273,
+      "step": 28840
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.984773769357211e-06,
+      "loss": 0.5012,
+      "step": 28841
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9841970181965e-06,
+      "loss": 0.4357,
+      "step": 28842
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.983620267035787e-06,
+      "loss": 0.43,
+      "step": 28843
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.983043515875076e-06,
+      "loss": 0.4863,
+      "step": 28844
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.982466764714364e-06,
+      "loss": 0.3723,
+      "step": 28845
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.981890013553653e-06,
+      "loss": 0.3824,
+      "step": 28846
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.981313262392941e-06,
+      "loss": 0.3247,
+      "step": 28847
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.980736511232229e-06,
+      "loss": 0.4464,
+      "step": 28848
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.980159760071518e-06,
+      "loss": 0.4086,
+      "step": 28849
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.979583008910805e-06,
+      "loss": 0.3956,
+      "step": 28850
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.979006257750094e-06,
+      "loss": 0.4626,
+      "step": 28851
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9784295065893825e-06,
+      "loss": 0.4443,
+      "step": 28852
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.97785275542867e-06,
+      "loss": 0.4877,
+      "step": 28853
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.977276004267959e-06,
+      "loss": 0.3764,
+      "step": 28854
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.976699253107247e-06,
+      "loss": 0.346,
+      "step": 28855
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.976122501946536e-06,
+      "loss": 0.4322,
+      "step": 28856
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.975545750785823e-06,
+      "loss": 0.4886,
+      "step": 28857
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.974968999625112e-06,
+      "loss": 0.2987,
+      "step": 28858
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9743922484644005e-06,
+      "loss": 0.3612,
+      "step": 28859
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.973815497303688e-06,
+      "loss": 0.3481,
+      "step": 28860
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.973238746142977e-06,
+      "loss": 0.4307,
+      "step": 28861
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.972661994982265e-06,
+      "loss": 0.507,
+      "step": 28862
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.972085243821554e-06,
+      "loss": 0.4348,
+      "step": 28863
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9715084926608414e-06,
+      "loss": 0.3926,
+      "step": 28864
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.97093174150013e-06,
+      "loss": 0.2747,
+      "step": 28865
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.9703549903394185e-06,
+      "loss": 0.3281,
+      "step": 28866
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.969778239178706e-06,
+      "loss": 0.4287,
+      "step": 28867
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.969201488017996e-06,
+      "loss": 0.4474,
+      "step": 28868
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.968624736857283e-06,
+      "loss": 0.3342,
+      "step": 28869
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.968047985696572e-06,
+      "loss": 0.5635,
+      "step": 28870
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9674712345358595e-06,
+      "loss": 0.4361,
+      "step": 28871
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.966894483375148e-06,
+      "loss": 0.4537,
+      "step": 28872
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9663177322144365e-06,
+      "loss": 0.3841,
+      "step": 28873
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.965740981053724e-06,
+      "loss": 0.4217,
+      "step": 28874
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.965164229893014e-06,
+      "loss": 0.2495,
+      "step": 28875
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.964587478732301e-06,
+      "loss": 0.2806,
+      "step": 28876
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.96401072757159e-06,
+      "loss": 0.3466,
+      "step": 28877
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.963433976410878e-06,
+      "loss": 0.4433,
+      "step": 28878
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.962857225250166e-06,
+      "loss": 0.305,
+      "step": 28879
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9622804740894546e-06,
+      "loss": 0.4422,
+      "step": 28880
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.961703722928742e-06,
+      "loss": 0.4275,
+      "step": 28881
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.961126971768031e-06,
+      "loss": 0.331,
+      "step": 28882
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.960550220607319e-06,
+      "loss": 0.4817,
+      "step": 28883
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.959973469446608e-06,
+      "loss": 0.4569,
+      "step": 28884
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.959396718285896e-06,
+      "loss": 0.4268,
+      "step": 28885
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.958819967125184e-06,
+      "loss": 0.4138,
+      "step": 28886
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.958243215964473e-06,
+      "loss": 0.3576,
+      "step": 28887
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.95766646480376e-06,
+      "loss": 0.3339,
+      "step": 28888
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.957089713643049e-06,
+      "loss": 0.3655,
+      "step": 28889
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.956512962482337e-06,
+      "loss": 0.36,
+      "step": 28890
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.955936211321625e-06,
+      "loss": 0.4458,
+      "step": 28891
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.955359460160914e-06,
+      "loss": 0.3701,
+      "step": 28892
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.954782709000202e-06,
+      "loss": 0.3762,
+      "step": 28893
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.954205957839491e-06,
+      "loss": 0.4434,
+      "step": 28894
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.953629206678779e-06,
+      "loss": 0.3217,
+      "step": 28895
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.953052455518067e-06,
+      "loss": 0.4281,
+      "step": 28896
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.952475704357355e-06,
+      "loss": 0.4059,
+      "step": 28897
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.951898953196643e-06,
+      "loss": 0.3807,
+      "step": 28898
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9513222020359324e-06,
+      "loss": 0.3093,
+      "step": 28899
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.95074545087522e-06,
+      "loss": 0.4843,
+      "step": 28900
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.950168699714509e-06,
+      "loss": 0.3981,
+      "step": 28901
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.949591948553797e-06,
+      "loss": 0.4712,
+      "step": 28902
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.949015197393085e-06,
+      "loss": 0.4392,
+      "step": 28903
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.948438446232373e-06,
+      "loss": 0.3896,
+      "step": 28904
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.947861695071661e-06,
+      "loss": 0.3748,
+      "step": 28905
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9472849439109505e-06,
+      "loss": 0.4121,
+      "step": 28906
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.946708192750238e-06,
+      "loss": 0.5557,
+      "step": 28907
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.946131441589527e-06,
+      "loss": 0.4125,
+      "step": 28908
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.945554690428815e-06,
+      "loss": 0.3746,
+      "step": 28909
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.944977939268103e-06,
+      "loss": 0.4541,
+      "step": 28910
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9444011881073914e-06,
+      "loss": 0.3787,
+      "step": 28911
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.94382443694668e-06,
+      "loss": 0.3362,
+      "step": 28912
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.943247685785968e-06,
+      "loss": 0.366,
+      "step": 28913
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.942670934625256e-06,
+      "loss": 0.258,
+      "step": 28914
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.942094183464545e-06,
+      "loss": 0.3243,
+      "step": 28915
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.941517432303833e-06,
+      "loss": 0.3948,
+      "step": 28916
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.940940681143121e-06,
+      "loss": 0.4214,
+      "step": 28917
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9403639299824095e-06,
+      "loss": 0.4045,
+      "step": 28918
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.939787178821698e-06,
+      "loss": 0.4053,
+      "step": 28919
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.939210427660986e-06,
+      "loss": 0.4455,
+      "step": 28920
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.938633676500274e-06,
+      "loss": 0.4506,
+      "step": 28921
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.938056925339562e-06,
+      "loss": 0.3285,
+      "step": 28922
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.937480174178851e-06,
+      "loss": 0.307,
+      "step": 28923
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.936903423018139e-06,
+      "loss": 0.4484,
+      "step": 28924
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9363266718574275e-06,
+      "loss": 0.436,
+      "step": 28925
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.935749920696716e-06,
+      "loss": 0.4316,
+      "step": 28926
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.935173169536004e-06,
+      "loss": 0.3851,
+      "step": 28927
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.934596418375292e-06,
+      "loss": 0.3782,
+      "step": 28928
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.934019667214581e-06,
+      "loss": 0.3605,
+      "step": 28929
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.933442916053869e-06,
+      "loss": 0.4226,
+      "step": 28930
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.932866164893157e-06,
+      "loss": 0.4071,
+      "step": 28931
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9322894137324455e-06,
+      "loss": 0.4407,
+      "step": 28932
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.931712662571734e-06,
+      "loss": 0.3277,
+      "step": 28933
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.931135911411022e-06,
+      "loss": 0.3924,
+      "step": 28934
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.93055916025031e-06,
+      "loss": 0.3633,
+      "step": 28935
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.929982409089599e-06,
+      "loss": 0.4137,
+      "step": 28936
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.929405657928887e-06,
+      "loss": 0.3984,
+      "step": 28937
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.928828906768175e-06,
+      "loss": 0.2967,
+      "step": 28938
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9282521556074636e-06,
+      "loss": 0.3701,
+      "step": 28939
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.927675404446752e-06,
+      "loss": 0.3123,
+      "step": 28940
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.92709865328604e-06,
+      "loss": 0.4161,
+      "step": 28941
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.926521902125328e-06,
+      "loss": 0.5306,
+      "step": 28942
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.925945150964617e-06,
+      "loss": 0.3704,
+      "step": 28943
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9253683998039045e-06,
+      "loss": 0.4442,
+      "step": 28944
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.924791648643193e-06,
+      "loss": 0.3944,
+      "step": 28945
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.924214897482482e-06,
+      "loss": 0.4226,
+      "step": 28946
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.92363814632177e-06,
+      "loss": 0.4311,
+      "step": 28947
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.923061395161058e-06,
+      "loss": 0.4225,
+      "step": 28948
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.922484644000346e-06,
+      "loss": 0.4578,
+      "step": 28949
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.921907892839635e-06,
+      "loss": 0.3786,
+      "step": 28950
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9213311416789226e-06,
+      "loss": 0.4261,
+      "step": 28951
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.920754390518211e-06,
+      "loss": 0.2846,
+      "step": 28952
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9201776393575e-06,
+      "loss": 0.3882,
+      "step": 28953
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.919600888196788e-06,
+      "loss": 0.3897,
+      "step": 28954
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.919024137036076e-06,
+      "loss": 0.4348,
+      "step": 28955
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.918447385875364e-06,
+      "loss": 0.4307,
+      "step": 28956
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.917870634714653e-06,
+      "loss": 0.3755,
+      "step": 28957
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.917293883553941e-06,
+      "loss": 0.3905,
+      "step": 28958
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.916717132393229e-06,
+      "loss": 0.3736,
+      "step": 28959
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.916140381232518e-06,
+      "loss": 0.4353,
+      "step": 28960
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.915563630071806e-06,
+      "loss": 0.4115,
+      "step": 28961
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.914986878911094e-06,
+      "loss": 0.5204,
+      "step": 28962
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.914410127750382e-06,
+      "loss": 0.3196,
+      "step": 28963
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.913833376589671e-06,
+      "loss": 0.3452,
+      "step": 28964
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.913256625428959e-06,
+      "loss": 0.4256,
+      "step": 28965
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.912679874268247e-06,
+      "loss": 0.3731,
+      "step": 28966
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.912103123107536e-06,
+      "loss": 0.3443,
+      "step": 28967
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.911526371946824e-06,
+      "loss": 0.4913,
+      "step": 28968
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.910949620786112e-06,
+      "loss": 0.3911,
+      "step": 28969
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9103728696254004e-06,
+      "loss": 0.4825,
+      "step": 28970
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.909796118464689e-06,
+      "loss": 0.433,
+      "step": 28971
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.909219367303977e-06,
+      "loss": 0.3852,
+      "step": 28972
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.908642616143265e-06,
+      "loss": 0.5577,
+      "step": 28973
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.908065864982554e-06,
+      "loss": 0.481,
+      "step": 28974
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.907489113821841e-06,
+      "loss": 0.3503,
+      "step": 28975
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.90691236266113e-06,
+      "loss": 0.4141,
+      "step": 28976
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9063356115004185e-06,
+      "loss": 0.3551,
+      "step": 28977
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.905758860339707e-06,
+      "loss": 0.3913,
+      "step": 28978
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.905182109178995e-06,
+      "loss": 0.3808,
+      "step": 28979
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.904605358018283e-06,
+      "loss": 0.4134,
+      "step": 28980
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.904028606857572e-06,
+      "loss": 0.4054,
+      "step": 28981
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9034518556968594e-06,
+      "loss": 0.4803,
+      "step": 28982
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.902875104536148e-06,
+      "loss": 0.3835,
+      "step": 28983
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.9022983533754365e-06,
+      "loss": 0.4631,
+      "step": 28984
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.901721602214725e-06,
+      "loss": 0.3685,
+      "step": 28985
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.901144851054013e-06,
+      "loss": 0.3863,
+      "step": 28986
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.900568099893301e-06,
+      "loss": 0.2741,
+      "step": 28987
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.89999134873259e-06,
+      "loss": 0.4155,
+      "step": 28988
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8994145975718775e-06,
+      "loss": 0.3874,
+      "step": 28989
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.898837846411167e-06,
+      "loss": 0.4115,
+      "step": 28990
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8982610952504545e-06,
+      "loss": 0.4054,
+      "step": 28991
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.897684344089743e-06,
+      "loss": 0.453,
+      "step": 28992
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.897107592929031e-06,
+      "loss": 0.3937,
+      "step": 28993
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.896530841768319e-06,
+      "loss": 0.4484,
+      "step": 28994
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.895954090607608e-06,
+      "loss": 0.3599,
+      "step": 28995
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8953773394468955e-06,
+      "loss": 0.5322,
+      "step": 28996
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.894800588286184e-06,
+      "loss": 0.3071,
+      "step": 28997
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8942238371254726e-06,
+      "loss": 0.3483,
+      "step": 28998
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.893647085964761e-06,
+      "loss": 0.4502,
+      "step": 28999
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.893070334804049e-06,
+      "loss": 0.2854,
+      "step": 29000
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.892493583643337e-06,
+      "loss": 0.4151,
+      "step": 29001
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.891916832482626e-06,
+      "loss": 0.4813,
+      "step": 29002
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8913400813219135e-06,
+      "loss": 0.4539,
+      "step": 29003
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.890763330161202e-06,
+      "loss": 0.438,
+      "step": 29004
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.890186579000491e-06,
+      "loss": 0.5315,
+      "step": 29005
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.889609827839778e-06,
+      "loss": 0.3522,
+      "step": 29006
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.889033076679068e-06,
+      "loss": 0.4637,
+      "step": 29007
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.888456325518355e-06,
+      "loss": 0.3614,
+      "step": 29008
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.887879574357644e-06,
+      "loss": 0.2853,
+      "step": 29009
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8873028231969316e-06,
+      "loss": 0.3534,
+      "step": 29010
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.88672607203622e-06,
+      "loss": 0.3491,
+      "step": 29011
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.886149320875509e-06,
+      "loss": 0.3217,
+      "step": 29012
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.885572569714796e-06,
+      "loss": 0.4157,
+      "step": 29013
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.884995818554086e-06,
+      "loss": 0.3673,
+      "step": 29014
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.884419067393373e-06,
+      "loss": 0.4084,
+      "step": 29015
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.883842316232662e-06,
+      "loss": 0.4424,
+      "step": 29016
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8832655650719504e-06,
+      "loss": 0.4533,
+      "step": 29017
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.882688813911238e-06,
+      "loss": 0.3857,
+      "step": 29018
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.882112062750527e-06,
+      "loss": 0.4284,
+      "step": 29019
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.881535311589814e-06,
+      "loss": 0.3637,
+      "step": 29020
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.880958560429104e-06,
+      "loss": 0.4327,
+      "step": 29021
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.880381809268391e-06,
+      "loss": 0.4414,
+      "step": 29022
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.87980505810768e-06,
+      "loss": 0.4234,
+      "step": 29023
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8792283069469685e-06,
+      "loss": 0.4429,
+      "step": 29024
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.878651555786256e-06,
+      "loss": 0.5332,
+      "step": 29025
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.878074804625545e-06,
+      "loss": 0.4688,
+      "step": 29026
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.877498053464832e-06,
+      "loss": 0.4359,
+      "step": 29027
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.876921302304122e-06,
+      "loss": 0.406,
+      "step": 29028
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8763445511434094e-06,
+      "loss": 0.4502,
+      "step": 29029
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.875767799982698e-06,
+      "loss": 0.4094,
+      "step": 29030
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8751910488219865e-06,
+      "loss": 0.4357,
+      "step": 29031
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.874614297661274e-06,
+      "loss": 0.4061,
+      "step": 29032
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.874037546500563e-06,
+      "loss": 0.5239,
+      "step": 29033
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.873460795339851e-06,
+      "loss": 0.3903,
+      "step": 29034
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.872884044179139e-06,
+      "loss": 0.3852,
+      "step": 29035
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8723072930184275e-06,
+      "loss": 0.4531,
+      "step": 29036
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.871730541857715e-06,
+      "loss": 0.4509,
+      "step": 29037
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8711537906970045e-06,
+      "loss": 0.4028,
+      "step": 29038
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.870577039536292e-06,
+      "loss": 0.4635,
+      "step": 29039
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.870000288375581e-06,
+      "loss": 0.3933,
+      "step": 29040
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.869423537214869e-06,
+      "loss": 0.3034,
+      "step": 29041
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.868846786054157e-06,
+      "loss": 0.4148,
+      "step": 29042
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8682700348934455e-06,
+      "loss": 0.5325,
+      "step": 29043
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.867693283732733e-06,
+      "loss": 0.3692,
+      "step": 29044
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.8671165325720226e-06,
+      "loss": 0.2713,
+      "step": 29045
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.86653978141131e-06,
+      "loss": 0.4563,
+      "step": 29046
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.865963030250599e-06,
+      "loss": 0.369,
+      "step": 29047
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.865386279089887e-06,
+      "loss": 0.466,
+      "step": 29048
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.864809527929175e-06,
+      "loss": 0.4589,
+      "step": 29049
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8642327767684635e-06,
+      "loss": 0.4253,
+      "step": 29050
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.863656025607752e-06,
+      "loss": 0.4501,
+      "step": 29051
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.863079274447041e-06,
+      "loss": 0.4131,
+      "step": 29052
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.862502523286328e-06,
+      "loss": 0.4495,
+      "step": 29053
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.861925772125617e-06,
+      "loss": 0.5119,
+      "step": 29054
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.861349020964905e-06,
+      "loss": 0.3704,
+      "step": 29055
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.860772269804193e-06,
+      "loss": 0.4871,
+      "step": 29056
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8601955186434816e-06,
+      "loss": 0.3645,
+      "step": 29057
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.85961876748277e-06,
+      "loss": 0.5021,
+      "step": 29058
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.859042016322059e-06,
+      "loss": 0.4526,
+      "step": 29059
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.858465265161346e-06,
+      "loss": 0.4124,
+      "step": 29060
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.857888514000635e-06,
+      "loss": 0.343,
+      "step": 29061
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.857311762839923e-06,
+      "loss": 0.4317,
+      "step": 29062
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.856735011679211e-06,
+      "loss": 0.4604,
+      "step": 29063
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8561582605185e-06,
+      "loss": 0.4534,
+      "step": 29064
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.855581509357788e-06,
+      "loss": 0.4136,
+      "step": 29065
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.855004758197076e-06,
+      "loss": 0.355,
+      "step": 29066
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.854428007036364e-06,
+      "loss": 0.3952,
+      "step": 29067
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.853851255875653e-06,
+      "loss": 0.4055,
+      "step": 29068
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.853274504714941e-06,
+      "loss": 0.502,
+      "step": 29069
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.852697753554229e-06,
+      "loss": 0.3713,
+      "step": 29070
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.852121002393518e-06,
+      "loss": 0.3742,
+      "step": 29071
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.851544251232806e-06,
+      "loss": 0.4542,
+      "step": 29072
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.850967500072094e-06,
+      "loss": 0.3162,
+      "step": 29073
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.850390748911382e-06,
+      "loss": 0.4141,
+      "step": 29074
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.849813997750671e-06,
+      "loss": 0.3522,
+      "step": 29075
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8492372465899594e-06,
+      "loss": 0.4067,
+      "step": 29076
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.848660495429247e-06,
+      "loss": 0.4734,
+      "step": 29077
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.848083744268536e-06,
+      "loss": 0.4403,
+      "step": 29078
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.847506993107824e-06,
+      "loss": 0.4863,
+      "step": 29079
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.846930241947112e-06,
+      "loss": 0.383,
+      "step": 29080
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8463534907864e-06,
+      "loss": 0.3791,
+      "step": 29081
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.845776739625689e-06,
+      "loss": 0.3583,
+      "step": 29082
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8451999884649775e-06,
+      "loss": 0.4417,
+      "step": 29083
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.844623237304265e-06,
+      "loss": 0.3394,
+      "step": 29084
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.844046486143554e-06,
+      "loss": 0.4572,
+      "step": 29085
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.843469734982842e-06,
+      "loss": 0.4213,
+      "step": 29086
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.84289298382213e-06,
+      "loss": 0.3154,
+      "step": 29087
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8423162326614184e-06,
+      "loss": 0.3558,
+      "step": 29088
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.841739481500707e-06,
+      "loss": 0.4843,
+      "step": 29089
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8411627303399955e-06,
+      "loss": 0.4454,
+      "step": 29090
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.840585979179283e-06,
+      "loss": 0.378,
+      "step": 29091
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.840009228018572e-06,
+      "loss": 0.3392,
+      "step": 29092
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.83943247685786e-06,
+      "loss": 0.413,
+      "step": 29093
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.838855725697148e-06,
+      "loss": 0.3388,
+      "step": 29094
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8382789745364365e-06,
+      "loss": 0.2939,
+      "step": 29095
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.837702223375725e-06,
+      "loss": 0.4047,
+      "step": 29096
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.837125472215013e-06,
+      "loss": 0.3731,
+      "step": 29097
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.836548721054301e-06,
+      "loss": 0.3662,
+      "step": 29098
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.83597196989359e-06,
+      "loss": 0.4334,
+      "step": 29099
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.835395218732878e-06,
+      "loss": 0.3961,
+      "step": 29100
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.834818467572166e-06,
+      "loss": 0.3331,
+      "step": 29101
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8342417164114545e-06,
+      "loss": 0.419,
+      "step": 29102
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.833664965250743e-06,
+      "loss": 0.3604,
+      "step": 29103
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.833088214090031e-06,
+      "loss": 0.3546,
+      "step": 29104
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.832511462929319e-06,
+      "loss": 0.4367,
+      "step": 29105
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.831934711768608e-06,
+      "loss": 0.4369,
+      "step": 29106
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.831357960607896e-06,
+      "loss": 0.4631,
+      "step": 29107
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.830781209447184e-06,
+      "loss": 0.4458,
+      "step": 29108
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8302044582864725e-06,
+      "loss": 0.441,
+      "step": 29109
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.829627707125761e-06,
+      "loss": 0.4431,
+      "step": 29110
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.829050955965049e-06,
+      "loss": 0.5839,
+      "step": 29111
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.828474204804338e-06,
+      "loss": 0.4579,
+      "step": 29112
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.827897453643626e-06,
+      "loss": 0.39,
+      "step": 29113
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.827320702482914e-06,
+      "loss": 0.3349,
+      "step": 29114
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.826743951322202e-06,
+      "loss": 0.3858,
+      "step": 29115
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8261672001614906e-06,
+      "loss": 0.4383,
+      "step": 29116
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.825590449000779e-06,
+      "loss": 0.4517,
+      "step": 29117
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.825013697840067e-06,
+      "loss": 0.388,
+      "step": 29118
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.824436946679355e-06,
+      "loss": 0.4758,
+      "step": 29119
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.823860195518644e-06,
+      "loss": 0.3957,
+      "step": 29120
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.823283444357932e-06,
+      "loss": 0.4231,
+      "step": 29121
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.82270669319722e-06,
+      "loss": 0.4044,
+      "step": 29122
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.822129942036509e-06,
+      "loss": 0.4863,
+      "step": 29123
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.821553190875797e-06,
+      "loss": 0.3727,
+      "step": 29124
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.820976439715085e-06,
+      "loss": 0.417,
+      "step": 29125
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.820399688554373e-06,
+      "loss": 0.46,
+      "step": 29126
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.819822937393662e-06,
+      "loss": 0.3552,
+      "step": 29127
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8192461862329496e-06,
+      "loss": 0.4179,
+      "step": 29128
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.818669435072239e-06,
+      "loss": 0.4711,
+      "step": 29129
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.818092683911527e-06,
+      "loss": 0.4671,
+      "step": 29130
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.817515932750815e-06,
+      "loss": 0.4066,
+      "step": 29131
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.816939181590103e-06,
+      "loss": 0.3352,
+      "step": 29132
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.816362430429391e-06,
+      "loss": 0.4027,
+      "step": 29133
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.81578567926868e-06,
+      "loss": 0.3076,
+      "step": 29134
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.815208928107968e-06,
+      "loss": 0.4453,
+      "step": 29135
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8146321769472565e-06,
+      "loss": 0.3214,
+      "step": 29136
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8140554257865447e-06,
+      "loss": 0.3179,
+      "step": 29137
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8134786746258328e-06,
+      "loss": 0.4951,
+      "step": 29138
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8129019234651217e-06,
+      "loss": 0.3937,
+      "step": 29139
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.81232517230441e-06,
+      "loss": 0.4818,
+      "step": 29140
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.811748421143698e-06,
+      "loss": 0.3702,
+      "step": 29141
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.811171669982986e-06,
+      "loss": 0.4294,
+      "step": 29142
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8105949188222746e-06,
+      "loss": 0.3812,
+      "step": 29143
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8100181676615627e-06,
+      "loss": 0.4076,
+      "step": 29144
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.809441416500851e-06,
+      "loss": 0.3382,
+      "step": 29145
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8088646653401397e-06,
+      "loss": 0.4012,
+      "step": 29146
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.808287914179428e-06,
+      "loss": 0.288,
+      "step": 29147
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.807711163018716e-06,
+      "loss": 0.5332,
+      "step": 29148
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.807134411858004e-06,
+      "loss": 0.47,
+      "step": 29149
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8065576606972926e-06,
+      "loss": 0.3565,
+      "step": 29150
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8059809095365807e-06,
+      "loss": 0.3831,
+      "step": 29151
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.805404158375869e-06,
+      "loss": 0.3802,
+      "step": 29152
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8048274072151574e-06,
+      "loss": 0.4774,
+      "step": 29153
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8042506560544455e-06,
+      "loss": 0.3806,
+      "step": 29154
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8036739048937336e-06,
+      "loss": 0.2756,
+      "step": 29155
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8030971537330225e-06,
+      "loss": 0.4371,
+      "step": 29156
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8025204025723106e-06,
+      "loss": 0.3973,
+      "step": 29157
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8019436514115987e-06,
+      "loss": 0.4103,
+      "step": 29158
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.801366900250887e-06,
+      "loss": 0.2838,
+      "step": 29159
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8007901490901754e-06,
+      "loss": 0.4464,
+      "step": 29160
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.8002133979294635e-06,
+      "loss": 0.3374,
+      "step": 29161
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7996366467687516e-06,
+      "loss": 0.4647,
+      "step": 29162
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7990598956080406e-06,
+      "loss": 0.4099,
+      "step": 29163
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7984831444473287e-06,
+      "loss": 0.3862,
+      "step": 29164
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7979063932866168e-06,
+      "loss": 0.4191,
+      "step": 29165
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.797329642125905e-06,
+      "loss": 0.3484,
+      "step": 29166
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7967528909651934e-06,
+      "loss": 0.3709,
+      "step": 29167
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7961761398044815e-06,
+      "loss": 0.3735,
+      "step": 29168
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7955993886437696e-06,
+      "loss": 0.4849,
+      "step": 29169
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7950226374830586e-06,
+      "loss": 0.4144,
+      "step": 29170
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7944458863223467e-06,
+      "loss": 0.3784,
+      "step": 29171
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.793869135161635e-06,
+      "loss": 0.4587,
+      "step": 29172
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7932923840009233e-06,
+      "loss": 0.4444,
+      "step": 29173
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7927156328402114e-06,
+      "loss": 0.4731,
+      "step": 29174
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7921388816794996e-06,
+      "loss": 0.4277,
+      "step": 29175
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7915621305187877e-06,
+      "loss": 0.4664,
+      "step": 29176
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7909853793580766e-06,
+      "loss": 0.3426,
+      "step": 29177
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7904086281973647e-06,
+      "loss": 0.4334,
+      "step": 29178
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.789831877036653e-06,
+      "loss": 0.4733,
+      "step": 29179
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7892551258759414e-06,
+      "loss": 0.4504,
+      "step": 29180
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7886783747152295e-06,
+      "loss": 0.4328,
+      "step": 29181
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7881016235545176e-06,
+      "loss": 0.4441,
+      "step": 29182
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7875248723938057e-06,
+      "loss": 0.4431,
+      "step": 29183
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7869481212330942e-06,
+      "loss": 0.4171,
+      "step": 29184
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7863713700723823e-06,
+      "loss": 0.4068,
+      "step": 29185
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.785794618911671e-06,
+      "loss": 0.4613,
+      "step": 29186
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7852178677509594e-06,
+      "loss": 0.426,
+      "step": 29187
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7846411165902475e-06,
+      "loss": 0.4197,
+      "step": 29188
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7840643654295356e-06,
+      "loss": 0.473,
+      "step": 29189
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.783487614268824e-06,
+      "loss": 0.48,
+      "step": 29190
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7829108631081123e-06,
+      "loss": 0.3508,
+      "step": 29191
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7823341119474004e-06,
+      "loss": 0.4074,
+      "step": 29192
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7817573607866885e-06,
+      "loss": 0.4836,
+      "step": 29193
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7811806096259774e-06,
+      "loss": 0.406,
+      "step": 29194
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7806038584652655e-06,
+      "loss": 0.3879,
+      "step": 29195
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7800271073045536e-06,
+      "loss": 0.2841,
+      "step": 29196
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.779450356143842e-06,
+      "loss": 0.3304,
+      "step": 29197
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7788736049831303e-06,
+      "loss": 0.4084,
+      "step": 29198
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7782968538224184e-06,
+      "loss": 0.4443,
+      "step": 29199
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7777201026617074e-06,
+      "loss": 0.3778,
+      "step": 29200
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7771433515009955e-06,
+      "loss": 0.5192,
+      "step": 29201
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7765666003402836e-06,
+      "loss": 0.3663,
+      "step": 29202
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7759898491795717e-06,
+      "loss": 0.3769,
+      "step": 29203
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7754130980188602e-06,
+      "loss": 0.485,
+      "step": 29204
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7748363468581483e-06,
+      "loss": 0.3314,
+      "step": 29205
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7742595956974364e-06,
+      "loss": 0.3784,
+      "step": 29206
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7736828445367254e-06,
+      "loss": 0.3132,
+      "step": 29207
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7731060933760135e-06,
+      "loss": 0.4547,
+      "step": 29208
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7725293422153016e-06,
+      "loss": 0.4002,
+      "step": 29209
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7719525910545897e-06,
+      "loss": 0.301,
+      "step": 29210
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7713758398938782e-06,
+      "loss": 0.3178,
+      "step": 29211
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7707990887331664e-06,
+      "loss": 0.3647,
+      "step": 29212
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7702223375724545e-06,
+      "loss": 0.4711,
+      "step": 29213
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.769645586411743e-06,
+      "loss": 0.4052,
+      "step": 29214
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.769068835251031e-06,
+      "loss": 0.3135,
+      "step": 29215
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.768492084090319e-06,
+      "loss": 0.3223,
+      "step": 29216
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.767915332929608e-06,
+      "loss": 0.4924,
+      "step": 29217
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7673385817688963e-06,
+      "loss": 0.37,
+      "step": 29218
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7667618306081844e-06,
+      "loss": 0.4014,
+      "step": 29219
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7661850794474725e-06,
+      "loss": 0.3462,
+      "step": 29220
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.765608328286761e-06,
+      "loss": 0.3702,
+      "step": 29221
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.765031577126049e-06,
+      "loss": 0.4684,
+      "step": 29222
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7644548259653372e-06,
+      "loss": 0.448,
+      "step": 29223
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.763878074804626e-06,
+      "loss": 0.4401,
+      "step": 29224
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 3.7633013236439143e-06,
+      "loss": 0.2827,
+      "step": 29225
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7627245724832024e-06,
+      "loss": 0.4726,
+      "step": 29226
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7621478213224905e-06,
+      "loss": 0.3912,
+      "step": 29227
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.761571070161779e-06,
+      "loss": 0.3205,
+      "step": 29228
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.760994319001067e-06,
+      "loss": 0.4281,
+      "step": 29229
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7604175678403553e-06,
+      "loss": 0.4164,
+      "step": 29230
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7598408166796442e-06,
+      "loss": 0.4022,
+      "step": 29231
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7592640655189323e-06,
+      "loss": 0.3718,
+      "step": 29232
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7586873143582204e-06,
+      "loss": 0.4052,
+      "step": 29233
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.758110563197509e-06,
+      "loss": 0.3991,
+      "step": 29234
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.757533812036797e-06,
+      "loss": 0.3758,
+      "step": 29235
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.756957060876085e-06,
+      "loss": 0.4259,
+      "step": 29236
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7563803097153733e-06,
+      "loss": 0.3861,
+      "step": 29237
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7558035585546623e-06,
+      "loss": 0.5043,
+      "step": 29238
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7552268073939504e-06,
+      "loss": 0.3079,
+      "step": 29239
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7546500562332385e-06,
+      "loss": 0.5059,
+      "step": 29240
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.754073305072527e-06,
+      "loss": 0.354,
+      "step": 29241
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.753496553911815e-06,
+      "loss": 0.4356,
+      "step": 29242
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7529198027511032e-06,
+      "loss": 0.4,
+      "step": 29243
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7523430515903913e-06,
+      "loss": 0.4923,
+      "step": 29244
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.75176630042968e-06,
+      "loss": 0.3982,
+      "step": 29245
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.751189549268968e-06,
+      "loss": 0.4782,
+      "step": 29246
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.750612798108256e-06,
+      "loss": 0.4239,
+      "step": 29247
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.750036046947545e-06,
+      "loss": 0.4011,
+      "step": 29248
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.749459295786833e-06,
+      "loss": 0.3178,
+      "step": 29249
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7488825446261213e-06,
+      "loss": 0.3963,
+      "step": 29250
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7483057934654098e-06,
+      "loss": 0.3034,
+      "step": 29251
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.747729042304698e-06,
+      "loss": 0.4807,
+      "step": 29252
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.747152291143986e-06,
+      "loss": 0.3736,
+      "step": 29253
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.746575539983274e-06,
+      "loss": 0.3927,
+      "step": 29254
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.745998788822563e-06,
+      "loss": 0.3174,
+      "step": 29255
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.745422037661851e-06,
+      "loss": 0.4668,
+      "step": 29256
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7448452865011393e-06,
+      "loss": 0.4677,
+      "step": 29257
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.744268535340428e-06,
+      "loss": 0.4091,
+      "step": 29258
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.743691784179716e-06,
+      "loss": 0.5272,
+      "step": 29259
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.743115033019004e-06,
+      "loss": 0.3389,
+      "step": 29260
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.742538281858293e-06,
+      "loss": 0.4219,
+      "step": 29261
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.741961530697581e-06,
+      "loss": 0.2792,
+      "step": 29262
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.741384779536869e-06,
+      "loss": 0.486,
+      "step": 29263
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7408080283761573e-06,
+      "loss": 0.2945,
+      "step": 29264
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.740231277215446e-06,
+      "loss": 0.2965,
+      "step": 29265
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.739654526054734e-06,
+      "loss": 0.5267,
+      "step": 29266
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.739077774894022e-06,
+      "loss": 0.3993,
+      "step": 29267
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7385010237333106e-06,
+      "loss": 0.3354,
+      "step": 29268
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.737924272572599e-06,
+      "loss": 0.3608,
+      "step": 29269
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7373475214118872e-06,
+      "loss": 0.409,
+      "step": 29270
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7367707702511753e-06,
+      "loss": 0.4108,
+      "step": 29271
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.736194019090464e-06,
+      "loss": 0.4937,
+      "step": 29272
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.735617267929752e-06,
+      "loss": 0.4418,
+      "step": 29273
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.73504051676904e-06,
+      "loss": 0.4014,
+      "step": 29274
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7344637656083286e-06,
+      "loss": 0.3694,
+      "step": 29275
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7338870144476167e-06,
+      "loss": 0.3248,
+      "step": 29276
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.733310263286905e-06,
+      "loss": 0.339,
+      "step": 29277
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.732733512126194e-06,
+      "loss": 0.4141,
+      "step": 29278
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.732156760965482e-06,
+      "loss": 0.3947,
+      "step": 29279
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.73158000980477e-06,
+      "loss": 0.3871,
+      "step": 29280
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.731003258644058e-06,
+      "loss": 0.31,
+      "step": 29281
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7304265074833467e-06,
+      "loss": 0.3461,
+      "step": 29282
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7298497563226348e-06,
+      "loss": 0.4601,
+      "step": 29283
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.729273005161923e-06,
+      "loss": 0.4214,
+      "step": 29284
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.728696254001212e-06,
+      "loss": 0.523,
+      "step": 29285
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7281195028405e-06,
+      "loss": 0.4808,
+      "step": 29286
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.727542751679788e-06,
+      "loss": 0.5441,
+      "step": 29287
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.726966000519076e-06,
+      "loss": 0.3616,
+      "step": 29288
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7263892493583647e-06,
+      "loss": 0.461,
+      "step": 29289
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.725812498197653e-06,
+      "loss": 0.3619,
+      "step": 29290
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.725235747036941e-06,
+      "loss": 0.4904,
+      "step": 29291
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.72465899587623e-06,
+      "loss": 0.4197,
+      "step": 29292
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.724082244715518e-06,
+      "loss": 0.435,
+      "step": 29293
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.723505493554806e-06,
+      "loss": 0.4271,
+      "step": 29294
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7229287423940946e-06,
+      "loss": 0.3844,
+      "step": 29295
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7223519912333827e-06,
+      "loss": 0.4487,
+      "step": 29296
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.721775240072671e-06,
+      "loss": 0.3877,
+      "step": 29297
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.721198488911959e-06,
+      "loss": 0.4202,
+      "step": 29298
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.720621737751248e-06,
+      "loss": 0.3739,
+      "step": 29299
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.720044986590536e-06,
+      "loss": 0.4198,
+      "step": 29300
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.719468235429824e-06,
+      "loss": 0.365,
+      "step": 29301
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7188914842691126e-06,
+      "loss": 0.3906,
+      "step": 29302
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7183147331084008e-06,
+      "loss": 0.4546,
+      "step": 29303
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.717737981947689e-06,
+      "loss": 0.4765,
+      "step": 29304
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.717161230786977e-06,
+      "loss": 0.3767,
+      "step": 29305
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7165844796262655e-06,
+      "loss": 0.5036,
+      "step": 29306
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7160077284655536e-06,
+      "loss": 0.3618,
+      "step": 29307
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7154309773048417e-06,
+      "loss": 0.3265,
+      "step": 29308
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7148542261441307e-06,
+      "loss": 0.3652,
+      "step": 29309
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7142774749834188e-06,
+      "loss": 0.3858,
+      "step": 29310
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.713700723822707e-06,
+      "loss": 0.397,
+      "step": 29311
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7131239726619954e-06,
+      "loss": 0.4435,
+      "step": 29312
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7125472215012835e-06,
+      "loss": 0.3368,
+      "step": 29313
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7119704703405716e-06,
+      "loss": 0.4705,
+      "step": 29314
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7113937191798598e-06,
+      "loss": 0.4691,
+      "step": 29315
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7108169680191487e-06,
+      "loss": 0.5299,
+      "step": 29316
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.710240216858437e-06,
+      "loss": 0.4531,
+      "step": 29317
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.709663465697725e-06,
+      "loss": 0.4635,
+      "step": 29318
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7090867145370135e-06,
+      "loss": 0.3298,
+      "step": 29319
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7085099633763016e-06,
+      "loss": 0.4676,
+      "step": 29320
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7079332122155897e-06,
+      "loss": 0.3512,
+      "step": 29321
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7073564610548786e-06,
+      "loss": 0.3822,
+      "step": 29322
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7067797098941667e-06,
+      "loss": 0.4113,
+      "step": 29323
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.706202958733455e-06,
+      "loss": 0.4374,
+      "step": 29324
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.705626207572743e-06,
+      "loss": 0.5272,
+      "step": 29325
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7050494564120315e-06,
+      "loss": 0.468,
+      "step": 29326
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7044727052513196e-06,
+      "loss": 0.3223,
+      "step": 29327
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7038959540906077e-06,
+      "loss": 0.4027,
+      "step": 29328
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7033192029298962e-06,
+      "loss": 0.4779,
+      "step": 29329
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7027424517691848e-06,
+      "loss": 0.3953,
+      "step": 29330
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.702165700608473e-06,
+      "loss": 0.3931,
+      "step": 29331
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.701588949447761e-06,
+      "loss": 0.4489,
+      "step": 29332
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7010121982870495e-06,
+      "loss": 0.4295,
+      "step": 29333
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7004354471263376e-06,
+      "loss": 0.4407,
+      "step": 29334
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6998586959656257e-06,
+      "loss": 0.4497,
+      "step": 29335
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6992819448049143e-06,
+      "loss": 0.4158,
+      "step": 29336
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6987051936442024e-06,
+      "loss": 0.3823,
+      "step": 29337
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6981284424834905e-06,
+      "loss": 0.5889,
+      "step": 29338
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6975516913227794e-06,
+      "loss": 0.3915,
+      "step": 29339
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6969749401620675e-06,
+      "loss": 0.3285,
+      "step": 29340
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6963981890013557e-06,
+      "loss": 0.3896,
+      "step": 29341
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6958214378406438e-06,
+      "loss": 0.3154,
+      "step": 29342
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6952446866799323e-06,
+      "loss": 0.3504,
+      "step": 29343
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6946679355192204e-06,
+      "loss": 0.5131,
+      "step": 29344
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6940911843585085e-06,
+      "loss": 0.4752,
+      "step": 29345
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6935144331977975e-06,
+      "loss": 0.4291,
+      "step": 29346
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6929376820370856e-06,
+      "loss": 0.4094,
+      "step": 29347
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6923609308763737e-06,
+      "loss": 0.3973,
+      "step": 29348
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.691784179715662e-06,
+      "loss": 0.5274,
+      "step": 29349
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6912074285549503e-06,
+      "loss": 0.4702,
+      "step": 29350
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6906306773942384e-06,
+      "loss": 0.3275,
+      "step": 29351
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6900539262335265e-06,
+      "loss": 0.4057,
+      "step": 29352
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6894771750728155e-06,
+      "loss": 0.4147,
+      "step": 29353
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6889004239121036e-06,
+      "loss": 0.3901,
+      "step": 29354
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6883236727513917e-06,
+      "loss": 0.4439,
+      "step": 29355
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6877469215906803e-06,
+      "loss": 0.3532,
+      "step": 29356
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6871701704299684e-06,
+      "loss": 0.5,
+      "step": 29357
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6865934192692565e-06,
+      "loss": 0.5189,
+      "step": 29358
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6860166681085446e-06,
+      "loss": 0.3666,
+      "step": 29359
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.685439916947833e-06,
+      "loss": 0.3394,
+      "step": 29360
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6848631657871216e-06,
+      "loss": 0.4085,
+      "step": 29361
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6842864146264097e-06,
+      "loss": 0.4462,
+      "step": 29362
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6837096634656983e-06,
+      "loss": 0.3259,
+      "step": 29363
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6831329123049864e-06,
+      "loss": 0.4697,
+      "step": 29364
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6825561611442745e-06,
+      "loss": 0.4348,
+      "step": 29365
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6819794099835626e-06,
+      "loss": 0.4172,
+      "step": 29366
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.681402658822851e-06,
+      "loss": 0.3327,
+      "step": 29367
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6808259076621392e-06,
+      "loss": 0.4985,
+      "step": 29368
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6802491565014274e-06,
+      "loss": 0.3485,
+      "step": 29369
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6796724053407163e-06,
+      "loss": 0.5408,
+      "step": 29370
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6790956541800044e-06,
+      "loss": 0.3548,
+      "step": 29371
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6785189030192925e-06,
+      "loss": 0.4816,
+      "step": 29372
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.677942151858581e-06,
+      "loss": 0.4406,
+      "step": 29373
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.677365400697869e-06,
+      "loss": 0.4541,
+      "step": 29374
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6767886495371573e-06,
+      "loss": 0.3041,
+      "step": 29375
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6762118983764454e-06,
+      "loss": 0.51,
+      "step": 29376
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6756351472157343e-06,
+      "loss": 0.4644,
+      "step": 29377
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6750583960550225e-06,
+      "loss": 0.4437,
+      "step": 29378
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6744816448943106e-06,
+      "loss": 0.3532,
+      "step": 29379
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.673904893733599e-06,
+      "loss": 0.3408,
+      "step": 29380
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.673328142572887e-06,
+      "loss": 0.336,
+      "step": 29381
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6727513914121753e-06,
+      "loss": 0.4209,
+      "step": 29382
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6721746402514634e-06,
+      "loss": 0.3857,
+      "step": 29383
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6715978890907524e-06,
+      "loss": 0.4623,
+      "step": 29384
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6710211379300405e-06,
+      "loss": 0.4956,
+      "step": 29385
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6704443867693286e-06,
+      "loss": 0.271,
+      "step": 29386
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.669867635608617e-06,
+      "loss": 0.444,
+      "step": 29387
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6692908844479052e-06,
+      "loss": 0.4653,
+      "step": 29388
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6687141332871933e-06,
+      "loss": 0.3672,
+      "step": 29389
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.668137382126482e-06,
+      "loss": 0.5405,
+      "step": 29390
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6675606309657704e-06,
+      "loss": 0.4484,
+      "step": 29391
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6669838798050585e-06,
+      "loss": 0.3883,
+      "step": 29392
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6664071286443466e-06,
+      "loss": 0.3728,
+      "step": 29393
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.665830377483635e-06,
+      "loss": 0.3787,
+      "step": 29394
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6652536263229233e-06,
+      "loss": 0.394,
+      "step": 29395
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6646768751622114e-06,
+      "loss": 0.4246,
+      "step": 29396
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6641001240015e-06,
+      "loss": 0.4687,
+      "step": 29397
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.663523372840788e-06,
+      "loss": 0.5008,
+      "step": 29398
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.662946621680076e-06,
+      "loss": 0.318,
+      "step": 29399
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.662369870519365e-06,
+      "loss": 0.3534,
+      "step": 29400
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.661793119358653e-06,
+      "loss": 0.4547,
+      "step": 29401
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6612163681979413e-06,
+      "loss": 0.4539,
+      "step": 29402
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.6606396170372294e-06,
+      "loss": 0.3958,
+      "step": 29403
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.660062865876518e-06,
+      "loss": 0.3384,
+      "step": 29404
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.659486114715806e-06,
+      "loss": 0.3819,
+      "step": 29405
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.658909363555094e-06,
+      "loss": 0.4911,
+      "step": 29406
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.658332612394383e-06,
+      "loss": 0.4146,
+      "step": 29407
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6577558612336712e-06,
+      "loss": 0.4946,
+      "step": 29408
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6571791100729593e-06,
+      "loss": 0.5029,
+      "step": 29409
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6566023589122474e-06,
+      "loss": 0.3826,
+      "step": 29410
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.656025607751536e-06,
+      "loss": 0.3884,
+      "step": 29411
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.655448856590824e-06,
+      "loss": 0.3483,
+      "step": 29412
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.654872105430112e-06,
+      "loss": 0.624,
+      "step": 29413
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.654295354269401e-06,
+      "loss": 0.4038,
+      "step": 29414
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6537186031086892e-06,
+      "loss": 0.3731,
+      "step": 29415
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6531418519479774e-06,
+      "loss": 0.4839,
+      "step": 29416
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.652565100787266e-06,
+      "loss": 0.2884,
+      "step": 29417
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.651988349626554e-06,
+      "loss": 0.4504,
+      "step": 29418
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.651411598465842e-06,
+      "loss": 0.3675,
+      "step": 29419
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6508348473051302e-06,
+      "loss": 0.4277,
+      "step": 29420
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6502580961444187e-06,
+      "loss": 0.4347,
+      "step": 29421
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6496813449837073e-06,
+      "loss": 0.4195,
+      "step": 29422
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6491045938229954e-06,
+      "loss": 0.41,
+      "step": 29423
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.648527842662284e-06,
+      "loss": 0.419,
+      "step": 29424
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.647951091501572e-06,
+      "loss": 0.4304,
+      "step": 29425
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.64737434034086e-06,
+      "loss": 0.3714,
+      "step": 29426
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6467975891801482e-06,
+      "loss": 0.3611,
+      "step": 29427
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6462208380194368e-06,
+      "loss": 0.307,
+      "step": 29428
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.645644086858725e-06,
+      "loss": 0.2633,
+      "step": 29429
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.645067335698013e-06,
+      "loss": 0.4296,
+      "step": 29430
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.644490584537302e-06,
+      "loss": 0.4479,
+      "step": 29431
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.64391383337659e-06,
+      "loss": 0.3547,
+      "step": 29432
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.643337082215878e-06,
+      "loss": 0.3408,
+      "step": 29433
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6427603310551667e-06,
+      "loss": 0.3997,
+      "step": 29434
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.642183579894455e-06,
+      "loss": 0.3855,
+      "step": 29435
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.641606828733743e-06,
+      "loss": 0.5259,
+      "step": 29436
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.641030077573031e-06,
+      "loss": 0.3494,
+      "step": 29437
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.64045332641232e-06,
+      "loss": 0.4848,
+      "step": 29438
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.639876575251608e-06,
+      "loss": 0.4532,
+      "step": 29439
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.639299824090896e-06,
+      "loss": 0.427,
+      "step": 29440
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6387230729301847e-06,
+      "loss": 0.378,
+      "step": 29441
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.638146321769473e-06,
+      "loss": 0.3696,
+      "step": 29442
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.637569570608761e-06,
+      "loss": 0.4998,
+      "step": 29443
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.636992819448049e-06,
+      "loss": 0.5004,
+      "step": 29444
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.636416068287338e-06,
+      "loss": 0.3922,
+      "step": 29445
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.635839317126626e-06,
+      "loss": 0.4487,
+      "step": 29446
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6352625659659142e-06,
+      "loss": 0.3778,
+      "step": 29447
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6346858148052028e-06,
+      "loss": 0.3226,
+      "step": 29448
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.634109063644491e-06,
+      "loss": 0.3292,
+      "step": 29449
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.633532312483779e-06,
+      "loss": 0.454,
+      "step": 29450
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6329555613230675e-06,
+      "loss": 0.4412,
+      "step": 29451
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6323788101623556e-06,
+      "loss": 0.4279,
+      "step": 29452
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.631802059001644e-06,
+      "loss": 0.3806,
+      "step": 29453
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6312253078409323e-06,
+      "loss": 0.4404,
+      "step": 29454
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.630648556680221e-06,
+      "loss": 0.4165,
+      "step": 29455
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.630071805519509e-06,
+      "loss": 0.4574,
+      "step": 29456
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.629495054358797e-06,
+      "loss": 0.4345,
+      "step": 29457
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6289183031980855e-06,
+      "loss": 0.3807,
+      "step": 29458
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6283415520373736e-06,
+      "loss": 0.3886,
+      "step": 29459
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6277648008766618e-06,
+      "loss": 0.3538,
+      "step": 29460
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6271880497159507e-06,
+      "loss": 0.4434,
+      "step": 29461
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.626611298555239e-06,
+      "loss": 0.4259,
+      "step": 29462
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.626034547394527e-06,
+      "loss": 0.3433,
+      "step": 29463
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.625457796233815e-06,
+      "loss": 0.2877,
+      "step": 29464
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6248810450731036e-06,
+      "loss": 0.4493,
+      "step": 29465
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6243042939123917e-06,
+      "loss": 0.42,
+      "step": 29466
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.62372754275168e-06,
+      "loss": 0.5239,
+      "step": 29467
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6231507915909687e-06,
+      "loss": 0.3655,
+      "step": 29468
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.622574040430257e-06,
+      "loss": 0.3805,
+      "step": 29469
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.621997289269545e-06,
+      "loss": 0.4107,
+      "step": 29470
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.621420538108833e-06,
+      "loss": 0.4211,
+      "step": 29471
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6208437869481216e-06,
+      "loss": 0.4631,
+      "step": 29472
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6202670357874097e-06,
+      "loss": 0.3767,
+      "step": 29473
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.619690284626698e-06,
+      "loss": 0.261,
+      "step": 29474
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6191135334659868e-06,
+      "loss": 0.3672,
+      "step": 29475
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.618536782305275e-06,
+      "loss": 0.3465,
+      "step": 29476
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.617960031144563e-06,
+      "loss": 0.4967,
+      "step": 29477
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6173832799838515e-06,
+      "loss": 0.3918,
+      "step": 29478
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6168065288231396e-06,
+      "loss": 0.4185,
+      "step": 29479
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6162297776624277e-06,
+      "loss": 0.4837,
+      "step": 29480
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.615653026501716e-06,
+      "loss": 0.3957,
+      "step": 29481
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6150762753410044e-06,
+      "loss": 0.3559,
+      "step": 29482
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.614499524180293e-06,
+      "loss": 0.3718,
+      "step": 29483
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.613922773019581e-06,
+      "loss": 0.3782,
+      "step": 29484
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6133460218588696e-06,
+      "loss": 0.5426,
+      "step": 29485
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6127692706981577e-06,
+      "loss": 0.4045,
+      "step": 29486
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6121925195374458e-06,
+      "loss": 0.3742,
+      "step": 29487
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.611615768376734e-06,
+      "loss": 0.3593,
+      "step": 29488
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6110390172160224e-06,
+      "loss": 0.25,
+      "step": 29489
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6104622660553105e-06,
+      "loss": 0.3093,
+      "step": 29490
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6098855148945986e-06,
+      "loss": 0.3846,
+      "step": 29491
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6093087637338876e-06,
+      "loss": 0.3522,
+      "step": 29492
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6087320125731757e-06,
+      "loss": 0.4719,
+      "step": 29493
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.608155261412464e-06,
+      "loss": 0.4247,
+      "step": 29494
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6075785102517523e-06,
+      "loss": 0.5448,
+      "step": 29495
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6070017590910404e-06,
+      "loss": 0.4239,
+      "step": 29496
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6064250079303286e-06,
+      "loss": 0.3918,
+      "step": 29497
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6058482567696167e-06,
+      "loss": 0.5094,
+      "step": 29498
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6052715056089056e-06,
+      "loss": 0.3891,
+      "step": 29499
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6046947544481937e-06,
+      "loss": 0.2667,
+      "step": 29500
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.604118003287482e-06,
+      "loss": 0.3703,
+      "step": 29501
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6035412521267704e-06,
+      "loss": 0.4018,
+      "step": 29502
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6029645009660585e-06,
+      "loss": 0.4803,
+      "step": 29503
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6023877498053466e-06,
+      "loss": 0.4065,
+      "step": 29504
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6018109986446347e-06,
+      "loss": 0.2976,
+      "step": 29505
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6012342474839236e-06,
+      "loss": 0.3618,
+      "step": 29506
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6006574963232118e-06,
+      "loss": 0.3852,
+      "step": 29507
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.6000807451625e-06,
+      "loss": 0.327,
+      "step": 29508
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5995039940017884e-06,
+      "loss": 0.454,
+      "step": 29509
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5989272428410765e-06,
+      "loss": 0.4501,
+      "step": 29510
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5983504916803646e-06,
+      "loss": 0.309,
+      "step": 29511
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.597773740519653e-06,
+      "loss": 0.4377,
+      "step": 29512
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5971969893589413e-06,
+      "loss": 0.3946,
+      "step": 29513
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.59662023819823e-06,
+      "loss": 0.3869,
+      "step": 29514
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.596043487037518e-06,
+      "loss": 0.3705,
+      "step": 29515
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5954667358768064e-06,
+      "loss": 0.377,
+      "step": 29516
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5948899847160945e-06,
+      "loss": 0.4224,
+      "step": 29517
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5943132335553826e-06,
+      "loss": 0.3808,
+      "step": 29518
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.593736482394671e-06,
+      "loss": 0.49,
+      "step": 29519
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5931597312339593e-06,
+      "loss": 0.387,
+      "step": 29520
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5925829800732474e-06,
+      "loss": 0.2504,
+      "step": 29521
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5920062289125364e-06,
+      "loss": 0.3867,
+      "step": 29522
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5914294777518245e-06,
+      "loss": 0.4006,
+      "step": 29523
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5908527265911126e-06,
+      "loss": 0.4296,
+      "step": 29524
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5902759754304007e-06,
+      "loss": 0.4848,
+      "step": 29525
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.589699224269689e-06,
+      "loss": 0.2514,
+      "step": 29526
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5891224731089773e-06,
+      "loss": 0.3952,
+      "step": 29527
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5885457219482654e-06,
+      "loss": 0.4373,
+      "step": 29528
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5879689707875544e-06,
+      "loss": 0.4283,
+      "step": 29529
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5873922196268425e-06,
+      "loss": 0.4701,
+      "step": 29530
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5868154684661306e-06,
+      "loss": 0.4885,
+      "step": 29531
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5862387173054187e-06,
+      "loss": 0.404,
+      "step": 29532
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5856619661447072e-06,
+      "loss": 0.3022,
+      "step": 29533
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5850852149839953e-06,
+      "loss": 0.4048,
+      "step": 29534
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5845084638232835e-06,
+      "loss": 0.5267,
+      "step": 29535
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5839317126625724e-06,
+      "loss": 0.4147,
+      "step": 29536
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5833549615018605e-06,
+      "loss": 0.4194,
+      "step": 29537
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5827782103411486e-06,
+      "loss": 0.377,
+      "step": 29538
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.582201459180437e-06,
+      "loss": 0.4535,
+      "step": 29539
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5816247080197253e-06,
+      "loss": 0.38,
+      "step": 29540
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5810479568590134e-06,
+      "loss": 0.361,
+      "step": 29541
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5804712056983015e-06,
+      "loss": 0.416,
+      "step": 29542
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.57989445453759e-06,
+      "loss": 0.4155,
+      "step": 29543
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.579317703376878e-06,
+      "loss": 0.3133,
+      "step": 29544
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5787409522161667e-06,
+      "loss": 0.3581,
+      "step": 29545
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.578164201055455e-06,
+      "loss": 0.4075,
+      "step": 29546
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5775874498947433e-06,
+      "loss": 0.4023,
+      "step": 29547
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5770106987340314e-06,
+      "loss": 0.4681,
+      "step": 29548
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5764339475733195e-06,
+      "loss": 0.3953,
+      "step": 29549
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.575857196412608e-06,
+      "loss": 0.2911,
+      "step": 29550
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.575280445251896e-06,
+      "loss": 0.394,
+      "step": 29551
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5747036940911843e-06,
+      "loss": 0.3195,
+      "step": 29552
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5741269429304732e-06,
+      "loss": 0.4294,
+      "step": 29553
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5735501917697613e-06,
+      "loss": 0.3874,
+      "step": 29554
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5729734406090494e-06,
+      "loss": 0.4601,
+      "step": 29555
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.572396689448338e-06,
+      "loss": 0.5094,
+      "step": 29556
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.571819938287626e-06,
+      "loss": 0.4428,
+      "step": 29557
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.571243187126914e-06,
+      "loss": 0.4555,
+      "step": 29558
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5706664359662023e-06,
+      "loss": 0.5166,
+      "step": 29559
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5700896848054913e-06,
+      "loss": 0.4101,
+      "step": 29560
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5695129336447794e-06,
+      "loss": 0.3723,
+      "step": 29561
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5689361824840675e-06,
+      "loss": 0.4372,
+      "step": 29562
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.568359431323356e-06,
+      "loss": 0.4726,
+      "step": 29563
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.567782680162644e-06,
+      "loss": 0.3534,
+      "step": 29564
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5672059290019322e-06,
+      "loss": 0.4685,
+      "step": 29565
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5666291778412203e-06,
+      "loss": 0.3802,
+      "step": 29566
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5660524266805093e-06,
+      "loss": 0.338,
+      "step": 29567
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5654756755197974e-06,
+      "loss": 0.4807,
+      "step": 29568
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5648989243590855e-06,
+      "loss": 0.4183,
+      "step": 29569
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.564322173198374e-06,
+      "loss": 0.3192,
+      "step": 29570
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.563745422037662e-06,
+      "loss": 0.4222,
+      "step": 29571
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5631686708769503e-06,
+      "loss": 0.5275,
+      "step": 29572
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5625919197162388e-06,
+      "loss": 0.3429,
+      "step": 29573
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.562015168555527e-06,
+      "loss": 0.4776,
+      "step": 29574
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5614384173948154e-06,
+      "loss": 0.4252,
+      "step": 29575
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5608616662341035e-06,
+      "loss": 0.4019,
+      "step": 29576
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.560284915073392e-06,
+      "loss": 0.4191,
+      "step": 29577
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.55970816391268e-06,
+      "loss": 0.3311,
+      "step": 29578
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.5591314127519683e-06,
+      "loss": 0.371,
+      "step": 29579
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.558554661591257e-06,
+      "loss": 0.536,
+      "step": 29580
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.557977910430545e-06,
+      "loss": 0.3406,
+      "step": 29581
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.557401159269833e-06,
+      "loss": 0.3525,
+      "step": 29582
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 3.556824408109122e-06,
+      "loss": 0.4787,
+      "step": 29583
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.55624765694841e-06,
+      "loss": 0.3328,
+      "step": 29584
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.555670905787698e-06,
+      "loss": 0.356,
+      "step": 29585
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5550941546269863e-06,
+      "loss": 0.3572,
+      "step": 29586
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.554517403466275e-06,
+      "loss": 0.2855,
+      "step": 29587
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.553940652305563e-06,
+      "loss": 0.4636,
+      "step": 29588
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.553363901144851e-06,
+      "loss": 0.4329,
+      "step": 29589
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.55278714998414e-06,
+      "loss": 0.2285,
+      "step": 29590
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.552210398823428e-06,
+      "loss": 0.4639,
+      "step": 29591
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5516336476627162e-06,
+      "loss": 0.377,
+      "step": 29592
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5510568965020043e-06,
+      "loss": 0.3766,
+      "step": 29593
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.550480145341293e-06,
+      "loss": 0.3875,
+      "step": 29594
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.549903394180581e-06,
+      "loss": 0.4576,
+      "step": 29595
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.549326643019869e-06,
+      "loss": 0.3183,
+      "step": 29596
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.548749891859158e-06,
+      "loss": 0.3155,
+      "step": 29597
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.548173140698446e-06,
+      "loss": 0.4604,
+      "step": 29598
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5475963895377343e-06,
+      "loss": 0.3179,
+      "step": 29599
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.547019638377023e-06,
+      "loss": 0.5087,
+      "step": 29600
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.546442887216311e-06,
+      "loss": 0.3868,
+      "step": 29601
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.545866136055599e-06,
+      "loss": 0.3632,
+      "step": 29602
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.545289384894887e-06,
+      "loss": 0.4119,
+      "step": 29603
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5447126337341757e-06,
+      "loss": 0.3482,
+      "step": 29604
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5441358825734638e-06,
+      "loss": 0.3701,
+      "step": 29605
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5435591314127523e-06,
+      "loss": 0.2719,
+      "step": 29606
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.542982380252041e-06,
+      "loss": 0.4513,
+      "step": 29607
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.542405629091329e-06,
+      "loss": 0.3891,
+      "step": 29608
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.541828877930617e-06,
+      "loss": 0.3888,
+      "step": 29609
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.541252126769905e-06,
+      "loss": 0.4431,
+      "step": 29610
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5406753756091937e-06,
+      "loss": 0.3877,
+      "step": 29611
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.540098624448482e-06,
+      "loss": 0.3928,
+      "step": 29612
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.53952187328777e-06,
+      "loss": 0.3978,
+      "step": 29613
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.538945122127059e-06,
+      "loss": 0.458,
+      "step": 29614
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.538368370966347e-06,
+      "loss": 0.3074,
+      "step": 29615
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.537791619805635e-06,
+      "loss": 0.365,
+      "step": 29616
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5372148686449236e-06,
+      "loss": 0.3503,
+      "step": 29617
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5366381174842117e-06,
+      "loss": 0.473,
+      "step": 29618
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5360613663235e-06,
+      "loss": 0.3539,
+      "step": 29619
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.535484615162788e-06,
+      "loss": 0.5581,
+      "step": 29620
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.534907864002077e-06,
+      "loss": 0.4069,
+      "step": 29621
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.534331112841365e-06,
+      "loss": 0.5215,
+      "step": 29622
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.533754361680653e-06,
+      "loss": 0.4323,
+      "step": 29623
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5331776105199416e-06,
+      "loss": 0.5271,
+      "step": 29624
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5326008593592298e-06,
+      "loss": 0.319,
+      "step": 29625
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.532024108198518e-06,
+      "loss": 0.4762,
+      "step": 29626
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.531447357037806e-06,
+      "loss": 0.3311,
+      "step": 29627
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.530870605877095e-06,
+      "loss": 0.4196,
+      "step": 29628
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.530293854716383e-06,
+      "loss": 0.3551,
+      "step": 29629
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.529717103555671e-06,
+      "loss": 0.4602,
+      "step": 29630
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5291403523949597e-06,
+      "loss": 0.4415,
+      "step": 29631
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5285636012342478e-06,
+      "loss": 0.3295,
+      "step": 29632
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.527986850073536e-06,
+      "loss": 0.3573,
+      "step": 29633
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5274100989128244e-06,
+      "loss": 0.3108,
+      "step": 29634
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5268333477521125e-06,
+      "loss": 0.427,
+      "step": 29635
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5262565965914006e-06,
+      "loss": 0.3982,
+      "step": 29636
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.525679845430689e-06,
+      "loss": 0.4381,
+      "step": 29637
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5251030942699777e-06,
+      "loss": 0.5379,
+      "step": 29638
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.524526343109266e-06,
+      "loss": 0.4935,
+      "step": 29639
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.523949591948554e-06,
+      "loss": 0.4481,
+      "step": 29640
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5233728407878425e-06,
+      "loss": 0.5111,
+      "step": 29641
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5227960896271306e-06,
+      "loss": 0.3726,
+      "step": 29642
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5222193384664187e-06,
+      "loss": 0.3339,
+      "step": 29643
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5216425873057076e-06,
+      "loss": 0.464,
+      "step": 29644
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5210658361449957e-06,
+      "loss": 0.3933,
+      "step": 29645
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.520489084984284e-06,
+      "loss": 0.4781,
+      "step": 29646
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.519912333823572e-06,
+      "loss": 0.4731,
+      "step": 29647
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5193355826628605e-06,
+      "loss": 0.5186,
+      "step": 29648
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5187588315021486e-06,
+      "loss": 0.3475,
+      "step": 29649
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5181820803414367e-06,
+      "loss": 0.3104,
+      "step": 29650
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5176053291807257e-06,
+      "loss": 0.3739,
+      "step": 29651
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5170285780200138e-06,
+      "loss": 0.3958,
+      "step": 29652
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.516451826859302e-06,
+      "loss": 0.4019,
+      "step": 29653
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.51587507569859e-06,
+      "loss": 0.3319,
+      "step": 29654
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5152983245378785e-06,
+      "loss": 0.4196,
+      "step": 29655
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5147215733771666e-06,
+      "loss": 0.3074,
+      "step": 29656
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5141448222164547e-06,
+      "loss": 0.429,
+      "step": 29657
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5135680710557437e-06,
+      "loss": 0.4129,
+      "step": 29658
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.512991319895032e-06,
+      "loss": 0.4342,
+      "step": 29659
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.51241456873432e-06,
+      "loss": 0.383,
+      "step": 29660
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5118378175736084e-06,
+      "loss": 0.3825,
+      "step": 29661
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5112610664128965e-06,
+      "loss": 0.4246,
+      "step": 29662
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5106843152521847e-06,
+      "loss": 0.5594,
+      "step": 29663
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5101075640914728e-06,
+      "loss": 0.3611,
+      "step": 29664
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5095308129307613e-06,
+      "loss": 0.3203,
+      "step": 29665
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5089540617700494e-06,
+      "loss": 0.3766,
+      "step": 29666
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.508377310609338e-06,
+      "loss": 0.5065,
+      "step": 29667
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5078005594486265e-06,
+      "loss": 0.4617,
+      "step": 29668
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5072238082879146e-06,
+      "loss": 0.3935,
+      "step": 29669
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5066470571272027e-06,
+      "loss": 0.4002,
+      "step": 29670
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.506070305966491e-06,
+      "loss": 0.3459,
+      "step": 29671
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5054935548057793e-06,
+      "loss": 0.4035,
+      "step": 29672
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5049168036450674e-06,
+      "loss": 0.4676,
+      "step": 29673
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5043400524843555e-06,
+      "loss": 0.4982,
+      "step": 29674
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5037633013236445e-06,
+      "loss": 0.4731,
+      "step": 29675
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5031865501629326e-06,
+      "loss": 0.362,
+      "step": 29676
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5026097990022207e-06,
+      "loss": 0.4639,
+      "step": 29677
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5020330478415092e-06,
+      "loss": 0.3825,
+      "step": 29678
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5014562966807974e-06,
+      "loss": 0.357,
+      "step": 29679
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5008795455200855e-06,
+      "loss": 0.4659,
+      "step": 29680
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.5003027943593736e-06,
+      "loss": 0.5043,
+      "step": 29681
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4997260431986625e-06,
+      "loss": 0.3627,
+      "step": 29682
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4991492920379506e-06,
+      "loss": 0.5377,
+      "step": 29683
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4985725408772387e-06,
+      "loss": 0.4625,
+      "step": 29684
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4979957897165273e-06,
+      "loss": 0.5255,
+      "step": 29685
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4974190385558154e-06,
+      "loss": 0.3313,
+      "step": 29686
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4968422873951035e-06,
+      "loss": 0.3855,
+      "step": 29687
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4962655362343916e-06,
+      "loss": 0.4075,
+      "step": 29688
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4956887850736806e-06,
+      "loss": 0.3511,
+      "step": 29689
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4951120339129687e-06,
+      "loss": 0.2811,
+      "step": 29690
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4945352827522568e-06,
+      "loss": 0.4677,
+      "step": 29691
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4939585315915453e-06,
+      "loss": 0.3974,
+      "step": 29692
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4933817804308334e-06,
+      "loss": 0.4352,
+      "step": 29693
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4928050292701215e-06,
+      "loss": 0.4673,
+      "step": 29694
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.49222827810941e-06,
+      "loss": 0.4472,
+      "step": 29695
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.491651526948698e-06,
+      "loss": 0.4993,
+      "step": 29696
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4910747757879863e-06,
+      "loss": 0.4649,
+      "step": 29697
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.490498024627275e-06,
+      "loss": 0.5375,
+      "step": 29698
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4899212734665633e-06,
+      "loss": 0.3736,
+      "step": 29699
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4893445223058514e-06,
+      "loss": 0.3722,
+      "step": 29700
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4887677711451396e-06,
+      "loss": 0.427,
+      "step": 29701
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.488191019984428e-06,
+      "loss": 0.4713,
+      "step": 29702
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.487614268823716e-06,
+      "loss": 0.4311,
+      "step": 29703
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4870375176630043e-06,
+      "loss": 0.4932,
+      "step": 29704
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4864607665022933e-06,
+      "loss": 0.3352,
+      "step": 29705
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4858840153415814e-06,
+      "loss": 0.4477,
+      "step": 29706
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4853072641808695e-06,
+      "loss": 0.3247,
+      "step": 29707
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4847305130201576e-06,
+      "loss": 0.3343,
+      "step": 29708
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.484153761859446e-06,
+      "loss": 0.3948,
+      "step": 29709
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4835770106987342e-06,
+      "loss": 0.3406,
+      "step": 29710
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4830002595380223e-06,
+      "loss": 0.355,
+      "step": 29711
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4824235083773113e-06,
+      "loss": 0.4843,
+      "step": 29712
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4818467572165994e-06,
+      "loss": 0.3818,
+      "step": 29713
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4812700060558875e-06,
+      "loss": 0.4162,
+      "step": 29714
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4806932548951756e-06,
+      "loss": 0.3403,
+      "step": 29715
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.480116503734464e-06,
+      "loss": 0.3472,
+      "step": 29716
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4795397525737523e-06,
+      "loss": 0.4691,
+      "step": 29717
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4789630014130404e-06,
+      "loss": 0.3841,
+      "step": 29718
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4783862502523293e-06,
+      "loss": 0.3787,
+      "step": 29719
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4778094990916174e-06,
+      "loss": 0.4611,
+      "step": 29720
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4772327479309055e-06,
+      "loss": 0.3456,
+      "step": 29721
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.476655996770194e-06,
+      "loss": 0.4016,
+      "step": 29722
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.476079245609482e-06,
+      "loss": 0.4005,
+      "step": 29723
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4755024944487703e-06,
+      "loss": 0.4063,
+      "step": 29724
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4749257432880584e-06,
+      "loss": 0.4244,
+      "step": 29725
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.474348992127347e-06,
+      "loss": 0.3313,
+      "step": 29726
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.473772240966635e-06,
+      "loss": 0.4032,
+      "step": 29727
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.473195489805923e-06,
+      "loss": 0.3633,
+      "step": 29728
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.472618738645212e-06,
+      "loss": 0.3658,
+      "step": 29729
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4720419874845002e-06,
+      "loss": 0.4,
+      "step": 29730
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4714652363237883e-06,
+      "loss": 0.3021,
+      "step": 29731
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4708884851630764e-06,
+      "loss": 0.4554,
+      "step": 29732
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.470311734002365e-06,
+      "loss": 0.349,
+      "step": 29733
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.469734982841653e-06,
+      "loss": 0.5025,
+      "step": 29734
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.469158231680941e-06,
+      "loss": 0.3985,
+      "step": 29735
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.46858148052023e-06,
+      "loss": 0.4513,
+      "step": 29736
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4680047293595182e-06,
+      "loss": 0.4473,
+      "step": 29737
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4674279781988064e-06,
+      "loss": 0.3268,
+      "step": 29738
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.466851227038095e-06,
+      "loss": 0.3187,
+      "step": 29739
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.466274475877383e-06,
+      "loss": 0.4928,
+      "step": 29740
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.465697724716671e-06,
+      "loss": 0.4863,
+      "step": 29741
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.465120973555959e-06,
+      "loss": 0.4838,
+      "step": 29742
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.464544222395248e-06,
+      "loss": 0.3778,
+      "step": 29743
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4639674712345363e-06,
+      "loss": 0.3552,
+      "step": 29744
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4633907200738244e-06,
+      "loss": 0.4698,
+      "step": 29745
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.462813968913113e-06,
+      "loss": 0.3687,
+      "step": 29746
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.462237217752401e-06,
+      "loss": 0.3453,
+      "step": 29747
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.461660466591689e-06,
+      "loss": 0.3346,
+      "step": 29748
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4610837154309772e-06,
+      "loss": 0.3611,
+      "step": 29749
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.460506964270266e-06,
+      "loss": 0.3796,
+      "step": 29750
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4599302131095543e-06,
+      "loss": 0.4944,
+      "step": 29751
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4593534619488424e-06,
+      "loss": 0.3259,
+      "step": 29752
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.458776710788131e-06,
+      "loss": 0.3652,
+      "step": 29753
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.458199959627419e-06,
+      "loss": 0.3874,
+      "step": 29754
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.457623208466707e-06,
+      "loss": 0.3512,
+      "step": 29755
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4570464573059957e-06,
+      "loss": 0.4605,
+      "step": 29756
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.456469706145284e-06,
+      "loss": 0.5108,
+      "step": 29757
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.455892954984572e-06,
+      "loss": 0.3533,
+      "step": 29758
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.4553162038238604e-06,
+      "loss": 0.3093,
+      "step": 29759
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.454739452663149e-06,
+      "loss": 0.3855,
+      "step": 29760
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.454162701502437e-06,
+      "loss": 0.4697,
+      "step": 29761
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.453585950341725e-06,
+      "loss": 0.4524,
+      "step": 29762
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4530091991810137e-06,
+      "loss": 0.3634,
+      "step": 29763
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.452432448020302e-06,
+      "loss": 0.4666,
+      "step": 29764
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.45185569685959e-06,
+      "loss": 0.5294,
+      "step": 29765
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.451278945698878e-06,
+      "loss": 0.4146,
+      "step": 29766
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.450702194538167e-06,
+      "loss": 0.3536,
+      "step": 29767
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.450125443377455e-06,
+      "loss": 0.4033,
+      "step": 29768
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4495486922167432e-06,
+      "loss": 0.4925,
+      "step": 29769
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4489719410560318e-06,
+      "loss": 0.4256,
+      "step": 29770
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.44839518989532e-06,
+      "loss": 0.4603,
+      "step": 29771
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.447818438734608e-06,
+      "loss": 0.3663,
+      "step": 29772
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.447241687573897e-06,
+      "loss": 0.5373,
+      "step": 29773
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.446664936413185e-06,
+      "loss": 0.4071,
+      "step": 29774
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.446088185252473e-06,
+      "loss": 0.3999,
+      "step": 29775
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4455114340917613e-06,
+      "loss": 0.37,
+      "step": 29776
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.44493468293105e-06,
+      "loss": 0.3722,
+      "step": 29777
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.444357931770338e-06,
+      "loss": 0.3055,
+      "step": 29778
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.443781180609626e-06,
+      "loss": 0.3607,
+      "step": 29779
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.443204429448915e-06,
+      "loss": 0.3781,
+      "step": 29780
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.442627678288203e-06,
+      "loss": 0.3521,
+      "step": 29781
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.442050927127491e-06,
+      "loss": 0.3639,
+      "step": 29782
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4414741759667797e-06,
+      "loss": 0.3605,
+      "step": 29783
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.440897424806068e-06,
+      "loss": 0.3078,
+      "step": 29784
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.440320673645356e-06,
+      "loss": 0.4479,
+      "step": 29785
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.439743922484644e-06,
+      "loss": 0.4018,
+      "step": 29786
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4391671713239326e-06,
+      "loss": 0.3656,
+      "step": 29787
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4385904201632207e-06,
+      "loss": 0.4376,
+      "step": 29788
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4380136690025088e-06,
+      "loss": 0.2179,
+      "step": 29789
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4374369178417977e-06,
+      "loss": 0.4325,
+      "step": 29790
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.436860166681086e-06,
+      "loss": 0.3545,
+      "step": 29791
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.436283415520374e-06,
+      "loss": 0.5103,
+      "step": 29792
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.435706664359662e-06,
+      "loss": 0.4227,
+      "step": 29793
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4351299131989506e-06,
+      "loss": 0.3187,
+      "step": 29794
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4345531620382387e-06,
+      "loss": 0.4228,
+      "step": 29795
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.433976410877527e-06,
+      "loss": 0.3915,
+      "step": 29796
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4333996597168158e-06,
+      "loss": 0.3862,
+      "step": 29797
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.432822908556104e-06,
+      "loss": 0.4618,
+      "step": 29798
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.432246157395392e-06,
+      "loss": 0.4796,
+      "step": 29799
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4316694062346805e-06,
+      "loss": 0.3655,
+      "step": 29800
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4310926550739686e-06,
+      "loss": 0.4875,
+      "step": 29801
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4305159039132567e-06,
+      "loss": 0.5084,
+      "step": 29802
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.429939152752545e-06,
+      "loss": 0.358,
+      "step": 29803
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.429362401591834e-06,
+      "loss": 0.3811,
+      "step": 29804
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.428785650431122e-06,
+      "loss": 0.4994,
+      "step": 29805
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.42820889927041e-06,
+      "loss": 0.3459,
+      "step": 29806
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4276321481096986e-06,
+      "loss": 0.4808,
+      "step": 29807
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4270553969489867e-06,
+      "loss": 0.4366,
+      "step": 29808
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4264786457882748e-06,
+      "loss": 0.437,
+      "step": 29809
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.425901894627563e-06,
+      "loss": 0.465,
+      "step": 29810
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.425325143466852e-06,
+      "loss": 0.3079,
+      "step": 29811
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.42474839230614e-06,
+      "loss": 0.2898,
+      "step": 29812
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.424171641145428e-06,
+      "loss": 0.4563,
+      "step": 29813
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4235948899847166e-06,
+      "loss": 0.4954,
+      "step": 29814
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4230181388240047e-06,
+      "loss": 0.3329,
+      "step": 29815
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.422441387663293e-06,
+      "loss": 0.3675,
+      "step": 29816
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4218646365025813e-06,
+      "loss": 0.3571,
+      "step": 29817
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4212878853418694e-06,
+      "loss": 0.3445,
+      "step": 29818
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4207111341811576e-06,
+      "loss": 0.3778,
+      "step": 29819
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4201343830204457e-06,
+      "loss": 0.4525,
+      "step": 29820
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4195576318597346e-06,
+      "loss": 0.4075,
+      "step": 29821
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4189808806990227e-06,
+      "loss": 0.5066,
+      "step": 29822
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.418404129538311e-06,
+      "loss": 0.4818,
+      "step": 29823
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4178273783775994e-06,
+      "loss": 0.4658,
+      "step": 29824
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4172506272168875e-06,
+      "loss": 0.43,
+      "step": 29825
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4166738760561756e-06,
+      "loss": 0.4371,
+      "step": 29826
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4160971248954637e-06,
+      "loss": 0.3549,
+      "step": 29827
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4155203737347526e-06,
+      "loss": 0.3564,
+      "step": 29828
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4149436225740408e-06,
+      "loss": 0.3393,
+      "step": 29829
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.414366871413329e-06,
+      "loss": 0.439,
+      "step": 29830
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4137901202526174e-06,
+      "loss": 0.4187,
+      "step": 29831
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4132133690919055e-06,
+      "loss": 0.3941,
+      "step": 29832
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4126366179311936e-06,
+      "loss": 0.41,
+      "step": 29833
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4120598667704826e-06,
+      "loss": 0.3266,
+      "step": 29834
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4114831156097707e-06,
+      "loss": 0.3558,
+      "step": 29835
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4109063644490588e-06,
+      "loss": 0.4554,
+      "step": 29836
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.410329613288347e-06,
+      "loss": 0.3175,
+      "step": 29837
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4097528621276354e-06,
+      "loss": 0.3036,
+      "step": 29838
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4091761109669235e-06,
+      "loss": 0.3111,
+      "step": 29839
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4085993598062116e-06,
+      "loss": 0.3706,
+      "step": 29840
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4080226086455e-06,
+      "loss": 0.3861,
+      "step": 29841
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4074458574847887e-06,
+      "loss": 0.5778,
+      "step": 29842
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.406869106324077e-06,
+      "loss": 0.5457,
+      "step": 29843
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4062923551633653e-06,
+      "loss": 0.3932,
+      "step": 29844
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4057156040026535e-06,
+      "loss": 0.2796,
+      "step": 29845
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4051388528419416e-06,
+      "loss": 0.384,
+      "step": 29846
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4045621016812297e-06,
+      "loss": 0.4811,
+      "step": 29847
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.403985350520518e-06,
+      "loss": 0.4295,
+      "step": 29848
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4034085993598063e-06,
+      "loss": 0.3712,
+      "step": 29849
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4028318481990944e-06,
+      "loss": 0.3157,
+      "step": 29850
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4022550970383834e-06,
+      "loss": 0.381,
+      "step": 29851
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4016783458776715e-06,
+      "loss": 0.3486,
+      "step": 29852
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4011015947169596e-06,
+      "loss": 0.4418,
+      "step": 29853
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.4005248435562477e-06,
+      "loss": 0.4278,
+      "step": 29854
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3999480923955362e-06,
+      "loss": 0.3955,
+      "step": 29855
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3993713412348243e-06,
+      "loss": 0.4383,
+      "step": 29856
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3987945900741125e-06,
+      "loss": 0.3494,
+      "step": 29857
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3982178389134014e-06,
+      "loss": 0.4418,
+      "step": 29858
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3976410877526895e-06,
+      "loss": 0.4312,
+      "step": 29859
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3970643365919776e-06,
+      "loss": 0.3392,
+      "step": 29860
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.396487585431266e-06,
+      "loss": 0.3923,
+      "step": 29861
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3959108342705543e-06,
+      "loss": 0.4909,
+      "step": 29862
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3953340831098424e-06,
+      "loss": 0.3644,
+      "step": 29863
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3947573319491305e-06,
+      "loss": 0.4191,
+      "step": 29864
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3941805807884194e-06,
+      "loss": 0.4209,
+      "step": 29865
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3936038296277075e-06,
+      "loss": 0.5034,
+      "step": 29866
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3930270784669957e-06,
+      "loss": 0.5499,
+      "step": 29867
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.392450327306284e-06,
+      "loss": 0.4906,
+      "step": 29868
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3918735761455723e-06,
+      "loss": 0.5695,
+      "step": 29869
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3912968249848604e-06,
+      "loss": 0.4519,
+      "step": 29870
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3907200738241485e-06,
+      "loss": 0.363,
+      "step": 29871
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3901433226634375e-06,
+      "loss": 0.4968,
+      "step": 29872
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3895665715027256e-06,
+      "loss": 0.4147,
+      "step": 29873
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3889898203420137e-06,
+      "loss": 0.5658,
+      "step": 29874
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3884130691813022e-06,
+      "loss": 0.4425,
+      "step": 29875
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3878363180205903e-06,
+      "loss": 0.3609,
+      "step": 29876
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3872595668598784e-06,
+      "loss": 0.4437,
+      "step": 29877
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.386682815699167e-06,
+      "loss": 0.4453,
+      "step": 29878
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.386106064538455e-06,
+      "loss": 0.3245,
+      "step": 29879
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.385529313377743e-06,
+      "loss": 0.3313,
+      "step": 29880
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3849525622170313e-06,
+      "loss": 0.2971,
+      "step": 29881
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3843758110563203e-06,
+      "loss": 0.4325,
+      "step": 29882
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3837990598956084e-06,
+      "loss": 0.4568,
+      "step": 29883
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3832223087348965e-06,
+      "loss": 0.3119,
+      "step": 29884
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.382645557574185e-06,
+      "loss": 0.5213,
+      "step": 29885
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.382068806413473e-06,
+      "loss": 0.3048,
+      "step": 29886
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3814920552527612e-06,
+      "loss": 0.4457,
+      "step": 29887
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3809153040920493e-06,
+      "loss": 0.3619,
+      "step": 29888
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3803385529313383e-06,
+      "loss": 0.4709,
+      "step": 29889
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3797618017706264e-06,
+      "loss": 0.3992,
+      "step": 29890
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3791850506099145e-06,
+      "loss": 0.3639,
+      "step": 29891
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.378608299449203e-06,
+      "loss": 0.3465,
+      "step": 29892
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.378031548288491e-06,
+      "loss": 0.3156,
+      "step": 29893
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3774547971277792e-06,
+      "loss": 0.4085,
+      "step": 29894
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.376878045967068e-06,
+      "loss": 0.4485,
+      "step": 29895
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3763012948063563e-06,
+      "loss": 0.3336,
+      "step": 29896
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3757245436456444e-06,
+      "loss": 0.329,
+      "step": 29897
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3751477924849325e-06,
+      "loss": 0.4656,
+      "step": 29898
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.374571041324221e-06,
+      "loss": 0.3182,
+      "step": 29899
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.373994290163509e-06,
+      "loss": 0.4436,
+      "step": 29900
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3734175390027973e-06,
+      "loss": 0.4216,
+      "step": 29901
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.372840787842086e-06,
+      "loss": 0.3485,
+      "step": 29902
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3722640366813743e-06,
+      "loss": 0.3649,
+      "step": 29903
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3716872855206625e-06,
+      "loss": 0.3972,
+      "step": 29904
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.371110534359951e-06,
+      "loss": 0.4897,
+      "step": 29905
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.370533783199239e-06,
+      "loss": 0.3641,
+      "step": 29906
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.369957032038527e-06,
+      "loss": 0.3838,
+      "step": 29907
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3693802808778153e-06,
+      "loss": 0.4516,
+      "step": 29908
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.368803529717104e-06,
+      "loss": 0.3636,
+      "step": 29909
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.368226778556392e-06,
+      "loss": 0.3954,
+      "step": 29910
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.36765002739568e-06,
+      "loss": 0.4824,
+      "step": 29911
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.367073276234969e-06,
+      "loss": 0.5011,
+      "step": 29912
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.366496525074257e-06,
+      "loss": 0.5084,
+      "step": 29913
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3659197739135452e-06,
+      "loss": 0.6036,
+      "step": 29914
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3653430227528333e-06,
+      "loss": 0.3941,
+      "step": 29915
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.364766271592122e-06,
+      "loss": 0.3969,
+      "step": 29916
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.36418952043141e-06,
+      "loss": 0.515,
+      "step": 29917
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.363612769270698e-06,
+      "loss": 0.4587,
+      "step": 29918
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.363036018109987e-06,
+      "loss": 0.4919,
+      "step": 29919
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.362459266949275e-06,
+      "loss": 0.4131,
+      "step": 29920
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3618825157885633e-06,
+      "loss": 0.2669,
+      "step": 29921
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.361305764627852e-06,
+      "loss": 0.4389,
+      "step": 29922
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.36072901346714e-06,
+      "loss": 0.368,
+      "step": 29923
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.360152262306428e-06,
+      "loss": 0.4267,
+      "step": 29924
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.359575511145716e-06,
+      "loss": 0.4209,
+      "step": 29925
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.358998759985005e-06,
+      "loss": 0.4289,
+      "step": 29926
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.358422008824293e-06,
+      "loss": 0.4203,
+      "step": 29927
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3578452576635813e-06,
+      "loss": 0.4656,
+      "step": 29928
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.35726850650287e-06,
+      "loss": 0.3417,
+      "step": 29929
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.356691755342158e-06,
+      "loss": 0.4306,
+      "step": 29930
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.356115004181446e-06,
+      "loss": 0.3425,
+      "step": 29931
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.355538253020734e-06,
+      "loss": 0.4564,
+      "step": 29932
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3549615018600227e-06,
+      "loss": 0.4608,
+      "step": 29933
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3543847506993112e-06,
+      "loss": 0.4709,
+      "step": 29934
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3538079995385993e-06,
+      "loss": 0.3302,
+      "step": 29935
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.353231248377888e-06,
+      "loss": 0.4456,
+      "step": 29936
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.352654497217176e-06,
+      "loss": 0.4605,
+      "step": 29937
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.352077746056464e-06,
+      "loss": 0.3505,
+      "step": 29938
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3515009948957526e-06,
+      "loss": 0.3045,
+      "step": 29939
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 3.3509242437350407e-06,
+      "loss": 0.34,
+      "step": 29940
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.350347492574329e-06,
+      "loss": 0.4298,
+      "step": 29941
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.349770741413617e-06,
+      "loss": 0.3487,
+      "step": 29942
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.349193990252906e-06,
+      "loss": 0.434,
+      "step": 29943
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.348617239092194e-06,
+      "loss": 0.4026,
+      "step": 29944
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.348040487931482e-06,
+      "loss": 0.396,
+      "step": 29945
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3474637367707706e-06,
+      "loss": 0.4388,
+      "step": 29946
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3468869856100587e-06,
+      "loss": 0.4377,
+      "step": 29947
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.346310234449347e-06,
+      "loss": 0.4492,
+      "step": 29948
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.345733483288635e-06,
+      "loss": 0.4428,
+      "step": 29949
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.345156732127924e-06,
+      "loss": 0.3234,
+      "step": 29950
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.344579980967212e-06,
+      "loss": 0.3248,
+      "step": 29951
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3440032298065e-06,
+      "loss": 0.4194,
+      "step": 29952
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3434264786457887e-06,
+      "loss": 0.4487,
+      "step": 29953
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3428497274850768e-06,
+      "loss": 0.3967,
+      "step": 29954
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.342272976324365e-06,
+      "loss": 0.4536,
+      "step": 29955
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.341696225163654e-06,
+      "loss": 0.344,
+      "step": 29956
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.341119474002942e-06,
+      "loss": 0.4393,
+      "step": 29957
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.34054272284223e-06,
+      "loss": 0.4241,
+      "step": 29958
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.339965971681518e-06,
+      "loss": 0.4358,
+      "step": 29959
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3393892205208067e-06,
+      "loss": 0.3415,
+      "step": 29960
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.338812469360095e-06,
+      "loss": 0.3295,
+      "step": 29961
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.338235718199383e-06,
+      "loss": 0.545,
+      "step": 29962
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3376589670386715e-06,
+      "loss": 0.3683,
+      "step": 29963
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.33708221587796e-06,
+      "loss": 0.3581,
+      "step": 29964
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.336505464717248e-06,
+      "loss": 0.3294,
+      "step": 29965
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3359287135565366e-06,
+      "loss": 0.3946,
+      "step": 29966
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3353519623958247e-06,
+      "loss": 0.4687,
+      "step": 29967
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.334775211235113e-06,
+      "loss": 0.3568,
+      "step": 29968
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.334198460074401e-06,
+      "loss": 0.3127,
+      "step": 29969
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3336217089136895e-06,
+      "loss": 0.3224,
+      "step": 29970
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3330449577529776e-06,
+      "loss": 0.3633,
+      "step": 29971
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3324682065922657e-06,
+      "loss": 0.3182,
+      "step": 29972
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3318914554315547e-06,
+      "loss": 0.5196,
+      "step": 29973
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3313147042708428e-06,
+      "loss": 0.3789,
+      "step": 29974
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.330737953110131e-06,
+      "loss": 0.412,
+      "step": 29975
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.330161201949419e-06,
+      "loss": 0.4304,
+      "step": 29976
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3295844507887075e-06,
+      "loss": 0.4883,
+      "step": 29977
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3290076996279956e-06,
+      "loss": 0.3904,
+      "step": 29978
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3284309484672837e-06,
+      "loss": 0.3158,
+      "step": 29979
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3278541973065727e-06,
+      "loss": 0.4682,
+      "step": 29980
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.327277446145861e-06,
+      "loss": 0.4579,
+      "step": 29981
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.326700694985149e-06,
+      "loss": 0.4456,
+      "step": 29982
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3261239438244374e-06,
+      "loss": 0.408,
+      "step": 29983
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3255471926637255e-06,
+      "loss": 0.4108,
+      "step": 29984
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3249704415030137e-06,
+      "loss": 0.3176,
+      "step": 29985
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3243936903423018e-06,
+      "loss": 0.3169,
+      "step": 29986
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3238169391815907e-06,
+      "loss": 0.3522,
+      "step": 29987
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.323240188020879e-06,
+      "loss": 0.4395,
+      "step": 29988
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.322663436860167e-06,
+      "loss": 0.4675,
+      "step": 29989
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3220866856994555e-06,
+      "loss": 0.4504,
+      "step": 29990
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3215099345387436e-06,
+      "loss": 0.4808,
+      "step": 29991
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3209331833780317e-06,
+      "loss": 0.4205,
+      "step": 29992
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.32035643221732e-06,
+      "loss": 0.5016,
+      "step": 29993
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3197796810566083e-06,
+      "loss": 0.449,
+      "step": 29994
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.319202929895897e-06,
+      "loss": 0.4068,
+      "step": 29995
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.318626178735185e-06,
+      "loss": 0.3606,
+      "step": 29996
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3180494275744735e-06,
+      "loss": 0.4063,
+      "step": 29997
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3174726764137616e-06,
+      "loss": 0.4869,
+      "step": 29998
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3168959252530497e-06,
+      "loss": 0.3558,
+      "step": 29999
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3163191740923382e-06,
+      "loss": 0.3255,
+      "step": 30000
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3157424229316264e-06,
+      "loss": 0.3977,
+      "step": 30001
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3151656717709145e-06,
+      "loss": 0.2957,
+      "step": 30002
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3145889206102026e-06,
+      "loss": 0.2884,
+      "step": 30003
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3140121694494915e-06,
+      "loss": 0.4152,
+      "step": 30004
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3134354182887796e-06,
+      "loss": 0.2894,
+      "step": 30005
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3128586671280677e-06,
+      "loss": 0.3966,
+      "step": 30006
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3122819159673563e-06,
+      "loss": 0.3482,
+      "step": 30007
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3117051648066444e-06,
+      "loss": 0.3179,
+      "step": 30008
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3111284136459325e-06,
+      "loss": 0.4042,
+      "step": 30009
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3105516624852206e-06,
+      "loss": 0.472,
+      "step": 30010
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3099749113245096e-06,
+      "loss": 0.2838,
+      "step": 30011
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3093981601637977e-06,
+      "loss": 0.335,
+      "step": 30012
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3088214090030858e-06,
+      "loss": 0.4268,
+      "step": 30013
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3082446578423743e-06,
+      "loss": 0.4901,
+      "step": 30014
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3076679066816624e-06,
+      "loss": 0.405,
+      "step": 30015
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3070911555209505e-06,
+      "loss": 0.4636,
+      "step": 30016
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3065144043602395e-06,
+      "loss": 0.3273,
+      "step": 30017
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3059376531995276e-06,
+      "loss": 0.2814,
+      "step": 30018
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3053609020388157e-06,
+      "loss": 0.372,
+      "step": 30019
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.304784150878104e-06,
+      "loss": 0.3515,
+      "step": 30020
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3042073997173923e-06,
+      "loss": 0.4558,
+      "step": 30021
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3036306485566804e-06,
+      "loss": 0.4486,
+      "step": 30022
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3030538973959686e-06,
+      "loss": 0.2801,
+      "step": 30023
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.302477146235257e-06,
+      "loss": 0.4863,
+      "step": 30024
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.301900395074545e-06,
+      "loss": 0.4027,
+      "step": 30025
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3013236439138337e-06,
+      "loss": 0.4332,
+      "step": 30026
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3007468927531223e-06,
+      "loss": 0.4284,
+      "step": 30027
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.3001701415924104e-06,
+      "loss": 0.3262,
+      "step": 30028
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2995933904316985e-06,
+      "loss": 0.4337,
+      "step": 30029
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2990166392709866e-06,
+      "loss": 0.4396,
+      "step": 30030
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.298439888110275e-06,
+      "loss": 0.531,
+      "step": 30031
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2978631369495632e-06,
+      "loss": 0.3747,
+      "step": 30032
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2972863857888513e-06,
+      "loss": 0.3391,
+      "step": 30033
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2967096346281403e-06,
+      "loss": 0.3105,
+      "step": 30034
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2961328834674284e-06,
+      "loss": 0.4133,
+      "step": 30035
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2955561323067165e-06,
+      "loss": 0.4245,
+      "step": 30036
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2949793811460046e-06,
+      "loss": 0.4837,
+      "step": 30037
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.294402629985293e-06,
+      "loss": 0.4947,
+      "step": 30038
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2938258788245813e-06,
+      "loss": 0.3536,
+      "step": 30039
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2932491276638694e-06,
+      "loss": 0.3491,
+      "step": 30040
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2926723765031583e-06,
+      "loss": 0.2762,
+      "step": 30041
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2920956253424464e-06,
+      "loss": 0.3777,
+      "step": 30042
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2915188741817345e-06,
+      "loss": 0.3615,
+      "step": 30043
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.290942123021023e-06,
+      "loss": 0.4677,
+      "step": 30044
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.290365371860311e-06,
+      "loss": 0.3518,
+      "step": 30045
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2897886206995993e-06,
+      "loss": 0.4684,
+      "step": 30046
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2892118695388874e-06,
+      "loss": 0.3612,
+      "step": 30047
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2886351183781764e-06,
+      "loss": 0.2953,
+      "step": 30048
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2880583672174645e-06,
+      "loss": 0.3918,
+      "step": 30049
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2874816160567526e-06,
+      "loss": 0.3421,
+      "step": 30050
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.286904864896041e-06,
+      "loss": 0.3858,
+      "step": 30051
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.286328113735329e-06,
+      "loss": 0.4105,
+      "step": 30052
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2857513625746173e-06,
+      "loss": 0.3507,
+      "step": 30053
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2851746114139054e-06,
+      "loss": 0.5002,
+      "step": 30054
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.284597860253194e-06,
+      "loss": 0.4981,
+      "step": 30055
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2840211090924825e-06,
+      "loss": 0.2998,
+      "step": 30056
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2834443579317706e-06,
+      "loss": 0.4569,
+      "step": 30057
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.282867606771059e-06,
+      "loss": 0.363,
+      "step": 30058
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2822908556103472e-06,
+      "loss": 0.2451,
+      "step": 30059
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2817141044496354e-06,
+      "loss": 0.5123,
+      "step": 30060
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.281137353288924e-06,
+      "loss": 0.4005,
+      "step": 30061
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.280560602128212e-06,
+      "loss": 0.4838,
+      "step": 30062
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2799838509675e-06,
+      "loss": 0.3993,
+      "step": 30063
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.279407099806788e-06,
+      "loss": 0.4916,
+      "step": 30064
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.278830348646077e-06,
+      "loss": 0.3889,
+      "step": 30065
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2782535974853653e-06,
+      "loss": 0.4555,
+      "step": 30066
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2776768463246534e-06,
+      "loss": 0.3573,
+      "step": 30067
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.277100095163942e-06,
+      "loss": 0.4151,
+      "step": 30068
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.27652334400323e-06,
+      "loss": 0.483,
+      "step": 30069
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.275946592842518e-06,
+      "loss": 0.4964,
+      "step": 30070
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2753698416818062e-06,
+      "loss": 0.468,
+      "step": 30071
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.274793090521095e-06,
+      "loss": 0.3942,
+      "step": 30072
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2742163393603833e-06,
+      "loss": 0.3675,
+      "step": 30073
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2736395881996714e-06,
+      "loss": 0.434,
+      "step": 30074
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.27306283703896e-06,
+      "loss": 0.4693,
+      "step": 30075
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.272486085878248e-06,
+      "loss": 0.39,
+      "step": 30076
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.271909334717536e-06,
+      "loss": 0.4156,
+      "step": 30077
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.271332583556825e-06,
+      "loss": 0.3235,
+      "step": 30078
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2707558323961132e-06,
+      "loss": 0.3666,
+      "step": 30079
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2701790812354013e-06,
+      "loss": 0.381,
+      "step": 30080
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2696023300746894e-06,
+      "loss": 0.4429,
+      "step": 30081
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.269025578913978e-06,
+      "loss": 0.3784,
+      "step": 30082
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.268448827753266e-06,
+      "loss": 0.418,
+      "step": 30083
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.267872076592554e-06,
+      "loss": 0.4973,
+      "step": 30084
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2672953254318427e-06,
+      "loss": 0.3831,
+      "step": 30085
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.266718574271131e-06,
+      "loss": 0.3406,
+      "step": 30086
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2661418231104194e-06,
+      "loss": 0.3319,
+      "step": 30087
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.265565071949708e-06,
+      "loss": 0.4382,
+      "step": 30088
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.264988320788996e-06,
+      "loss": 0.4052,
+      "step": 30089
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.264411569628284e-06,
+      "loss": 0.3999,
+      "step": 30090
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2638348184675722e-06,
+      "loss": 0.4337,
+      "step": 30091
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2632580673068608e-06,
+      "loss": 0.4211,
+      "step": 30092
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.262681316146149e-06,
+      "loss": 0.3762,
+      "step": 30093
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.262104564985437e-06,
+      "loss": 0.3076,
+      "step": 30094
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.261527813824726e-06,
+      "loss": 0.334,
+      "step": 30095
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.260951062664014e-06,
+      "loss": 0.3846,
+      "step": 30096
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.260374311503302e-06,
+      "loss": 0.4575,
+      "step": 30097
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2597975603425903e-06,
+      "loss": 0.5827,
+      "step": 30098
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2592208091818788e-06,
+      "loss": 0.5996,
+      "step": 30099
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.258644058021167e-06,
+      "loss": 0.3132,
+      "step": 30100
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.258067306860455e-06,
+      "loss": 0.5843,
+      "step": 30101
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.257490555699744e-06,
+      "loss": 0.2832,
+      "step": 30102
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.256913804539032e-06,
+      "loss": 0.5532,
+      "step": 30103
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.25633705337832e-06,
+      "loss": 0.5042,
+      "step": 30104
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2557603022176087e-06,
+      "loss": 0.333,
+      "step": 30105
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.255183551056897e-06,
+      "loss": 0.3453,
+      "step": 30106
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.254606799896185e-06,
+      "loss": 0.3886,
+      "step": 30107
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.254030048735473e-06,
+      "loss": 0.444,
+      "step": 30108
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.253453297574762e-06,
+      "loss": 0.3991,
+      "step": 30109
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.25287654641405e-06,
+      "loss": 0.4803,
+      "step": 30110
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.252299795253338e-06,
+      "loss": 0.4295,
+      "step": 30111
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2517230440926267e-06,
+      "loss": 0.397,
+      "step": 30112
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.251146292931915e-06,
+      "loss": 0.3337,
+      "step": 30113
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.250569541771203e-06,
+      "loss": 0.3833,
+      "step": 30114
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.249992790610491e-06,
+      "loss": 0.4751,
+      "step": 30115
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2494160394497796e-06,
+      "loss": 0.3582,
+      "step": 30116
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2488392882890677e-06,
+      "loss": 0.3994,
+      "step": 30117
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2482625371283562e-06,
+      "loss": 0.3429,
+      "step": 30118
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.2476857859676448e-06,
+      "loss": 0.3221,
+      "step": 30119
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.247109034806933e-06,
+      "loss": 0.5001,
+      "step": 30120
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.246532283646221e-06,
+      "loss": 0.4911,
+      "step": 30121
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2459555324855095e-06,
+      "loss": 0.3465,
+      "step": 30122
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2453787813247976e-06,
+      "loss": 0.4266,
+      "step": 30123
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2448020301640857e-06,
+      "loss": 0.3712,
+      "step": 30124
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.244225279003374e-06,
+      "loss": 0.3398,
+      "step": 30125
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.243648527842663e-06,
+      "loss": 0.4136,
+      "step": 30126
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.243071776681951e-06,
+      "loss": 0.4417,
+      "step": 30127
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.242495025521239e-06,
+      "loss": 0.4136,
+      "step": 30128
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2419182743605276e-06,
+      "loss": 0.5746,
+      "step": 30129
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2413415231998157e-06,
+      "loss": 0.3545,
+      "step": 30130
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2407647720391038e-06,
+      "loss": 0.416,
+      "step": 30131
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.240188020878392e-06,
+      "loss": 0.3943,
+      "step": 30132
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.239611269717681e-06,
+      "loss": 0.4639,
+      "step": 30133
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.239034518556969e-06,
+      "loss": 0.3631,
+      "step": 30134
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.238457767396257e-06,
+      "loss": 0.3677,
+      "step": 30135
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2378810162355456e-06,
+      "loss": 0.4003,
+      "step": 30136
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2373042650748337e-06,
+      "loss": 0.4396,
+      "step": 30137
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.236727513914122e-06,
+      "loss": 0.5137,
+      "step": 30138
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2361507627534108e-06,
+      "loss": 0.3623,
+      "step": 30139
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.235574011592699e-06,
+      "loss": 0.4658,
+      "step": 30140
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.234997260431987e-06,
+      "loss": 0.3317,
+      "step": 30141
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.234420509271275e-06,
+      "loss": 0.3354,
+      "step": 30142
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2338437581105636e-06,
+      "loss": 0.4847,
+      "step": 30143
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2332670069498517e-06,
+      "loss": 0.5098,
+      "step": 30144
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.23269025578914e-06,
+      "loss": 0.5023,
+      "step": 30145
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2321135046284284e-06,
+      "loss": 0.2468,
+      "step": 30146
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2315367534677165e-06,
+      "loss": 0.4104,
+      "step": 30147
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.230960002307005e-06,
+      "loss": 0.4295,
+      "step": 30148
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.230383251146293e-06,
+      "loss": 0.5678,
+      "step": 30149
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2298064999855816e-06,
+      "loss": 0.4068,
+      "step": 30150
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2292297488248698e-06,
+      "loss": 0.3863,
+      "step": 30151
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.228652997664158e-06,
+      "loss": 0.4417,
+      "step": 30152
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2280762465034464e-06,
+      "loss": 0.4342,
+      "step": 30153
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2274994953427345e-06,
+      "loss": 0.3814,
+      "step": 30154
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2269227441820226e-06,
+      "loss": 0.4446,
+      "step": 30155
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2263459930213116e-06,
+      "loss": 0.3436,
+      "step": 30156
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2257692418605997e-06,
+      "loss": 0.4987,
+      "step": 30157
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2251924906998878e-06,
+      "loss": 0.3862,
+      "step": 30158
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.224615739539176e-06,
+      "loss": 0.3943,
+      "step": 30159
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2240389883784644e-06,
+      "loss": 0.342,
+      "step": 30160
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2234622372177525e-06,
+      "loss": 0.4838,
+      "step": 30161
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2228854860570406e-06,
+      "loss": 0.4056,
+      "step": 30162
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2223087348963296e-06,
+      "loss": 0.2889,
+      "step": 30163
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2217319837356177e-06,
+      "loss": 0.3152,
+      "step": 30164
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.221155232574906e-06,
+      "loss": 0.367,
+      "step": 30165
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2205784814141943e-06,
+      "loss": 0.4108,
+      "step": 30166
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2200017302534825e-06,
+      "loss": 0.3908,
+      "step": 30167
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2194249790927706e-06,
+      "loss": 0.3536,
+      "step": 30168
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2188482279320587e-06,
+      "loss": 0.3794,
+      "step": 30169
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2182714767713476e-06,
+      "loss": 0.3665,
+      "step": 30170
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2176947256106357e-06,
+      "loss": 0.4908,
+      "step": 30171
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.217117974449924e-06,
+      "loss": 0.414,
+      "step": 30172
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2165412232892124e-06,
+      "loss": 0.6025,
+      "step": 30173
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2159644721285005e-06,
+      "loss": 0.4328,
+      "step": 30174
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2153877209677886e-06,
+      "loss": 0.423,
+      "step": 30175
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2148109698070767e-06,
+      "loss": 0.3535,
+      "step": 30176
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2142342186463652e-06,
+      "loss": 0.4682,
+      "step": 30177
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2136574674856533e-06,
+      "loss": 0.4263,
+      "step": 30178
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.213080716324942e-06,
+      "loss": 0.3628,
+      "step": 30179
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2125039651642304e-06,
+      "loss": 0.3379,
+      "step": 30180
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2119272140035185e-06,
+      "loss": 0.3416,
+      "step": 30181
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2113504628428066e-06,
+      "loss": 0.3787,
+      "step": 30182
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.210773711682095e-06,
+      "loss": 0.4348,
+      "step": 30183
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2101969605213833e-06,
+      "loss": 0.5147,
+      "step": 30184
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2096202093606714e-06,
+      "loss": 0.381,
+      "step": 30185
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2090434581999595e-06,
+      "loss": 0.3585,
+      "step": 30186
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2084667070392484e-06,
+      "loss": 0.5055,
+      "step": 30187
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2078899558785365e-06,
+      "loss": 0.355,
+      "step": 30188
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2073132047178247e-06,
+      "loss": 0.3705,
+      "step": 30189
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.206736453557113e-06,
+      "loss": 0.4208,
+      "step": 30190
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2061597023964013e-06,
+      "loss": 0.4289,
+      "step": 30191
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2055829512356894e-06,
+      "loss": 0.4845,
+      "step": 30192
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2050062000749775e-06,
+      "loss": 0.3422,
+      "step": 30193
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2044294489142665e-06,
+      "loss": 0.4596,
+      "step": 30194
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2038526977535546e-06,
+      "loss": 0.4378,
+      "step": 30195
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2032759465928427e-06,
+      "loss": 0.4033,
+      "step": 30196
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2026991954321312e-06,
+      "loss": 0.425,
+      "step": 30197
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2021224442714193e-06,
+      "loss": 0.4042,
+      "step": 30198
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2015456931107074e-06,
+      "loss": 0.4364,
+      "step": 30199
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2009689419499964e-06,
+      "loss": 0.3129,
+      "step": 30200
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.2003921907892845e-06,
+      "loss": 0.4515,
+      "step": 30201
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1998154396285726e-06,
+      "loss": 0.3082,
+      "step": 30202
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1992386884678607e-06,
+      "loss": 0.3495,
+      "step": 30203
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1986619373071492e-06,
+      "loss": 0.4213,
+      "step": 30204
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1980851861464374e-06,
+      "loss": 0.4109,
+      "step": 30205
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1975084349857255e-06,
+      "loss": 0.3748,
+      "step": 30206
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.196931683825014e-06,
+      "loss": 0.363,
+      "step": 30207
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.196354932664302e-06,
+      "loss": 0.3962,
+      "step": 30208
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1957781815035902e-06,
+      "loss": 0.2975,
+      "step": 30209
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1952014303428787e-06,
+      "loss": 0.3421,
+      "step": 30210
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1946246791821673e-06,
+      "loss": 0.4062,
+      "step": 30211
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1940479280214554e-06,
+      "loss": 0.4251,
+      "step": 30212
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1934711768607435e-06,
+      "loss": 0.4059,
+      "step": 30213
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.192894425700032e-06,
+      "loss": 0.4533,
+      "step": 30214
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.19231767453932e-06,
+      "loss": 0.4415,
+      "step": 30215
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1917409233786082e-06,
+      "loss": 0.5351,
+      "step": 30216
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.191164172217897e-06,
+      "loss": 0.3759,
+      "step": 30217
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1905874210571853e-06,
+      "loss": 0.3982,
+      "step": 30218
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1900106698964734e-06,
+      "loss": 0.4001,
+      "step": 30219
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1894339187357615e-06,
+      "loss": 0.3582,
+      "step": 30220
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.18885716757505e-06,
+      "loss": 0.398,
+      "step": 30221
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.188280416414338e-06,
+      "loss": 0.4794,
+      "step": 30222
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1877036652536263e-06,
+      "loss": 0.4188,
+      "step": 30223
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1871269140929152e-06,
+      "loss": 0.3402,
+      "step": 30224
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1865501629322033e-06,
+      "loss": 0.4559,
+      "step": 30225
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1859734117714915e-06,
+      "loss": 0.288,
+      "step": 30226
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.18539666061078e-06,
+      "loss": 0.4896,
+      "step": 30227
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.184819909450068e-06,
+      "loss": 0.4147,
+      "step": 30228
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.184243158289356e-06,
+      "loss": 0.3648,
+      "step": 30229
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1836664071286443e-06,
+      "loss": 0.4218,
+      "step": 30230
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1830896559679333e-06,
+      "loss": 0.4768,
+      "step": 30231
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1825129048072214e-06,
+      "loss": 0.3283,
+      "step": 30232
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1819361536465095e-06,
+      "loss": 0.386,
+      "step": 30233
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.181359402485798e-06,
+      "loss": 0.4452,
+      "step": 30234
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.180782651325086e-06,
+      "loss": 0.2846,
+      "step": 30235
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1802059001643742e-06,
+      "loss": 0.3778,
+      "step": 30236
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1796291490036623e-06,
+      "loss": 0.3496,
+      "step": 30237
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.179052397842951e-06,
+      "loss": 0.3549,
+      "step": 30238
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.178475646682239e-06,
+      "loss": 0.3433,
+      "step": 30239
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1778988955215275e-06,
+      "loss": 0.426,
+      "step": 30240
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.177322144360816e-06,
+      "loss": 0.5416,
+      "step": 30241
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.176745393200104e-06,
+      "loss": 0.376,
+      "step": 30242
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1761686420393923e-06,
+      "loss": 0.3542,
+      "step": 30243
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.175591890878681e-06,
+      "loss": 0.504,
+      "step": 30244
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.175015139717969e-06,
+      "loss": 0.3729,
+      "step": 30245
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.174438388557257e-06,
+      "loss": 0.4007,
+      "step": 30246
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.173861637396545e-06,
+      "loss": 0.3918,
+      "step": 30247
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.173284886235834e-06,
+      "loss": 0.3172,
+      "step": 30248
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.172708135075122e-06,
+      "loss": 0.4433,
+      "step": 30249
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1721313839144103e-06,
+      "loss": 0.2783,
+      "step": 30250
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.171554632753699e-06,
+      "loss": 0.3818,
+      "step": 30251
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.170977881592987e-06,
+      "loss": 0.4308,
+      "step": 30252
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.170401130432275e-06,
+      "loss": 0.4128,
+      "step": 30253
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.169824379271563e-06,
+      "loss": 0.4033,
+      "step": 30254
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.169247628110852e-06,
+      "loss": 0.3305,
+      "step": 30255
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1686708769501402e-06,
+      "loss": 0.4077,
+      "step": 30256
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1680941257894283e-06,
+      "loss": 0.3159,
+      "step": 30257
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.167517374628717e-06,
+      "loss": 0.3452,
+      "step": 30258
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.166940623468005e-06,
+      "loss": 0.2995,
+      "step": 30259
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.166363872307293e-06,
+      "loss": 0.4719,
+      "step": 30260
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.165787121146582e-06,
+      "loss": 0.3832,
+      "step": 30261
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.16521036998587e-06,
+      "loss": 0.4344,
+      "step": 30262
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1646336188251582e-06,
+      "loss": 0.5045,
+      "step": 30263
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1640568676644464e-06,
+      "loss": 0.3195,
+      "step": 30264
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.163480116503735e-06,
+      "loss": 0.3762,
+      "step": 30265
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.162903365343023e-06,
+      "loss": 0.338,
+      "step": 30266
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.162326614182311e-06,
+      "loss": 0.4436,
+      "step": 30267
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1617498630215996e-06,
+      "loss": 0.4322,
+      "step": 30268
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1611731118608877e-06,
+      "loss": 0.3985,
+      "step": 30269
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.160596360700176e-06,
+      "loss": 0.4462,
+      "step": 30270
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1600196095394644e-06,
+      "loss": 0.4767,
+      "step": 30271
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.159442858378753e-06,
+      "loss": 0.4776,
+      "step": 30272
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.158866107218041e-06,
+      "loss": 0.4098,
+      "step": 30273
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.158289356057329e-06,
+      "loss": 0.5903,
+      "step": 30274
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1577126048966177e-06,
+      "loss": 0.4751,
+      "step": 30275
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1571358537359058e-06,
+      "loss": 0.4341,
+      "step": 30276
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.156559102575194e-06,
+      "loss": 0.3992,
+      "step": 30277
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.155982351414483e-06,
+      "loss": 0.4208,
+      "step": 30278
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.155405600253771e-06,
+      "loss": 0.2367,
+      "step": 30279
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.154828849093059e-06,
+      "loss": 0.4331,
+      "step": 30280
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.154252097932347e-06,
+      "loss": 0.3581,
+      "step": 30281
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1536753467716357e-06,
+      "loss": 0.4822,
+      "step": 30282
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.153098595610924e-06,
+      "loss": 0.4285,
+      "step": 30283
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.152521844450212e-06,
+      "loss": 0.33,
+      "step": 30284
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.151945093289501e-06,
+      "loss": 0.3603,
+      "step": 30285
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.151368342128789e-06,
+      "loss": 0.4739,
+      "step": 30286
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.150791590968077e-06,
+      "loss": 0.3673,
+      "step": 30287
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1502148398073656e-06,
+      "loss": 0.4115,
+      "step": 30288
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1496380886466537e-06,
+      "loss": 0.5363,
+      "step": 30289
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.149061337485942e-06,
+      "loss": 0.2618,
+      "step": 30290
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.14848458632523e-06,
+      "loss": 0.4448,
+      "step": 30291
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.147907835164519e-06,
+      "loss": 0.4152,
+      "step": 30292
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.147331084003807e-06,
+      "loss": 0.4422,
+      "step": 30293
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.146754332843095e-06,
+      "loss": 0.3774,
+      "step": 30294
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1461775816823837e-06,
+      "loss": 0.4477,
+      "step": 30295
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.1456008305216718e-06,
+      "loss": 0.3422,
+      "step": 30296
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.14502407936096e-06,
+      "loss": 0.4041,
+      "step": 30297
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 3.144447328200248e-06,
+      "loss": 0.3677,
+      "step": 30298
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1438705770395365e-06,
+      "loss": 0.4421,
+      "step": 30299
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1432938258788246e-06,
+      "loss": 0.38,
+      "step": 30300
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1427170747181127e-06,
+      "loss": 0.416,
+      "step": 30301
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1421403235574017e-06,
+      "loss": 0.4272,
+      "step": 30302
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.14156357239669e-06,
+      "loss": 0.3809,
+      "step": 30303
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.140986821235978e-06,
+      "loss": 0.3822,
+      "step": 30304
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1404100700752664e-06,
+      "loss": 0.303,
+      "step": 30305
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1398333189145545e-06,
+      "loss": 0.5146,
+      "step": 30306
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1392565677538426e-06,
+      "loss": 0.3203,
+      "step": 30307
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1386798165931308e-06,
+      "loss": 0.4178,
+      "step": 30308
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1381030654324197e-06,
+      "loss": 0.5867,
+      "step": 30309
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.137526314271708e-06,
+      "loss": 0.4223,
+      "step": 30310
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.136949563110996e-06,
+      "loss": 0.4275,
+      "step": 30311
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1363728119502845e-06,
+      "loss": 0.3935,
+      "step": 30312
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1357960607895726e-06,
+      "loss": 0.4225,
+      "step": 30313
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1352193096288607e-06,
+      "loss": 0.5133,
+      "step": 30314
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.134642558468149e-06,
+      "loss": 0.3414,
+      "step": 30315
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1340658073074377e-06,
+      "loss": 0.3831,
+      "step": 30316
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.133489056146726e-06,
+      "loss": 0.4673,
+      "step": 30317
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.132912304986014e-06,
+      "loss": 0.4412,
+      "step": 30318
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1323355538253025e-06,
+      "loss": 0.4196,
+      "step": 30319
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1317588026645906e-06,
+      "loss": 0.4701,
+      "step": 30320
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1311820515038787e-06,
+      "loss": 0.3241,
+      "step": 30321
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1306053003431672e-06,
+      "loss": 0.3951,
+      "step": 30322
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1300285491824558e-06,
+      "loss": 0.4077,
+      "step": 30323
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.129451798021744e-06,
+      "loss": 0.4679,
+      "step": 30324
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.128875046861032e-06,
+      "loss": 0.4378,
+      "step": 30325
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1282982957003205e-06,
+      "loss": 0.4321,
+      "step": 30326
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1277215445396086e-06,
+      "loss": 0.3309,
+      "step": 30327
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1271447933788967e-06,
+      "loss": 0.3885,
+      "step": 30328
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1265680422181853e-06,
+      "loss": 0.581,
+      "step": 30329
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1259912910574734e-06,
+      "loss": 0.4629,
+      "step": 30330
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1254145398967615e-06,
+      "loss": 0.5069,
+      "step": 30331
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.12483778873605e-06,
+      "loss": 0.4353,
+      "step": 30332
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1242610375753386e-06,
+      "loss": 0.3606,
+      "step": 30333
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1236842864146267e-06,
+      "loss": 0.4636,
+      "step": 30334
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1231075352539148e-06,
+      "loss": 0.3667,
+      "step": 30335
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1225307840932033e-06,
+      "loss": 0.434,
+      "step": 30336
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1219540329324914e-06,
+      "loss": 0.477,
+      "step": 30337
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1213772817717795e-06,
+      "loss": 0.4003,
+      "step": 30338
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1208005306110685e-06,
+      "loss": 0.3663,
+      "step": 30339
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1202237794503566e-06,
+      "loss": 0.3384,
+      "step": 30340
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1196470282896447e-06,
+      "loss": 0.3019,
+      "step": 30341
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.119070277128933e-06,
+      "loss": 0.4628,
+      "step": 30342
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1184935259682213e-06,
+      "loss": 0.3707,
+      "step": 30343
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1179167748075094e-06,
+      "loss": 0.3522,
+      "step": 30344
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1173400236467976e-06,
+      "loss": 0.6247,
+      "step": 30345
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1167632724860865e-06,
+      "loss": 0.4053,
+      "step": 30346
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1161865213253746e-06,
+      "loss": 0.2914,
+      "step": 30347
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1156097701646627e-06,
+      "loss": 0.4281,
+      "step": 30348
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1150330190039513e-06,
+      "loss": 0.4073,
+      "step": 30349
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1144562678432394e-06,
+      "loss": 0.3361,
+      "step": 30350
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1138795166825275e-06,
+      "loss": 0.4844,
+      "step": 30351
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1133027655218156e-06,
+      "loss": 0.4237,
+      "step": 30352
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1127260143611045e-06,
+      "loss": 0.4566,
+      "step": 30353
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1121492632003926e-06,
+      "loss": 0.3926,
+      "step": 30354
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1115725120396808e-06,
+      "loss": 0.3619,
+      "step": 30355
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1109957608789693e-06,
+      "loss": 0.4271,
+      "step": 30356
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1104190097182574e-06,
+      "loss": 0.3918,
+      "step": 30357
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1098422585575455e-06,
+      "loss": 0.5137,
+      "step": 30358
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1092655073968336e-06,
+      "loss": 0.3087,
+      "step": 30359
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.108688756236122e-06,
+      "loss": 0.3612,
+      "step": 30360
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1081120050754103e-06,
+      "loss": 0.3369,
+      "step": 30361
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1075352539146984e-06,
+      "loss": 0.3992,
+      "step": 30362
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1069585027539873e-06,
+      "loss": 0.3983,
+      "step": 30363
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1063817515932754e-06,
+      "loss": 0.3602,
+      "step": 30364
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1058050004325635e-06,
+      "loss": 0.3356,
+      "step": 30365
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.105228249271852e-06,
+      "loss": 0.4633,
+      "step": 30366
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.10465149811114e-06,
+      "loss": 0.4393,
+      "step": 30367
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1040747469504283e-06,
+      "loss": 0.2567,
+      "step": 30368
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1034979957897164e-06,
+      "loss": 0.3556,
+      "step": 30369
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1029212446290054e-06,
+      "loss": 0.3163,
+      "step": 30370
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1023444934682935e-06,
+      "loss": 0.2591,
+      "step": 30371
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1017677423075816e-06,
+      "loss": 0.3825,
+      "step": 30372
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.10119099114687e-06,
+      "loss": 0.4438,
+      "step": 30373
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.100614239986158e-06,
+      "loss": 0.3884,
+      "step": 30374
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.1000374888254463e-06,
+      "loss": 0.3873,
+      "step": 30375
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0994607376647344e-06,
+      "loss": 0.2998,
+      "step": 30376
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0988839865040234e-06,
+      "loss": 0.3459,
+      "step": 30377
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0983072353433115e-06,
+      "loss": 0.2471,
+      "step": 30378
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0977304841825996e-06,
+      "loss": 0.5226,
+      "step": 30379
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.097153733021888e-06,
+      "loss": 0.3833,
+      "step": 30380
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0965769818611762e-06,
+      "loss": 0.3556,
+      "step": 30381
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0960002307004643e-06,
+      "loss": 0.3401,
+      "step": 30382
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.095423479539753e-06,
+      "loss": 0.4762,
+      "step": 30383
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0948467283790414e-06,
+      "loss": 0.3435,
+      "step": 30384
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0942699772183295e-06,
+      "loss": 0.4224,
+      "step": 30385
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0936932260576176e-06,
+      "loss": 0.3713,
+      "step": 30386
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.093116474896906e-06,
+      "loss": 0.3393,
+      "step": 30387
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0925397237361943e-06,
+      "loss": 0.2826,
+      "step": 30388
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0919629725754824e-06,
+      "loss": 0.3026,
+      "step": 30389
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.091386221414771e-06,
+      "loss": 0.4117,
+      "step": 30390
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.090809470254059e-06,
+      "loss": 0.3933,
+      "step": 30391
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.090232719093347e-06,
+      "loss": 0.2799,
+      "step": 30392
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0896559679326352e-06,
+      "loss": 0.3665,
+      "step": 30393
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.089079216771924e-06,
+      "loss": 0.5313,
+      "step": 30394
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0885024656112123e-06,
+      "loss": 0.4087,
+      "step": 30395
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0879257144505004e-06,
+      "loss": 0.3987,
+      "step": 30396
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.087348963289789e-06,
+      "loss": 0.3188,
+      "step": 30397
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.086772212129077e-06,
+      "loss": 0.4586,
+      "step": 30398
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.086195460968365e-06,
+      "loss": 0.3073,
+      "step": 30399
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.085618709807654e-06,
+      "loss": 0.4272,
+      "step": 30400
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0850419586469422e-06,
+      "loss": 0.5454,
+      "step": 30401
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0844652074862303e-06,
+      "loss": 0.5324,
+      "step": 30402
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0838884563255184e-06,
+      "loss": 0.3311,
+      "step": 30403
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.083311705164807e-06,
+      "loss": 0.1456,
+      "step": 30404
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.082734954004095e-06,
+      "loss": 0.5154,
+      "step": 30405
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.082158202843383e-06,
+      "loss": 0.4334,
+      "step": 30406
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.081581451682672e-06,
+      "loss": 0.5237,
+      "step": 30407
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0810047005219603e-06,
+      "loss": 0.3402,
+      "step": 30408
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0804279493612484e-06,
+      "loss": 0.4268,
+      "step": 30409
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.079851198200537e-06,
+      "loss": 0.5177,
+      "step": 30410
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.079274447039825e-06,
+      "loss": 0.5015,
+      "step": 30411
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.078697695879113e-06,
+      "loss": 0.3969,
+      "step": 30412
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0781209447184012e-06,
+      "loss": 0.3998,
+      "step": 30413
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0775441935576898e-06,
+      "loss": 0.4598,
+      "step": 30414
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0769674423969783e-06,
+      "loss": 0.4098,
+      "step": 30415
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0763906912362664e-06,
+      "loss": 0.324,
+      "step": 30416
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.075813940075555e-06,
+      "loss": 0.3982,
+      "step": 30417
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.075237188914843e-06,
+      "loss": 0.438,
+      "step": 30418
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.074660437754131e-06,
+      "loss": 0.3823,
+      "step": 30419
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0740836865934193e-06,
+      "loss": 0.4082,
+      "step": 30420
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0735069354327078e-06,
+      "loss": 0.5009,
+      "step": 30421
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.072930184271996e-06,
+      "loss": 0.419,
+      "step": 30422
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.072353433111284e-06,
+      "loss": 0.3808,
+      "step": 30423
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.071776681950573e-06,
+      "loss": 0.4726,
+      "step": 30424
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.071199930789861e-06,
+      "loss": 0.3964,
+      "step": 30425
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.070623179629149e-06,
+      "loss": 0.3736,
+      "step": 30426
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0700464284684377e-06,
+      "loss": 0.3436,
+      "step": 30427
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.069469677307726e-06,
+      "loss": 0.4421,
+      "step": 30428
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.068892926147014e-06,
+      "loss": 0.5049,
+      "step": 30429
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.068316174986302e-06,
+      "loss": 0.4306,
+      "step": 30430
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.067739423825591e-06,
+      "loss": 0.4887,
+      "step": 30431
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.067162672664879e-06,
+      "loss": 0.4084,
+      "step": 30432
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.066585921504167e-06,
+      "loss": 0.3402,
+      "step": 30433
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0660091703434557e-06,
+      "loss": 0.4973,
+      "step": 30434
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.065432419182744e-06,
+      "loss": 0.4821,
+      "step": 30435
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.064855668022032e-06,
+      "loss": 0.3301,
+      "step": 30436
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.06427891686132e-06,
+      "loss": 0.3491,
+      "step": 30437
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.063702165700609e-06,
+      "loss": 0.3448,
+      "step": 30438
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.063125414539897e-06,
+      "loss": 0.3431,
+      "step": 30439
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0625486633791852e-06,
+      "loss": 0.4265,
+      "step": 30440
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0619719122184738e-06,
+      "loss": 0.4791,
+      "step": 30441
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.061395161057762e-06,
+      "loss": 0.3838,
+      "step": 30442
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.06081840989705e-06,
+      "loss": 0.4658,
+      "step": 30443
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0602416587363385e-06,
+      "loss": 0.2876,
+      "step": 30444
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.059664907575627e-06,
+      "loss": 0.4107,
+      "step": 30445
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.059088156414915e-06,
+      "loss": 0.3635,
+      "step": 30446
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0585114052542033e-06,
+      "loss": 0.2921,
+      "step": 30447
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.057934654093492e-06,
+      "loss": 0.3292,
+      "step": 30448
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.05735790293278e-06,
+      "loss": 0.4853,
+      "step": 30449
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.056781151772068e-06,
+      "loss": 0.4687,
+      "step": 30450
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0562044006113565e-06,
+      "loss": 0.393,
+      "step": 30451
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0556276494506447e-06,
+      "loss": 0.4181,
+      "step": 30452
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0550508982899328e-06,
+      "loss": 0.377,
+      "step": 30453
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.054474147129221e-06,
+      "loss": 0.4323,
+      "step": 30454
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.05389739596851e-06,
+      "loss": 0.5771,
+      "step": 30455
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.053320644807798e-06,
+      "loss": 0.3848,
+      "step": 30456
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.052743893647086e-06,
+      "loss": 0.3539,
+      "step": 30457
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0521671424863746e-06,
+      "loss": 0.4331,
+      "step": 30458
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0515903913256627e-06,
+      "loss": 0.3953,
+      "step": 30459
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.051013640164951e-06,
+      "loss": 0.2992,
+      "step": 30460
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0504368890042398e-06,
+      "loss": 0.3738,
+      "step": 30461
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.049860137843528e-06,
+      "loss": 0.403,
+      "step": 30462
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.049283386682816e-06,
+      "loss": 0.4044,
+      "step": 30463
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.048706635522104e-06,
+      "loss": 0.3857,
+      "step": 30464
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0481298843613926e-06,
+      "loss": 0.3975,
+      "step": 30465
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0475531332006807e-06,
+      "loss": 0.5217,
+      "step": 30466
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.046976382039969e-06,
+      "loss": 0.442,
+      "step": 30467
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0463996308792578e-06,
+      "loss": 0.3863,
+      "step": 30468
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.045822879718546e-06,
+      "loss": 0.4371,
+      "step": 30469
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.045246128557834e-06,
+      "loss": 0.3111,
+      "step": 30470
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.044669377397122e-06,
+      "loss": 0.4238,
+      "step": 30471
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0440926262364106e-06,
+      "loss": 0.4256,
+      "step": 30472
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0435158750756987e-06,
+      "loss": 0.3772,
+      "step": 30473
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.042939123914987e-06,
+      "loss": 0.4006,
+      "step": 30474
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.0423623727542754e-06,
+      "loss": 0.4038,
+      "step": 30475
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.041785621593564e-06,
+      "loss": 0.4747,
+      "step": 30476
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.041208870432852e-06,
+      "loss": 0.3033,
+      "step": 30477
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0406321192721406e-06,
+      "loss": 0.3357,
+      "step": 30478
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0400553681114287e-06,
+      "loss": 0.3944,
+      "step": 30479
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0394786169507168e-06,
+      "loss": 0.5021,
+      "step": 30480
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.038901865790005e-06,
+      "loss": 0.3988,
+      "step": 30481
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0383251146292934e-06,
+      "loss": 0.4292,
+      "step": 30482
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0377483634685815e-06,
+      "loss": 0.4757,
+      "step": 30483
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0371716123078696e-06,
+      "loss": 0.4239,
+      "step": 30484
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0365948611471586e-06,
+      "loss": 0.3681,
+      "step": 30485
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0360181099864467e-06,
+      "loss": 0.3589,
+      "step": 30486
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.035441358825735e-06,
+      "loss": 0.4964,
+      "step": 30487
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0348646076650233e-06,
+      "loss": 0.3469,
+      "step": 30488
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0342878565043115e-06,
+      "loss": 0.2838,
+      "step": 30489
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0337111053435996e-06,
+      "loss": 0.4257,
+      "step": 30490
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0331343541828877e-06,
+      "loss": 0.4963,
+      "step": 30491
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0325576030221766e-06,
+      "loss": 0.4521,
+      "step": 30492
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0319808518614647e-06,
+      "loss": 0.3141,
+      "step": 30493
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.031404100700753e-06,
+      "loss": 0.3303,
+      "step": 30494
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0308273495400414e-06,
+      "loss": 0.3856,
+      "step": 30495
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0302505983793295e-06,
+      "loss": 0.3877,
+      "step": 30496
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0296738472186176e-06,
+      "loss": 0.3918,
+      "step": 30497
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0290970960579057e-06,
+      "loss": 0.4396,
+      "step": 30498
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0285203448971947e-06,
+      "loss": 0.4676,
+      "step": 30499
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0279435937364828e-06,
+      "loss": 0.4494,
+      "step": 30500
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.027366842575771e-06,
+      "loss": 0.4207,
+      "step": 30501
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0267900914150594e-06,
+      "loss": 0.4588,
+      "step": 30502
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0262133402543475e-06,
+      "loss": 0.3631,
+      "step": 30503
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0256365890936356e-06,
+      "loss": 0.3587,
+      "step": 30504
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.025059837932924e-06,
+      "loss": 0.3679,
+      "step": 30505
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0244830867722123e-06,
+      "loss": 0.3475,
+      "step": 30506
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.023906335611501e-06,
+      "loss": 0.4217,
+      "step": 30507
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.023329584450789e-06,
+      "loss": 0.3997,
+      "step": 30508
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0227528332900774e-06,
+      "loss": 0.3075,
+      "step": 30509
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0221760821293655e-06,
+      "loss": 0.5242,
+      "step": 30510
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0215993309686537e-06,
+      "loss": 0.4907,
+      "step": 30511
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.021022579807942e-06,
+      "loss": 0.4324,
+      "step": 30512
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0204458286472303e-06,
+      "loss": 0.3785,
+      "step": 30513
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0198690774865184e-06,
+      "loss": 0.3862,
+      "step": 30514
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0192923263258065e-06,
+      "loss": 0.3659,
+      "step": 30515
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0187155751650955e-06,
+      "loss": 0.4472,
+      "step": 30516
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0181388240043836e-06,
+      "loss": 0.3995,
+      "step": 30517
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0175620728436717e-06,
+      "loss": 0.397,
+      "step": 30518
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0169853216829602e-06,
+      "loss": 0.3885,
+      "step": 30519
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0164085705222483e-06,
+      "loss": 0.4052,
+      "step": 30520
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0158318193615364e-06,
+      "loss": 0.4004,
+      "step": 30521
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0152550682008254e-06,
+      "loss": 0.4361,
+      "step": 30522
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0146783170401135e-06,
+      "loss": 0.3326,
+      "step": 30523
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0141015658794016e-06,
+      "loss": 0.4537,
+      "step": 30524
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0135248147186897e-06,
+      "loss": 0.3702,
+      "step": 30525
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0129480635579782e-06,
+      "loss": 0.4116,
+      "step": 30526
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0123713123972664e-06,
+      "loss": 0.3455,
+      "step": 30527
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0117945612365545e-06,
+      "loss": 0.4359,
+      "step": 30528
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0112178100758434e-06,
+      "loss": 0.3604,
+      "step": 30529
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0106410589151315e-06,
+      "loss": 0.4082,
+      "step": 30530
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0100643077544196e-06,
+      "loss": 0.3644,
+      "step": 30531
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0094875565937077e-06,
+      "loss": 0.4088,
+      "step": 30532
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0089108054329963e-06,
+      "loss": 0.3094,
+      "step": 30533
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0083340542722844e-06,
+      "loss": 0.4071,
+      "step": 30534
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0077573031115725e-06,
+      "loss": 0.3658,
+      "step": 30535
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.007180551950861e-06,
+      "loss": 0.4263,
+      "step": 30536
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0066038007901496e-06,
+      "loss": 0.4617,
+      "step": 30537
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0060270496294377e-06,
+      "loss": 0.3903,
+      "step": 30538
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.005450298468726e-06,
+      "loss": 0.432,
+      "step": 30539
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0048735473080143e-06,
+      "loss": 0.3507,
+      "step": 30540
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0042967961473024e-06,
+      "loss": 0.5197,
+      "step": 30541
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0037200449865905e-06,
+      "loss": 0.4364,
+      "step": 30542
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.003143293825879e-06,
+      "loss": 0.3219,
+      "step": 30543
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.002566542665167e-06,
+      "loss": 0.3751,
+      "step": 30544
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0019897915044553e-06,
+      "loss": 0.3446,
+      "step": 30545
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0014130403437442e-06,
+      "loss": 0.5327,
+      "step": 30546
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0008362891830323e-06,
+      "loss": 0.4873,
+      "step": 30547
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.0002595380223204e-06,
+      "loss": 0.447,
+      "step": 30548
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.999682786861609e-06,
+      "loss": 0.4924,
+      "step": 30549
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.999106035700897e-06,
+      "loss": 0.3073,
+      "step": 30550
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.998529284540185e-06,
+      "loss": 0.4452,
+      "step": 30551
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9979525333794733e-06,
+      "loss": 0.3211,
+      "step": 30552
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9973757822187623e-06,
+      "loss": 0.4468,
+      "step": 30553
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9967990310580504e-06,
+      "loss": 0.3901,
+      "step": 30554
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9962222798973385e-06,
+      "loss": 0.3441,
+      "step": 30555
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.995645528736627e-06,
+      "loss": 0.3146,
+      "step": 30556
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.995068777575915e-06,
+      "loss": 0.3838,
+      "step": 30557
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9944920264152032e-06,
+      "loss": 0.4299,
+      "step": 30558
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9939152752544913e-06,
+      "loss": 0.3996,
+      "step": 30559
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9933385240937803e-06,
+      "loss": 0.3498,
+      "step": 30560
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9927617729330684e-06,
+      "loss": 0.4707,
+      "step": 30561
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9921850217723565e-06,
+      "loss": 0.392,
+      "step": 30562
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.991608270611645e-06,
+      "loss": 0.4246,
+      "step": 30563
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.991031519450933e-06,
+      "loss": 0.4778,
+      "step": 30564
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9904547682902213e-06,
+      "loss": 0.376,
+      "step": 30565
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.98987801712951e-06,
+      "loss": 0.3603,
+      "step": 30566
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.989301265968798e-06,
+      "loss": 0.3479,
+      "step": 30567
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9887245148080864e-06,
+      "loss": 0.4417,
+      "step": 30568
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9881477636473745e-06,
+      "loss": 0.4745,
+      "step": 30569
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.987571012486663e-06,
+      "loss": 0.4253,
+      "step": 30570
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.986994261325951e-06,
+      "loss": 0.4662,
+      "step": 30571
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9864175101652393e-06,
+      "loss": 0.3817,
+      "step": 30572
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.985840759004528e-06,
+      "loss": 0.3713,
+      "step": 30573
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.985264007843816e-06,
+      "loss": 0.3335,
+      "step": 30574
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.984687256683104e-06,
+      "loss": 0.3348,
+      "step": 30575
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.984110505522392e-06,
+      "loss": 0.3547,
+      "step": 30576
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.983533754361681e-06,
+      "loss": 0.3083,
+      "step": 30577
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9829570032009692e-06,
+      "loss": 0.4344,
+      "step": 30578
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9823802520402573e-06,
+      "loss": 0.5796,
+      "step": 30579
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.981803500879546e-06,
+      "loss": 0.3547,
+      "step": 30580
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.981226749718834e-06,
+      "loss": 0.3677,
+      "step": 30581
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.980649998558122e-06,
+      "loss": 0.3503,
+      "step": 30582
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.980073247397411e-06,
+      "loss": 0.4416,
+      "step": 30583
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.979496496236699e-06,
+      "loss": 0.3667,
+      "step": 30584
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9789197450759872e-06,
+      "loss": 0.3277,
+      "step": 30585
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9783429939152754e-06,
+      "loss": 0.3878,
+      "step": 30586
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.977766242754564e-06,
+      "loss": 0.4414,
+      "step": 30587
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.977189491593852e-06,
+      "loss": 0.3846,
+      "step": 30588
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.97661274043314e-06,
+      "loss": 0.3765,
+      "step": 30589
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.976035989272429e-06,
+      "loss": 0.4552,
+      "step": 30590
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.975459238111717e-06,
+      "loss": 0.3863,
+      "step": 30591
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9748824869510053e-06,
+      "loss": 0.4148,
+      "step": 30592
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9743057357902934e-06,
+      "loss": 0.3835,
+      "step": 30593
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.973728984629582e-06,
+      "loss": 0.3749,
+      "step": 30594
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.97315223346887e-06,
+      "loss": 0.3484,
+      "step": 30595
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.972575482308158e-06,
+      "loss": 0.4239,
+      "step": 30596
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9719987311474467e-06,
+      "loss": 0.4819,
+      "step": 30597
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9714219799867348e-06,
+      "loss": 0.3938,
+      "step": 30598
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9708452288260233e-06,
+      "loss": 0.4456,
+      "step": 30599
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.970268477665312e-06,
+      "loss": 0.3524,
+      "step": 30600
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9696917265046e-06,
+      "loss": 0.333,
+      "step": 30601
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.969114975343888e-06,
+      "loss": 0.5218,
+      "step": 30602
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.968538224183176e-06,
+      "loss": 0.4917,
+      "step": 30603
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9679614730224647e-06,
+      "loss": 0.3557,
+      "step": 30604
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.967384721861753e-06,
+      "loss": 0.4468,
+      "step": 30605
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.966807970701041e-06,
+      "loss": 0.3296,
+      "step": 30606
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.96623121954033e-06,
+      "loss": 0.4568,
+      "step": 30607
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.965654468379618e-06,
+      "loss": 0.4432,
+      "step": 30608
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.965077717218906e-06,
+      "loss": 0.4271,
+      "step": 30609
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9645009660581946e-06,
+      "loss": 0.4418,
+      "step": 30610
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9639242148974827e-06,
+      "loss": 0.3588,
+      "step": 30611
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.963347463736771e-06,
+      "loss": 0.3637,
+      "step": 30612
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.962770712576059e-06,
+      "loss": 0.29,
+      "step": 30613
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.962193961415348e-06,
+      "loss": 0.4326,
+      "step": 30614
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.961617210254636e-06,
+      "loss": 0.4029,
+      "step": 30615
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.961040459093924e-06,
+      "loss": 0.4434,
+      "step": 30616
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9604637079332126e-06,
+      "loss": 0.3553,
+      "step": 30617
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9598869567725008e-06,
+      "loss": 0.3281,
+      "step": 30618
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.959310205611789e-06,
+      "loss": 0.3425,
+      "step": 30619
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.958733454451077e-06,
+      "loss": 0.3317,
+      "step": 30620
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.958156703290366e-06,
+      "loss": 0.3871,
+      "step": 30621
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.957579952129654e-06,
+      "loss": 0.3744,
+      "step": 30622
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.957003200968942e-06,
+      "loss": 0.3298,
+      "step": 30623
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9564264498082307e-06,
+      "loss": 0.4159,
+      "step": 30624
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9558496986475188e-06,
+      "loss": 0.5119,
+      "step": 30625
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.955272947486807e-06,
+      "loss": 0.3686,
+      "step": 30626
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9546961963260954e-06,
+      "loss": 0.3284,
+      "step": 30627
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9541194451653835e-06,
+      "loss": 0.362,
+      "step": 30628
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.953542694004672e-06,
+      "loss": 0.3966,
+      "step": 30629
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.95296594284396e-06,
+      "loss": 0.5707,
+      "step": 30630
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9523891916832487e-06,
+      "loss": 0.3923,
+      "step": 30631
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.951812440522537e-06,
+      "loss": 0.3748,
+      "step": 30632
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.951235689361825e-06,
+      "loss": 0.4178,
+      "step": 30633
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9506589382011135e-06,
+      "loss": 0.4156,
+      "step": 30634
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9500821870404016e-06,
+      "loss": 0.3211,
+      "step": 30635
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9495054358796897e-06,
+      "loss": 0.4949,
+      "step": 30636
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9489286847189778e-06,
+      "loss": 0.4694,
+      "step": 30637
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9483519335582667e-06,
+      "loss": 0.3464,
+      "step": 30638
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.947775182397555e-06,
+      "loss": 0.2808,
+      "step": 30639
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.947198431236843e-06,
+      "loss": 0.4501,
+      "step": 30640
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9466216800761315e-06,
+      "loss": 0.4227,
+      "step": 30641
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9460449289154196e-06,
+      "loss": 0.3822,
+      "step": 30642
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9454681777547077e-06,
+      "loss": 0.3961,
+      "step": 30643
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9448914265939967e-06,
+      "loss": 0.3994,
+      "step": 30644
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9443146754332848e-06,
+      "loss": 0.3726,
+      "step": 30645
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.943737924272573e-06,
+      "loss": 0.3981,
+      "step": 30646
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.943161173111861e-06,
+      "loss": 0.4506,
+      "step": 30647
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9425844219511495e-06,
+      "loss": 0.3956,
+      "step": 30648
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9420076707904376e-06,
+      "loss": 0.3919,
+      "step": 30649
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9414309196297257e-06,
+      "loss": 0.4163,
+      "step": 30650
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9408541684690147e-06,
+      "loss": 0.3329,
+      "step": 30651
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.940277417308303e-06,
+      "loss": 0.4419,
+      "step": 30652
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.939700666147591e-06,
+      "loss": 0.4073,
+      "step": 30653
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.939123914986879e-06,
+      "loss": 0.4623,
+      "step": 30654
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2.9385471638261676e-06,
+      "loss": 0.5645,
+      "step": 30655
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9379704126654557e-06,
+      "loss": 0.5268,
+      "step": 30656
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9373936615047438e-06,
+      "loss": 0.3155,
+      "step": 30657
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9368169103440323e-06,
+      "loss": 0.3423,
+      "step": 30658
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9362401591833204e-06,
+      "loss": 0.4185,
+      "step": 30659
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.935663408022609e-06,
+      "loss": 0.4848,
+      "step": 30660
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9350866568618975e-06,
+      "loss": 0.3679,
+      "step": 30661
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9345099057011856e-06,
+      "loss": 0.3585,
+      "step": 30662
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9339331545404737e-06,
+      "loss": 0.4842,
+      "step": 30663
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.933356403379762e-06,
+      "loss": 0.4136,
+      "step": 30664
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9327796522190503e-06,
+      "loss": 0.378,
+      "step": 30665
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9322029010583384e-06,
+      "loss": 0.4265,
+      "step": 30666
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9316261498976265e-06,
+      "loss": 0.3875,
+      "step": 30667
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9310493987369155e-06,
+      "loss": 0.3359,
+      "step": 30668
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9304726475762036e-06,
+      "loss": 0.4817,
+      "step": 30669
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9298958964154917e-06,
+      "loss": 0.3071,
+      "step": 30670
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9293191452547803e-06,
+      "loss": 0.3607,
+      "step": 30671
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9287423940940684e-06,
+      "loss": 0.2693,
+      "step": 30672
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9281656429333565e-06,
+      "loss": 0.4087,
+      "step": 30673
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9275888917726446e-06,
+      "loss": 0.4373,
+      "step": 30674
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9270121406119335e-06,
+      "loss": 0.3986,
+      "step": 30675
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9264353894512216e-06,
+      "loss": 0.4417,
+      "step": 30676
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9258586382905098e-06,
+      "loss": 0.3349,
+      "step": 30677
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9252818871297983e-06,
+      "loss": 0.3725,
+      "step": 30678
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9247051359690864e-06,
+      "loss": 0.4346,
+      "step": 30679
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9241283848083745e-06,
+      "loss": 0.4164,
+      "step": 30680
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9235516336476626e-06,
+      "loss": 0.4033,
+      "step": 30681
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9229748824869516e-06,
+      "loss": 0.3437,
+      "step": 30682
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9223981313262397e-06,
+      "loss": 0.4291,
+      "step": 30683
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9218213801655278e-06,
+      "loss": 0.3411,
+      "step": 30684
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9212446290048163e-06,
+      "loss": 0.4462,
+      "step": 30685
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9206678778441044e-06,
+      "loss": 0.3649,
+      "step": 30686
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9200911266833925e-06,
+      "loss": 0.3785,
+      "step": 30687
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.919514375522681e-06,
+      "loss": 0.4225,
+      "step": 30688
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.918937624361969e-06,
+      "loss": 0.3539,
+      "step": 30689
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9183608732012573e-06,
+      "loss": 0.3557,
+      "step": 30690
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.917784122040546e-06,
+      "loss": 0.4226,
+      "step": 30691
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9172073708798343e-06,
+      "loss": 0.4464,
+      "step": 30692
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9166306197191225e-06,
+      "loss": 0.2191,
+      "step": 30693
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9160538685584106e-06,
+      "loss": 0.2702,
+      "step": 30694
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.915477117397699e-06,
+      "loss": 0.4438,
+      "step": 30695
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.914900366236987e-06,
+      "loss": 0.4732,
+      "step": 30696
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9143236150762753e-06,
+      "loss": 0.3363,
+      "step": 30697
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9137468639155634e-06,
+      "loss": 0.5459,
+      "step": 30698
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9131701127548524e-06,
+      "loss": 0.3194,
+      "step": 30699
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9125933615941405e-06,
+      "loss": 0.2933,
+      "step": 30700
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9120166104334286e-06,
+      "loss": 0.4105,
+      "step": 30701
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.911439859272717e-06,
+      "loss": 0.36,
+      "step": 30702
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9108631081120052e-06,
+      "loss": 0.4213,
+      "step": 30703
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9102863569512933e-06,
+      "loss": 0.5143,
+      "step": 30704
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9097096057905823e-06,
+      "loss": 0.4703,
+      "step": 30705
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9091328546298704e-06,
+      "loss": 0.3761,
+      "step": 30706
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9085561034691585e-06,
+      "loss": 0.4372,
+      "step": 30707
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9079793523084466e-06,
+      "loss": 0.3943,
+      "step": 30708
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.907402601147735e-06,
+      "loss": 0.5149,
+      "step": 30709
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9068258499870233e-06,
+      "loss": 0.3524,
+      "step": 30710
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9062490988263114e-06,
+      "loss": 0.3971,
+      "step": 30711
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9056723476656003e-06,
+      "loss": 0.4467,
+      "step": 30712
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9050955965048884e-06,
+      "loss": 0.3455,
+      "step": 30713
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9045188453441765e-06,
+      "loss": 0.2693,
+      "step": 30714
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9039420941834647e-06,
+      "loss": 0.556,
+      "step": 30715
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.903365343022753e-06,
+      "loss": 0.3703,
+      "step": 30716
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9027885918620413e-06,
+      "loss": 0.3672,
+      "step": 30717
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9022118407013294e-06,
+      "loss": 0.3618,
+      "step": 30718
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.901635089540618e-06,
+      "loss": 0.3944,
+      "step": 30719
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.901058338379906e-06,
+      "loss": 0.4549,
+      "step": 30720
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.9004815872191946e-06,
+      "loss": 0.3917,
+      "step": 30721
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.899904836058483e-06,
+      "loss": 0.4355,
+      "step": 30722
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8993280848977712e-06,
+      "loss": 0.3861,
+      "step": 30723
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8987513337370593e-06,
+      "loss": 0.4163,
+      "step": 30724
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8981745825763474e-06,
+      "loss": 0.3561,
+      "step": 30725
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.897597831415636e-06,
+      "loss": 0.3825,
+      "step": 30726
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.897021080254924e-06,
+      "loss": 0.3014,
+      "step": 30727
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.896444329094212e-06,
+      "loss": 0.3517,
+      "step": 30728
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.895867577933501e-06,
+      "loss": 0.3607,
+      "step": 30729
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8952908267727893e-06,
+      "loss": 0.4373,
+      "step": 30730
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8947140756120774e-06,
+      "loss": 0.3918,
+      "step": 30731
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.894137324451366e-06,
+      "loss": 0.5695,
+      "step": 30732
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.893560573290654e-06,
+      "loss": 0.3945,
+      "step": 30733
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.892983822129942e-06,
+      "loss": 0.4351,
+      "step": 30734
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8924070709692302e-06,
+      "loss": 0.3848,
+      "step": 30735
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.891830319808519e-06,
+      "loss": 0.363,
+      "step": 30736
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8912535686478073e-06,
+      "loss": 0.4826,
+      "step": 30737
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8906768174870954e-06,
+      "loss": 0.4478,
+      "step": 30738
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.890100066326384e-06,
+      "loss": 0.4585,
+      "step": 30739
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.889523315165672e-06,
+      "loss": 0.3345,
+      "step": 30740
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.88894656400496e-06,
+      "loss": 0.5115,
+      "step": 30741
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8883698128442482e-06,
+      "loss": 0.384,
+      "step": 30742
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.887793061683537e-06,
+      "loss": 0.4978,
+      "step": 30743
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8872163105228253e-06,
+      "loss": 0.4033,
+      "step": 30744
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8866395593621134e-06,
+      "loss": 0.2974,
+      "step": 30745
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.886062808201402e-06,
+      "loss": 0.463,
+      "step": 30746
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.88548605704069e-06,
+      "loss": 0.2953,
+      "step": 30747
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.884909305879978e-06,
+      "loss": 0.3078,
+      "step": 30748
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8843325547192667e-06,
+      "loss": 0.4313,
+      "step": 30749
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.883755803558555e-06,
+      "loss": 0.4111,
+      "step": 30750
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.883179052397843e-06,
+      "loss": 0.31,
+      "step": 30751
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8826023012371315e-06,
+      "loss": 0.3367,
+      "step": 30752
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.88202555007642e-06,
+      "loss": 0.5396,
+      "step": 30753
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.881448798915708e-06,
+      "loss": 0.4109,
+      "step": 30754
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.880872047754996e-06,
+      "loss": 0.4261,
+      "step": 30755
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8802952965942847e-06,
+      "loss": 0.4614,
+      "step": 30756
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.879718545433573e-06,
+      "loss": 0.4059,
+      "step": 30757
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.879141794272861e-06,
+      "loss": 0.4575,
+      "step": 30758
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.878565043112149e-06,
+      "loss": 0.3649,
+      "step": 30759
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.877988291951438e-06,
+      "loss": 0.4623,
+      "step": 30760
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.877411540790726e-06,
+      "loss": 0.3554,
+      "step": 30761
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8768347896300142e-06,
+      "loss": 0.3022,
+      "step": 30762
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8762580384693028e-06,
+      "loss": 0.5186,
+      "step": 30763
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.875681287308591e-06,
+      "loss": 0.2851,
+      "step": 30764
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.875104536147879e-06,
+      "loss": 0.5326,
+      "step": 30765
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.874527784987168e-06,
+      "loss": 0.4351,
+      "step": 30766
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.873951033826456e-06,
+      "loss": 0.4302,
+      "step": 30767
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.873374282665744e-06,
+      "loss": 0.3835,
+      "step": 30768
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8727975315050323e-06,
+      "loss": 0.3855,
+      "step": 30769
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.872220780344321e-06,
+      "loss": 0.4277,
+      "step": 30770
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.871644029183609e-06,
+      "loss": 0.4185,
+      "step": 30771
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.871067278022897e-06,
+      "loss": 0.4557,
+      "step": 30772
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.870490526862186e-06,
+      "loss": 0.3755,
+      "step": 30773
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.869913775701474e-06,
+      "loss": 0.547,
+      "step": 30774
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.869337024540762e-06,
+      "loss": 0.2884,
+      "step": 30775
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8687602733800503e-06,
+      "loss": 0.3617,
+      "step": 30776
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.868183522219339e-06,
+      "loss": 0.3534,
+      "step": 30777
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.867606771058627e-06,
+      "loss": 0.4165,
+      "step": 30778
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.867030019897915e-06,
+      "loss": 0.411,
+      "step": 30779
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8664532687372036e-06,
+      "loss": 0.3679,
+      "step": 30780
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8658765175764917e-06,
+      "loss": 0.3976,
+      "step": 30781
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.86529976641578e-06,
+      "loss": 0.4782,
+      "step": 30782
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8647230152550687e-06,
+      "loss": 0.4197,
+      "step": 30783
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.864146264094357e-06,
+      "loss": 0.3742,
+      "step": 30784
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.863569512933645e-06,
+      "loss": 0.3715,
+      "step": 30785
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.862992761772933e-06,
+      "loss": 0.5044,
+      "step": 30786
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8624160106122216e-06,
+      "loss": 0.4587,
+      "step": 30787
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8618392594515097e-06,
+      "loss": 0.4204,
+      "step": 30788
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.861262508290798e-06,
+      "loss": 0.4054,
+      "step": 30789
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8606857571300868e-06,
+      "loss": 0.4279,
+      "step": 30790
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.860109005969375e-06,
+      "loss": 0.4178,
+      "step": 30791
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.859532254808663e-06,
+      "loss": 0.4468,
+      "step": 30792
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8589555036479515e-06,
+      "loss": 0.4115,
+      "step": 30793
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8583787524872396e-06,
+      "loss": 0.3841,
+      "step": 30794
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8578020013265277e-06,
+      "loss": 0.3532,
+      "step": 30795
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.857225250165816e-06,
+      "loss": 0.3826,
+      "step": 30796
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.856648499005105e-06,
+      "loss": 0.3265,
+      "step": 30797
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.856071747844393e-06,
+      "loss": 0.4675,
+      "step": 30798
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.855494996683681e-06,
+      "loss": 0.3723,
+      "step": 30799
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8549182455229696e-06,
+      "loss": 0.4677,
+      "step": 30800
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8543414943622577e-06,
+      "loss": 0.4242,
+      "step": 30801
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8537647432015458e-06,
+      "loss": 0.3482,
+      "step": 30802
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.853187992040834e-06,
+      "loss": 0.4547,
+      "step": 30803
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.852611240880123e-06,
+      "loss": 0.3529,
+      "step": 30804
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.852034489719411e-06,
+      "loss": 0.384,
+      "step": 30805
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.851457738558699e-06,
+      "loss": 0.4436,
+      "step": 30806
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8508809873979876e-06,
+      "loss": 0.3506,
+      "step": 30807
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8503042362372757e-06,
+      "loss": 0.3257,
+      "step": 30808
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.849727485076564e-06,
+      "loss": 0.426,
+      "step": 30809
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8491507339158523e-06,
+      "loss": 0.3997,
+      "step": 30810
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8485739827551404e-06,
+      "loss": 0.4167,
+      "step": 30811
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8479972315944286e-06,
+      "loss": 0.3485,
+      "step": 30812
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.847420480433717e-06,
+      "loss": 0.295,
+      "step": 30813
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8468437292730056e-06,
+      "loss": 0.273,
+      "step": 30814
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8462669781122937e-06,
+      "loss": 0.3627,
+      "step": 30815
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.845690226951582e-06,
+      "loss": 0.4573,
+      "step": 30816
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8451134757908704e-06,
+      "loss": 0.3694,
+      "step": 30817
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8445367246301585e-06,
+      "loss": 0.3237,
+      "step": 30818
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8439599734694466e-06,
+      "loss": 0.3735,
+      "step": 30819
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8433832223087347e-06,
+      "loss": 0.2906,
+      "step": 30820
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8428064711480237e-06,
+      "loss": 0.3907,
+      "step": 30821
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8422297199873118e-06,
+      "loss": 0.3767,
+      "step": 30822
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8416529688266e-06,
+      "loss": 0.3916,
+      "step": 30823
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8410762176658884e-06,
+      "loss": 0.4005,
+      "step": 30824
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8404994665051765e-06,
+      "loss": 0.3547,
+      "step": 30825
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8399227153444646e-06,
+      "loss": 0.5118,
+      "step": 30826
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8393459641837536e-06,
+      "loss": 0.3125,
+      "step": 30827
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8387692130230417e-06,
+      "loss": 0.5115,
+      "step": 30828
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.83819246186233e-06,
+      "loss": 0.3978,
+      "step": 30829
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.837615710701618e-06,
+      "loss": 0.4327,
+      "step": 30830
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8370389595409064e-06,
+      "loss": 0.4436,
+      "step": 30831
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8364622083801945e-06,
+      "loss": 0.3272,
+      "step": 30832
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8358854572194827e-06,
+      "loss": 0.4029,
+      "step": 30833
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2.8353087060587716e-06,
+      "loss": 0.3476,
+      "step": 30834
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8347319548980597e-06,
+      "loss": 0.3121,
+      "step": 30835
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.834155203737348e-06,
+      "loss": 0.4009,
+      "step": 30836
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.833578452576636e-06,
+      "loss": 0.362,
+      "step": 30837
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8330017014159245e-06,
+      "loss": 0.4622,
+      "step": 30838
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8324249502552126e-06,
+      "loss": 0.3729,
+      "step": 30839
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8318481990945007e-06,
+      "loss": 0.4452,
+      "step": 30840
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8312714479337892e-06,
+      "loss": 0.4643,
+      "step": 30841
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8306946967730773e-06,
+      "loss": 0.3776,
+      "step": 30842
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8301179456123654e-06,
+      "loss": 0.3625,
+      "step": 30843
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8295411944516544e-06,
+      "loss": 0.4635,
+      "step": 30844
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8289644432909425e-06,
+      "loss": 0.3504,
+      "step": 30845
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8283876921302306e-06,
+      "loss": 0.3538,
+      "step": 30846
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8278109409695187e-06,
+      "loss": 0.4143,
+      "step": 30847
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8272341898088072e-06,
+      "loss": 0.3814,
+      "step": 30848
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8266574386480954e-06,
+      "loss": 0.3107,
+      "step": 30849
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8260806874873835e-06,
+      "loss": 0.4407,
+      "step": 30850
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8255039363266724e-06,
+      "loss": 0.445,
+      "step": 30851
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8249271851659605e-06,
+      "loss": 0.299,
+      "step": 30852
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8243504340052486e-06,
+      "loss": 0.3115,
+      "step": 30853
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8237736828445367e-06,
+      "loss": 0.476,
+      "step": 30854
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8231969316838253e-06,
+      "loss": 0.3122,
+      "step": 30855
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8226201805231134e-06,
+      "loss": 0.3044,
+      "step": 30856
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8220434293624015e-06,
+      "loss": 0.4675,
+      "step": 30857
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8214666782016904e-06,
+      "loss": 0.3634,
+      "step": 30858
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8208899270409786e-06,
+      "loss": 0.3481,
+      "step": 30859
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8203131758802667e-06,
+      "loss": 0.3699,
+      "step": 30860
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.819736424719555e-06,
+      "loss": 0.5579,
+      "step": 30861
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8191596735588433e-06,
+      "loss": 0.3464,
+      "step": 30862
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8185829223981314e-06,
+      "loss": 0.4741,
+      "step": 30863
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8180061712374195e-06,
+      "loss": 0.432,
+      "step": 30864
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8174294200767085e-06,
+      "loss": 0.4764,
+      "step": 30865
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8168526689159966e-06,
+      "loss": 0.4642,
+      "step": 30866
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8162759177552847e-06,
+      "loss": 0.3422,
+      "step": 30867
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8156991665945732e-06,
+      "loss": 0.5219,
+      "step": 30868
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8151224154338613e-06,
+      "loss": 0.2946,
+      "step": 30869
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8145456642731494e-06,
+      "loss": 0.4164,
+      "step": 30870
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.813968913112438e-06,
+      "loss": 0.4543,
+      "step": 30871
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.813392161951726e-06,
+      "loss": 0.3288,
+      "step": 30872
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.812815410791014e-06,
+      "loss": 0.4253,
+      "step": 30873
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8122386596303023e-06,
+      "loss": 0.3831,
+      "step": 30874
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8116619084695913e-06,
+      "loss": 0.3356,
+      "step": 30875
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8110851573088794e-06,
+      "loss": 0.3983,
+      "step": 30876
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8105084061481675e-06,
+      "loss": 0.4258,
+      "step": 30877
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.809931654987456e-06,
+      "loss": 0.4003,
+      "step": 30878
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.809354903826744e-06,
+      "loss": 0.5028,
+      "step": 30879
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8087781526660322e-06,
+      "loss": 0.4353,
+      "step": 30880
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8082014015053203e-06,
+      "loss": 0.2752,
+      "step": 30881
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8076246503446093e-06,
+      "loss": 0.3837,
+      "step": 30882
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8070478991838974e-06,
+      "loss": 0.3556,
+      "step": 30883
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8064711480231855e-06,
+      "loss": 0.3968,
+      "step": 30884
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.805894396862474e-06,
+      "loss": 0.4088,
+      "step": 30885
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.805317645701762e-06,
+      "loss": 0.367,
+      "step": 30886
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8047408945410503e-06,
+      "loss": 0.4544,
+      "step": 30887
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8041641433803392e-06,
+      "loss": 0.4351,
+      "step": 30888
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8035873922196273e-06,
+      "loss": 0.3986,
+      "step": 30889
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8030106410589154e-06,
+      "loss": 0.3859,
+      "step": 30890
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8024338898982035e-06,
+      "loss": 0.3985,
+      "step": 30891
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.801857138737492e-06,
+      "loss": 0.4102,
+      "step": 30892
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.80128038757678e-06,
+      "loss": 0.4036,
+      "step": 30893
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.8007036364160683e-06,
+      "loss": 0.3202,
+      "step": 30894
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.800126885255357e-06,
+      "loss": 0.4192,
+      "step": 30895
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7995501340946454e-06,
+      "loss": 0.4945,
+      "step": 30896
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7989733829339335e-06,
+      "loss": 0.3318,
+      "step": 30897
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7983966317732216e-06,
+      "loss": 0.4049,
+      "step": 30898
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.79781988061251e-06,
+      "loss": 0.4953,
+      "step": 30899
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.797243129451798e-06,
+      "loss": 0.3059,
+      "step": 30900
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7966663782910863e-06,
+      "loss": 0.3135,
+      "step": 30901
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.796089627130375e-06,
+      "loss": 0.4,
+      "step": 30902
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.795512875969663e-06,
+      "loss": 0.4525,
+      "step": 30903
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.794936124808951e-06,
+      "loss": 0.444,
+      "step": 30904
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.79435937364824e-06,
+      "loss": 0.4388,
+      "step": 30905
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.793782622487528e-06,
+      "loss": 0.539,
+      "step": 30906
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7932058713268162e-06,
+      "loss": 0.4032,
+      "step": 30907
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7926291201661043e-06,
+      "loss": 0.4047,
+      "step": 30908
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.792052369005393e-06,
+      "loss": 0.4012,
+      "step": 30909
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.791475617844681e-06,
+      "loss": 0.3661,
+      "step": 30910
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.790898866683969e-06,
+      "loss": 0.4565,
+      "step": 30911
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.790322115523258e-06,
+      "loss": 0.4359,
+      "step": 30912
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.789745364362546e-06,
+      "loss": 0.371,
+      "step": 30913
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7891686132018343e-06,
+      "loss": 0.4226,
+      "step": 30914
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7885918620411224e-06,
+      "loss": 0.4819,
+      "step": 30915
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.788015110880411e-06,
+      "loss": 0.4573,
+      "step": 30916
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.787438359719699e-06,
+      "loss": 0.3693,
+      "step": 30917
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.786861608558987e-06,
+      "loss": 0.433,
+      "step": 30918
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.786284857398276e-06,
+      "loss": 0.4449,
+      "step": 30919
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.785708106237564e-06,
+      "loss": 0.5478,
+      "step": 30920
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7851313550768523e-06,
+      "loss": 0.3964,
+      "step": 30921
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.784554603916141e-06,
+      "loss": 0.3987,
+      "step": 30922
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.783977852755429e-06,
+      "loss": 0.4088,
+      "step": 30923
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.783401101594717e-06,
+      "loss": 0.3771,
+      "step": 30924
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.782824350434005e-06,
+      "loss": 0.3834,
+      "step": 30925
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.782247599273294e-06,
+      "loss": 0.5156,
+      "step": 30926
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7816708481125822e-06,
+      "loss": 0.4528,
+      "step": 30927
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7810940969518703e-06,
+      "loss": 0.3661,
+      "step": 30928
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.780517345791159e-06,
+      "loss": 0.3388,
+      "step": 30929
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.779940594630447e-06,
+      "loss": 0.4287,
+      "step": 30930
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.779363843469735e-06,
+      "loss": 0.4964,
+      "step": 30931
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7787870923090236e-06,
+      "loss": 0.3802,
+      "step": 30932
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7782103411483117e-06,
+      "loss": 0.3638,
+      "step": 30933
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7776335899876e-06,
+      "loss": 0.4324,
+      "step": 30934
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.777056838826888e-06,
+      "loss": 0.2758,
+      "step": 30935
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.776480087666177e-06,
+      "loss": 0.3531,
+      "step": 30936
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.775903336505465e-06,
+      "loss": 0.3262,
+      "step": 30937
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.775326585344753e-06,
+      "loss": 0.3247,
+      "step": 30938
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7747498341840416e-06,
+      "loss": 0.4058,
+      "step": 30939
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7741730830233298e-06,
+      "loss": 0.4891,
+      "step": 30940
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.773596331862618e-06,
+      "loss": 0.4178,
+      "step": 30941
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.773019580701906e-06,
+      "loss": 0.3407,
+      "step": 30942
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.772442829541195e-06,
+      "loss": 0.4643,
+      "step": 30943
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.771866078380483e-06,
+      "loss": 0.4339,
+      "step": 30944
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.771289327219771e-06,
+      "loss": 0.3664,
+      "step": 30945
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7707125760590597e-06,
+      "loss": 0.4608,
+      "step": 30946
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7701358248983478e-06,
+      "loss": 0.3666,
+      "step": 30947
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.769559073737636e-06,
+      "loss": 0.4311,
+      "step": 30948
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.768982322576925e-06,
+      "loss": 0.3218,
+      "step": 30949
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.768405571416213e-06,
+      "loss": 0.4637,
+      "step": 30950
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.767828820255501e-06,
+      "loss": 0.362,
+      "step": 30951
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.767252069094789e-06,
+      "loss": 0.3342,
+      "step": 30952
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7666753179340777e-06,
+      "loss": 0.3504,
+      "step": 30953
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.766098566773366e-06,
+      "loss": 0.3402,
+      "step": 30954
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.765521815612654e-06,
+      "loss": 0.4192,
+      "step": 30955
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7649450644519425e-06,
+      "loss": 0.3386,
+      "step": 30956
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.764368313291231e-06,
+      "loss": 0.4452,
+      "step": 30957
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.763791562130519e-06,
+      "loss": 0.4916,
+      "step": 30958
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.763214810969807e-06,
+      "loss": 0.3817,
+      "step": 30959
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7626380598090957e-06,
+      "loss": 0.4234,
+      "step": 30960
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.762061308648384e-06,
+      "loss": 0.472,
+      "step": 30961
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.761484557487672e-06,
+      "loss": 0.3994,
+      "step": 30962
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7609078063269605e-06,
+      "loss": 0.441,
+      "step": 30963
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7603310551662486e-06,
+      "loss": 0.4139,
+      "step": 30964
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7597543040055367e-06,
+      "loss": 0.5324,
+      "step": 30965
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7591775528448257e-06,
+      "loss": 0.5184,
+      "step": 30966
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7586008016841138e-06,
+      "loss": 0.4552,
+      "step": 30967
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.758024050523402e-06,
+      "loss": 0.3331,
+      "step": 30968
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.75744729936269e-06,
+      "loss": 0.3774,
+      "step": 30969
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7568705482019785e-06,
+      "loss": 0.5096,
+      "step": 30970
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7562937970412666e-06,
+      "loss": 0.4014,
+      "step": 30971
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7557170458805547e-06,
+      "loss": 0.4268,
+      "step": 30972
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7551402947198437e-06,
+      "loss": 0.4472,
+      "step": 30973
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.754563543559132e-06,
+      "loss": 0.316,
+      "step": 30974
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.75398679239842e-06,
+      "loss": 0.3598,
+      "step": 30975
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.753410041237708e-06,
+      "loss": 0.5242,
+      "step": 30976
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7528332900769965e-06,
+      "loss": 0.5218,
+      "step": 30977
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7522565389162847e-06,
+      "loss": 0.3481,
+      "step": 30978
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7516797877555728e-06,
+      "loss": 0.4121,
+      "step": 30979
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7511030365948617e-06,
+      "loss": 0.5001,
+      "step": 30980
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.75052628543415e-06,
+      "loss": 0.3602,
+      "step": 30981
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.749949534273438e-06,
+      "loss": 0.4217,
+      "step": 30982
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7493727831127265e-06,
+      "loss": 0.362,
+      "step": 30983
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7487960319520146e-06,
+      "loss": 0.4139,
+      "step": 30984
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7482192807913027e-06,
+      "loss": 0.4318,
+      "step": 30985
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.747642529630591e-06,
+      "loss": 0.4045,
+      "step": 30986
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7470657784698793e-06,
+      "loss": 0.3276,
+      "step": 30987
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.746489027309168e-06,
+      "loss": 0.4199,
+      "step": 30988
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.745912276148456e-06,
+      "loss": 0.3468,
+      "step": 30989
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7453355249877445e-06,
+      "loss": 0.4645,
+      "step": 30990
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7447587738270326e-06,
+      "loss": 0.3542,
+      "step": 30991
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7441820226663207e-06,
+      "loss": 0.324,
+      "step": 30992
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7436052715056093e-06,
+      "loss": 0.4243,
+      "step": 30993
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7430285203448974e-06,
+      "loss": 0.4279,
+      "step": 30994
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7424517691841855e-06,
+      "loss": 0.4115,
+      "step": 30995
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7418750180234736e-06,
+      "loss": 0.3622,
+      "step": 30996
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7412982668627625e-06,
+      "loss": 0.3978,
+      "step": 30997
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7407215157020506e-06,
+      "loss": 0.4734,
+      "step": 30998
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7401447645413388e-06,
+      "loss": 0.4652,
+      "step": 30999
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7395680133806273e-06,
+      "loss": 0.4631,
+      "step": 31000
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7389912622199154e-06,
+      "loss": 0.4395,
+      "step": 31001
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7384145110592035e-06,
+      "loss": 0.3343,
+      "step": 31002
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7378377598984916e-06,
+      "loss": 0.3445,
+      "step": 31003
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7372610087377806e-06,
+      "loss": 0.4286,
+      "step": 31004
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7366842575770687e-06,
+      "loss": 0.3557,
+      "step": 31005
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7361075064163568e-06,
+      "loss": 0.3498,
+      "step": 31006
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7355307552556453e-06,
+      "loss": 0.4459,
+      "step": 31007
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7349540040949334e-06,
+      "loss": 0.4287,
+      "step": 31008
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7343772529342215e-06,
+      "loss": 0.2706,
+      "step": 31009
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7338005017735105e-06,
+      "loss": 0.3593,
+      "step": 31010
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7332237506127986e-06,
+      "loss": 0.3919,
+      "step": 31011
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.7326469994520867e-06,
+      "loss": 0.428,
+      "step": 31012
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2.732070248291375e-06,
+      "loss": 0.395,
+      "step": 31013
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7314934971306633e-06,
+      "loss": 0.4266,
+      "step": 31014
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7309167459699515e-06,
+      "loss": 0.406,
+      "step": 31015
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7303399948092396e-06,
+      "loss": 0.2869,
+      "step": 31016
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.729763243648528e-06,
+      "loss": 0.3433,
+      "step": 31017
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7291864924878166e-06,
+      "loss": 0.3024,
+      "step": 31018
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7286097413271047e-06,
+      "loss": 0.3618,
+      "step": 31019
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.728032990166393e-06,
+      "loss": 0.4678,
+      "step": 31020
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7274562390056814e-06,
+      "loss": 0.3072,
+      "step": 31021
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7268794878449695e-06,
+      "loss": 0.3804,
+      "step": 31022
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7263027366842576e-06,
+      "loss": 0.412,
+      "step": 31023
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.725725985523546e-06,
+      "loss": 0.4683,
+      "step": 31024
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7251492343628342e-06,
+      "loss": 0.4796,
+      "step": 31025
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7245724832021223e-06,
+      "loss": 0.4262,
+      "step": 31026
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7239957320414113e-06,
+      "loss": 0.3825,
+      "step": 31027
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7234189808806994e-06,
+      "loss": 0.4452,
+      "step": 31028
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7228422297199875e-06,
+      "loss": 0.3594,
+      "step": 31029
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7222654785592756e-06,
+      "loss": 0.4237,
+      "step": 31030
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.721688727398564e-06,
+      "loss": 0.3263,
+      "step": 31031
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7211119762378523e-06,
+      "loss": 0.4516,
+      "step": 31032
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7205352250771404e-06,
+      "loss": 0.3379,
+      "step": 31033
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7199584739164293e-06,
+      "loss": 0.4658,
+      "step": 31034
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7193817227557174e-06,
+      "loss": 0.4284,
+      "step": 31035
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7188049715950055e-06,
+      "loss": 0.38,
+      "step": 31036
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7182282204342937e-06,
+      "loss": 0.3481,
+      "step": 31037
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.717651469273582e-06,
+      "loss": 0.4896,
+      "step": 31038
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7170747181128703e-06,
+      "loss": 0.3921,
+      "step": 31039
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7164979669521584e-06,
+      "loss": 0.3932,
+      "step": 31040
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7159212157914474e-06,
+      "loss": 0.3163,
+      "step": 31041
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7153444646307355e-06,
+      "loss": 0.4867,
+      "step": 31042
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7147677134700236e-06,
+      "loss": 0.424,
+      "step": 31043
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.714190962309312e-06,
+      "loss": 0.4574,
+      "step": 31044
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7136142111486002e-06,
+      "loss": 0.5354,
+      "step": 31045
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7130374599878883e-06,
+      "loss": 0.4011,
+      "step": 31046
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7124607088271764e-06,
+      "loss": 0.3043,
+      "step": 31047
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.711883957666465e-06,
+      "loss": 0.3831,
+      "step": 31048
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7113072065057535e-06,
+      "loss": 0.3663,
+      "step": 31049
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7107304553450416e-06,
+      "loss": 0.441,
+      "step": 31050
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.71015370418433e-06,
+      "loss": 0.4513,
+      "step": 31051
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7095769530236182e-06,
+      "loss": 0.42,
+      "step": 31052
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7090002018629064e-06,
+      "loss": 0.3218,
+      "step": 31053
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.708423450702195e-06,
+      "loss": 0.3292,
+      "step": 31054
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.707846699541483e-06,
+      "loss": 0.3845,
+      "step": 31055
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.707269948380771e-06,
+      "loss": 0.3916,
+      "step": 31056
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7066931972200592e-06,
+      "loss": 0.328,
+      "step": 31057
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.706116446059348e-06,
+      "loss": 0.4778,
+      "step": 31058
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7055396948986363e-06,
+      "loss": 0.4189,
+      "step": 31059
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7049629437379244e-06,
+      "loss": 0.3495,
+      "step": 31060
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.704386192577213e-06,
+      "loss": 0.2884,
+      "step": 31061
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.703809441416501e-06,
+      "loss": 0.2946,
+      "step": 31062
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.703232690255789e-06,
+      "loss": 0.4225,
+      "step": 31063
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7026559390950772e-06,
+      "loss": 0.4078,
+      "step": 31064
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.702079187934366e-06,
+      "loss": 0.4423,
+      "step": 31065
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7015024367736543e-06,
+      "loss": 0.4744,
+      "step": 31066
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.7009256856129424e-06,
+      "loss": 0.3471,
+      "step": 31067
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.700348934452231e-06,
+      "loss": 0.3617,
+      "step": 31068
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.699772183291519e-06,
+      "loss": 0.3792,
+      "step": 31069
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.699195432130807e-06,
+      "loss": 0.4166,
+      "step": 31070
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.698618680970096e-06,
+      "loss": 0.5113,
+      "step": 31071
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6980419298093842e-06,
+      "loss": 0.5367,
+      "step": 31072
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6974651786486723e-06,
+      "loss": 0.4259,
+      "step": 31073
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6968884274879604e-06,
+      "loss": 0.4896,
+      "step": 31074
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.696311676327249e-06,
+      "loss": 0.4249,
+      "step": 31075
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.695734925166537e-06,
+      "loss": 0.41,
+      "step": 31076
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.695158174005825e-06,
+      "loss": 0.483,
+      "step": 31077
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6945814228451137e-06,
+      "loss": 0.4497,
+      "step": 31078
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.694004671684402e-06,
+      "loss": 0.4231,
+      "step": 31079
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6934279205236904e-06,
+      "loss": 0.3794,
+      "step": 31080
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6928511693629785e-06,
+      "loss": 0.3652,
+      "step": 31081
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.692274418202267e-06,
+      "loss": 0.3313,
+      "step": 31082
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.691697667041555e-06,
+      "loss": 0.3874,
+      "step": 31083
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6911209158808432e-06,
+      "loss": 0.4489,
+      "step": 31084
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6905441647201318e-06,
+      "loss": 0.551,
+      "step": 31085
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.68996741355942e-06,
+      "loss": 0.4212,
+      "step": 31086
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.689390662398708e-06,
+      "loss": 0.3448,
+      "step": 31087
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.688813911237997e-06,
+      "loss": 0.3579,
+      "step": 31088
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.688237160077285e-06,
+      "loss": 0.3738,
+      "step": 31089
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.687660408916573e-06,
+      "loss": 0.414,
+      "step": 31090
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6870836577558613e-06,
+      "loss": 0.4712,
+      "step": 31091
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.68650690659515e-06,
+      "loss": 0.4764,
+      "step": 31092
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.685930155434438e-06,
+      "loss": 0.3869,
+      "step": 31093
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.685353404273726e-06,
+      "loss": 0.3941,
+      "step": 31094
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.684776653113015e-06,
+      "loss": 0.4456,
+      "step": 31095
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.684199901952303e-06,
+      "loss": 0.359,
+      "step": 31096
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.683623150791591e-06,
+      "loss": 0.3114,
+      "step": 31097
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6830463996308793e-06,
+      "loss": 0.2822,
+      "step": 31098
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.682469648470168e-06,
+      "loss": 0.3207,
+      "step": 31099
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.681892897309456e-06,
+      "loss": 0.4104,
+      "step": 31100
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.681316146148744e-06,
+      "loss": 0.3698,
+      "step": 31101
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.680739394988033e-06,
+      "loss": 0.3903,
+      "step": 31102
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.680162643827321e-06,
+      "loss": 0.4997,
+      "step": 31103
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6795858926666092e-06,
+      "loss": 0.4731,
+      "step": 31104
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6790091415058977e-06,
+      "loss": 0.4605,
+      "step": 31105
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.678432390345186e-06,
+      "loss": 0.4121,
+      "step": 31106
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.677855639184474e-06,
+      "loss": 0.3885,
+      "step": 31107
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.677278888023762e-06,
+      "loss": 0.485,
+      "step": 31108
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6767021368630506e-06,
+      "loss": 0.4208,
+      "step": 31109
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.676125385702339e-06,
+      "loss": 0.3261,
+      "step": 31110
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6755486345416272e-06,
+      "loss": 0.3035,
+      "step": 31111
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6749718833809158e-06,
+      "loss": 0.3833,
+      "step": 31112
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.674395132220204e-06,
+      "loss": 0.5094,
+      "step": 31113
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.673818381059492e-06,
+      "loss": 0.4483,
+      "step": 31114
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6732416298987805e-06,
+      "loss": 0.3072,
+      "step": 31115
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6726648787380686e-06,
+      "loss": 0.4242,
+      "step": 31116
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6720881275773567e-06,
+      "loss": 0.4977,
+      "step": 31117
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.671511376416645e-06,
+      "loss": 0.3676,
+      "step": 31118
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.670934625255934e-06,
+      "loss": 0.4442,
+      "step": 31119
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.670357874095222e-06,
+      "loss": 0.3459,
+      "step": 31120
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.66978112293451e-06,
+      "loss": 0.4027,
+      "step": 31121
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6692043717737986e-06,
+      "loss": 0.4106,
+      "step": 31122
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6686276206130867e-06,
+      "loss": 0.3504,
+      "step": 31123
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6680508694523748e-06,
+      "loss": 0.3544,
+      "step": 31124
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.667474118291663e-06,
+      "loss": 0.3968,
+      "step": 31125
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.666897367130952e-06,
+      "loss": 0.4657,
+      "step": 31126
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.66632061597024e-06,
+      "loss": 0.3833,
+      "step": 31127
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.665743864809528e-06,
+      "loss": 0.4321,
+      "step": 31128
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6651671136488166e-06,
+      "loss": 0.3004,
+      "step": 31129
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6645903624881047e-06,
+      "loss": 0.4333,
+      "step": 31130
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.664013611327393e-06,
+      "loss": 0.504,
+      "step": 31131
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6634368601666818e-06,
+      "loss": 0.3775,
+      "step": 31132
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.66286010900597e-06,
+      "loss": 0.4943,
+      "step": 31133
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.662283357845258e-06,
+      "loss": 0.3981,
+      "step": 31134
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.661706606684546e-06,
+      "loss": 0.4033,
+      "step": 31135
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6611298555238346e-06,
+      "loss": 0.3719,
+      "step": 31136
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6605531043631227e-06,
+      "loss": 0.3886,
+      "step": 31137
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.659976353202411e-06,
+      "loss": 0.3289,
+      "step": 31138
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6593996020416994e-06,
+      "loss": 0.4786,
+      "step": 31139
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6588228508809875e-06,
+      "loss": 0.4282,
+      "step": 31140
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.658246099720276e-06,
+      "loss": 0.3108,
+      "step": 31141
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.657669348559564e-06,
+      "loss": 0.4882,
+      "step": 31142
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6570925973988527e-06,
+      "loss": 0.2889,
+      "step": 31143
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6565158462381408e-06,
+      "loss": 0.4847,
+      "step": 31144
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.655939095077429e-06,
+      "loss": 0.4649,
+      "step": 31145
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6553623439167174e-06,
+      "loss": 0.4449,
+      "step": 31146
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6547855927560055e-06,
+      "loss": 0.4207,
+      "step": 31147
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6542088415952936e-06,
+      "loss": 0.4911,
+      "step": 31148
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6536320904345826e-06,
+      "loss": 0.4669,
+      "step": 31149
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6530553392738707e-06,
+      "loss": 0.4183,
+      "step": 31150
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.652478588113159e-06,
+      "loss": 0.4382,
+      "step": 31151
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.651901836952447e-06,
+      "loss": 0.43,
+      "step": 31152
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6513250857917354e-06,
+      "loss": 0.4534,
+      "step": 31153
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6507483346310235e-06,
+      "loss": 0.3995,
+      "step": 31154
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6501715834703116e-06,
+      "loss": 0.4661,
+      "step": 31155
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6495948323096006e-06,
+      "loss": 0.5141,
+      "step": 31156
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6490180811488887e-06,
+      "loss": 0.3104,
+      "step": 31157
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.648441329988177e-06,
+      "loss": 0.3584,
+      "step": 31158
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.647864578827465e-06,
+      "loss": 0.3908,
+      "step": 31159
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6472878276667535e-06,
+      "loss": 0.4139,
+      "step": 31160
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6467110765060416e-06,
+      "loss": 0.4083,
+      "step": 31161
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6461343253453297e-06,
+      "loss": 0.4896,
+      "step": 31162
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6455575741846186e-06,
+      "loss": 0.3423,
+      "step": 31163
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6449808230239067e-06,
+      "loss": 0.4454,
+      "step": 31164
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.644404071863195e-06,
+      "loss": 0.4214,
+      "step": 31165
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6438273207024834e-06,
+      "loss": 0.3674,
+      "step": 31166
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6432505695417715e-06,
+      "loss": 0.4366,
+      "step": 31167
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6426738183810596e-06,
+      "loss": 0.3158,
+      "step": 31168
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6420970672203477e-06,
+      "loss": 0.3556,
+      "step": 31169
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6415203160596362e-06,
+      "loss": 0.3679,
+      "step": 31170
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6409435648989244e-06,
+      "loss": 0.4831,
+      "step": 31171
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.640366813738213e-06,
+      "loss": 0.5132,
+      "step": 31172
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6397900625775014e-06,
+      "loss": 0.4381,
+      "step": 31173
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6392133114167895e-06,
+      "loss": 0.4219,
+      "step": 31174
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6386365602560776e-06,
+      "loss": 0.4562,
+      "step": 31175
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.638059809095366e-06,
+      "loss": 0.4612,
+      "step": 31176
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6374830579346543e-06,
+      "loss": 0.4724,
+      "step": 31177
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6369063067739424e-06,
+      "loss": 0.4424,
+      "step": 31178
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6363295556132305e-06,
+      "loss": 0.5241,
+      "step": 31179
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6357528044525194e-06,
+      "loss": 0.416,
+      "step": 31180
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6351760532918076e-06,
+      "loss": 0.3591,
+      "step": 31181
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6345993021310957e-06,
+      "loss": 0.2413,
+      "step": 31182
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.634022550970384e-06,
+      "loss": 0.3013,
+      "step": 31183
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6334457998096723e-06,
+      "loss": 0.4958,
+      "step": 31184
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6328690486489604e-06,
+      "loss": 0.4306,
+      "step": 31185
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6322922974882485e-06,
+      "loss": 0.4912,
+      "step": 31186
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6317155463275375e-06,
+      "loss": 0.4753,
+      "step": 31187
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6311387951668256e-06,
+      "loss": 0.2991,
+      "step": 31188
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6305620440061137e-06,
+      "loss": 0.2438,
+      "step": 31189
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6299852928454022e-06,
+      "loss": 0.412,
+      "step": 31190
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2.6294085416846903e-06,
+      "loss": 0.3016,
+      "step": 31191
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6288317905239784e-06,
+      "loss": 0.4912,
+      "step": 31192
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6282550393632674e-06,
+      "loss": 0.4106,
+      "step": 31193
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6276782882025555e-06,
+      "loss": 0.4953,
+      "step": 31194
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6271015370418436e-06,
+      "loss": 0.4715,
+      "step": 31195
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6265247858811317e-06,
+      "loss": 0.3171,
+      "step": 31196
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6259480347204203e-06,
+      "loss": 0.3333,
+      "step": 31197
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6253712835597084e-06,
+      "loss": 0.4537,
+      "step": 31198
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6247945323989965e-06,
+      "loss": 0.4672,
+      "step": 31199
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.624217781238285e-06,
+      "loss": 0.3518,
+      "step": 31200
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.623641030077573e-06,
+      "loss": 0.4149,
+      "step": 31201
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6230642789168616e-06,
+      "loss": 0.4935,
+      "step": 31202
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6224875277561498e-06,
+      "loss": 0.4703,
+      "step": 31203
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6219107765954383e-06,
+      "loss": 0.4309,
+      "step": 31204
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6213340254347264e-06,
+      "loss": 0.496,
+      "step": 31205
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6207572742740145e-06,
+      "loss": 0.4472,
+      "step": 31206
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.620180523113303e-06,
+      "loss": 0.3969,
+      "step": 31207
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.619603771952591e-06,
+      "loss": 0.3794,
+      "step": 31208
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6190270207918793e-06,
+      "loss": 0.388,
+      "step": 31209
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.618450269631168e-06,
+      "loss": 0.4618,
+      "step": 31210
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6178735184704563e-06,
+      "loss": 0.4537,
+      "step": 31211
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6172967673097444e-06,
+      "loss": 0.4381,
+      "step": 31212
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6167200161490325e-06,
+      "loss": 0.4263,
+      "step": 31213
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.616143264988321e-06,
+      "loss": 0.3858,
+      "step": 31214
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.615566513827609e-06,
+      "loss": 0.3731,
+      "step": 31215
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6149897626668973e-06,
+      "loss": 0.4664,
+      "step": 31216
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6144130115061862e-06,
+      "loss": 0.445,
+      "step": 31217
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6138362603454743e-06,
+      "loss": 0.4524,
+      "step": 31218
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6132595091847625e-06,
+      "loss": 0.3427,
+      "step": 31219
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6126827580240506e-06,
+      "loss": 0.423,
+      "step": 31220
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.612106006863339e-06,
+      "loss": 0.3809,
+      "step": 31221
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.611529255702627e-06,
+      "loss": 0.3885,
+      "step": 31222
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6109525045419153e-06,
+      "loss": 0.4425,
+      "step": 31223
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6103757533812043e-06,
+      "loss": 0.4493,
+      "step": 31224
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6097990022204924e-06,
+      "loss": 0.407,
+      "step": 31225
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6092222510597805e-06,
+      "loss": 0.359,
+      "step": 31226
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.608645499899069e-06,
+      "loss": 0.2855,
+      "step": 31227
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.608068748738357e-06,
+      "loss": 0.3705,
+      "step": 31228
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6074919975776452e-06,
+      "loss": 0.456,
+      "step": 31229
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6069152464169333e-06,
+      "loss": 0.3204,
+      "step": 31230
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.606338495256222e-06,
+      "loss": 0.3244,
+      "step": 31231
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.60576174409551e-06,
+      "loss": 0.4101,
+      "step": 31232
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6051849929347985e-06,
+      "loss": 0.5389,
+      "step": 31233
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.604608241774087e-06,
+      "loss": 0.4846,
+      "step": 31234
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.604031490613375e-06,
+      "loss": 0.4348,
+      "step": 31235
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6034547394526633e-06,
+      "loss": 0.379,
+      "step": 31236
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.6028779882919514e-06,
+      "loss": 0.3852,
+      "step": 31237
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.60230123713124e-06,
+      "loss": 0.4067,
+      "step": 31238
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.601724485970528e-06,
+      "loss": 0.4874,
+      "step": 31239
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.601147734809816e-06,
+      "loss": 0.3664,
+      "step": 31240
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.600570983649105e-06,
+      "loss": 0.4025,
+      "step": 31241
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.599994232488393e-06,
+      "loss": 0.4221,
+      "step": 31242
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5994174813276813e-06,
+      "loss": 0.3544,
+      "step": 31243
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.59884073016697e-06,
+      "loss": 0.4425,
+      "step": 31244
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.598263979006258e-06,
+      "loss": 0.3806,
+      "step": 31245
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.597687227845546e-06,
+      "loss": 0.416,
+      "step": 31246
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.597110476684834e-06,
+      "loss": 0.3571,
+      "step": 31247
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.596533725524123e-06,
+      "loss": 0.39,
+      "step": 31248
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5959569743634112e-06,
+      "loss": 0.3541,
+      "step": 31249
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5953802232026993e-06,
+      "loss": 0.5222,
+      "step": 31250
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.594803472041988e-06,
+      "loss": 0.4286,
+      "step": 31251
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.594226720881276e-06,
+      "loss": 0.3345,
+      "step": 31252
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.593649969720564e-06,
+      "loss": 0.3114,
+      "step": 31253
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.593073218559853e-06,
+      "loss": 0.3459,
+      "step": 31254
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.592496467399141e-06,
+      "loss": 0.4827,
+      "step": 31255
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5919197162384293e-06,
+      "loss": 0.3553,
+      "step": 31256
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5913429650777174e-06,
+      "loss": 0.2852,
+      "step": 31257
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.590766213917006e-06,
+      "loss": 0.4454,
+      "step": 31258
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.590189462756294e-06,
+      "loss": 0.542,
+      "step": 31259
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.589612711595582e-06,
+      "loss": 0.3711,
+      "step": 31260
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5890359604348706e-06,
+      "loss": 0.4057,
+      "step": 31261
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5884592092741588e-06,
+      "loss": 0.3252,
+      "step": 31262
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.587882458113447e-06,
+      "loss": 0.4036,
+      "step": 31263
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5873057069527354e-06,
+      "loss": 0.3234,
+      "step": 31264
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.586728955792024e-06,
+      "loss": 0.4305,
+      "step": 31265
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.586152204631312e-06,
+      "loss": 0.394,
+      "step": 31266
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5855754534706e-06,
+      "loss": 0.3246,
+      "step": 31267
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5849987023098887e-06,
+      "loss": 0.3017,
+      "step": 31268
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5844219511491768e-06,
+      "loss": 0.3091,
+      "step": 31269
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.583845199988465e-06,
+      "loss": 0.3851,
+      "step": 31270
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.583268448827754e-06,
+      "loss": 0.4368,
+      "step": 31271
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.582691697667042e-06,
+      "loss": 0.3901,
+      "step": 31272
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.58211494650633e-06,
+      "loss": 0.3753,
+      "step": 31273
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.581538195345618e-06,
+      "loss": 0.4388,
+      "step": 31274
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5809614441849067e-06,
+      "loss": 0.3947,
+      "step": 31275
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.580384693024195e-06,
+      "loss": 0.4092,
+      "step": 31276
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.579807941863483e-06,
+      "loss": 0.5355,
+      "step": 31277
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.579231190702772e-06,
+      "loss": 0.4,
+      "step": 31278
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.57865443954206e-06,
+      "loss": 0.4148,
+      "step": 31279
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.578077688381348e-06,
+      "loss": 0.4662,
+      "step": 31280
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.577500937220636e-06,
+      "loss": 0.3919,
+      "step": 31281
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5769241860599247e-06,
+      "loss": 0.3891,
+      "step": 31282
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.576347434899213e-06,
+      "loss": 0.4004,
+      "step": 31283
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.575770683738501e-06,
+      "loss": 0.4234,
+      "step": 31284
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.57519393257779e-06,
+      "loss": 0.4207,
+      "step": 31285
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.574617181417078e-06,
+      "loss": 0.54,
+      "step": 31286
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.574040430256366e-06,
+      "loss": 0.4123,
+      "step": 31287
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5734636790956547e-06,
+      "loss": 0.3558,
+      "step": 31288
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5728869279349428e-06,
+      "loss": 0.3727,
+      "step": 31289
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.572310176774231e-06,
+      "loss": 0.2752,
+      "step": 31290
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.571733425613519e-06,
+      "loss": 0.3436,
+      "step": 31291
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5711566744528075e-06,
+      "loss": 0.3378,
+      "step": 31292
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5705799232920956e-06,
+      "loss": 0.4908,
+      "step": 31293
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.570003172131384e-06,
+      "loss": 0.3137,
+      "step": 31294
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5694264209706727e-06,
+      "loss": 0.4482,
+      "step": 31295
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.568849669809961e-06,
+      "loss": 0.3848,
+      "step": 31296
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.568272918649249e-06,
+      "loss": 0.4731,
+      "step": 31297
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.567696167488537e-06,
+      "loss": 0.3686,
+      "step": 31298
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5671194163278255e-06,
+      "loss": 0.4773,
+      "step": 31299
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5665426651671137e-06,
+      "loss": 0.3646,
+      "step": 31300
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5659659140064018e-06,
+      "loss": 0.2944,
+      "step": 31301
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5653891628456907e-06,
+      "loss": 0.4273,
+      "step": 31302
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.564812411684979e-06,
+      "loss": 0.4178,
+      "step": 31303
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.564235660524267e-06,
+      "loss": 0.5006,
+      "step": 31304
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5636589093635555e-06,
+      "loss": 0.4526,
+      "step": 31305
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5630821582028436e-06,
+      "loss": 0.4109,
+      "step": 31306
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5625054070421317e-06,
+      "loss": 0.4469,
+      "step": 31307
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.56192865588142e-06,
+      "loss": 0.263,
+      "step": 31308
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5613519047207088e-06,
+      "loss": 0.3989,
+      "step": 31309
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.560775153559997e-06,
+      "loss": 0.4599,
+      "step": 31310
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.560198402399285e-06,
+      "loss": 0.3111,
+      "step": 31311
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5596216512385735e-06,
+      "loss": 0.3371,
+      "step": 31312
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5590449000778616e-06,
+      "loss": 0.45,
+      "step": 31313
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5584681489171497e-06,
+      "loss": 0.3454,
+      "step": 31314
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5578913977564387e-06,
+      "loss": 0.3159,
+      "step": 31315
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5573146465957268e-06,
+      "loss": 0.4279,
+      "step": 31316
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.556737895435015e-06,
+      "loss": 0.4305,
+      "step": 31317
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.556161144274303e-06,
+      "loss": 0.4253,
+      "step": 31318
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5555843931135915e-06,
+      "loss": 0.3675,
+      "step": 31319
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5550076419528796e-06,
+      "loss": 0.3877,
+      "step": 31320
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5544308907921677e-06,
+      "loss": 0.4552,
+      "step": 31321
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5538541396314563e-06,
+      "loss": 0.3735,
+      "step": 31322
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5532773884707444e-06,
+      "loss": 0.4668,
+      "step": 31323
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5527006373100325e-06,
+      "loss": 0.443,
+      "step": 31324
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.552123886149321e-06,
+      "loss": 0.3689,
+      "step": 31325
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5515471349886096e-06,
+      "loss": 0.3465,
+      "step": 31326
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5509703838278977e-06,
+      "loss": 0.4427,
+      "step": 31327
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5503936326671858e-06,
+      "loss": 0.4354,
+      "step": 31328
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5498168815064743e-06,
+      "loss": 0.3823,
+      "step": 31329
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5492401303457624e-06,
+      "loss": 0.4425,
+      "step": 31330
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5486633791850505e-06,
+      "loss": 0.3685,
+      "step": 31331
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5480866280243395e-06,
+      "loss": 0.38,
+      "step": 31332
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5475098768636276e-06,
+      "loss": 0.4921,
+      "step": 31333
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5469331257029157e-06,
+      "loss": 0.4376,
+      "step": 31334
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.546356374542204e-06,
+      "loss": 0.3454,
+      "step": 31335
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5457796233814923e-06,
+      "loss": 0.499,
+      "step": 31336
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5452028722207805e-06,
+      "loss": 0.378,
+      "step": 31337
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5446261210600686e-06,
+      "loss": 0.3207,
+      "step": 31338
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5440493698993575e-06,
+      "loss": 0.4421,
+      "step": 31339
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5434726187386456e-06,
+      "loss": 0.4221,
+      "step": 31340
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5428958675779337e-06,
+      "loss": 0.3395,
+      "step": 31341
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.542319116417222e-06,
+      "loss": 0.5038,
+      "step": 31342
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5417423652565104e-06,
+      "loss": 0.4672,
+      "step": 31343
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5411656140957985e-06,
+      "loss": 0.419,
+      "step": 31344
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5405888629350866e-06,
+      "loss": 0.3324,
+      "step": 31345
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5400121117743755e-06,
+      "loss": 0.3918,
+      "step": 31346
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5394353606136637e-06,
+      "loss": 0.4604,
+      "step": 31347
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5388586094529518e-06,
+      "loss": 0.3475,
+      "step": 31348
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5382818582922403e-06,
+      "loss": 0.3552,
+      "step": 31349
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5377051071315284e-06,
+      "loss": 0.4114,
+      "step": 31350
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5371283559708165e-06,
+      "loss": 0.2759,
+      "step": 31351
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5365516048101046e-06,
+      "loss": 0.3625,
+      "step": 31352
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.535974853649393e-06,
+      "loss": 0.3947,
+      "step": 31353
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5353981024886813e-06,
+      "loss": 0.3859,
+      "step": 31354
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5348213513279694e-06,
+      "loss": 0.4631,
+      "step": 31355
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5342446001672583e-06,
+      "loss": 0.4404,
+      "step": 31356
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5336678490065464e-06,
+      "loss": 0.434,
+      "step": 31357
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5330910978458345e-06,
+      "loss": 0.4451,
+      "step": 31358
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5325143466851227e-06,
+      "loss": 0.4564,
+      "step": 31359
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.531937595524411e-06,
+      "loss": 0.4421,
+      "step": 31360
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5313608443636993e-06,
+      "loss": 0.3965,
+      "step": 31361
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5307840932029874e-06,
+      "loss": 0.2863,
+      "step": 31362
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5302073420422764e-06,
+      "loss": 0.2574,
+      "step": 31363
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5296305908815645e-06,
+      "loss": 0.4088,
+      "step": 31364
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5290538397208526e-06,
+      "loss": 0.3171,
+      "step": 31365
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.528477088560141e-06,
+      "loss": 0.5778,
+      "step": 31366
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5279003373994292e-06,
+      "loss": 0.3599,
+      "step": 31367
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5273235862387173e-06,
+      "loss": 0.4098,
+      "step": 31368
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5267468350780054e-06,
+      "loss": 0.4474,
+      "step": 31369
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2.5261700839172944e-06,
+      "loss": 0.4122,
+      "step": 31370
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5255933327565825e-06,
+      "loss": 0.4289,
+      "step": 31371
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5250165815958706e-06,
+      "loss": 0.3513,
+      "step": 31372
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.524439830435159e-06,
+      "loss": 0.3345,
+      "step": 31373
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5238630792744472e-06,
+      "loss": 0.4267,
+      "step": 31374
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5232863281137354e-06,
+      "loss": 0.4236,
+      "step": 31375
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.522709576953024e-06,
+      "loss": 0.4942,
+      "step": 31376
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5221328257923124e-06,
+      "loss": 0.3786,
+      "step": 31377
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5215560746316005e-06,
+      "loss": 0.4646,
+      "step": 31378
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5209793234708886e-06,
+      "loss": 0.3586,
+      "step": 31379
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.520402572310177e-06,
+      "loss": 0.3399,
+      "step": 31380
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5198258211494653e-06,
+      "loss": 0.3444,
+      "step": 31381
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5192490699887534e-06,
+      "loss": 0.3484,
+      "step": 31382
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.518672318828042e-06,
+      "loss": 0.5375,
+      "step": 31383
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.51809556766733e-06,
+      "loss": 0.3661,
+      "step": 31384
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.517518816506618e-06,
+      "loss": 0.4577,
+      "step": 31385
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5169420653459067e-06,
+      "loss": 0.5121,
+      "step": 31386
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.516365314185195e-06,
+      "loss": 0.4312,
+      "step": 31387
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5157885630244833e-06,
+      "loss": 0.4014,
+      "step": 31388
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5152118118637714e-06,
+      "loss": 0.3073,
+      "step": 31389
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.51463506070306e-06,
+      "loss": 0.4614,
+      "step": 31390
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.514058309542348e-06,
+      "loss": 0.414,
+      "step": 31391
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.513481558381636e-06,
+      "loss": 0.3339,
+      "step": 31392
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.512904807220925e-06,
+      "loss": 0.2992,
+      "step": 31393
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5123280560602132e-06,
+      "loss": 0.3996,
+      "step": 31394
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5117513048995013e-06,
+      "loss": 0.3794,
+      "step": 31395
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5111745537387894e-06,
+      "loss": 0.528,
+      "step": 31396
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.510597802578078e-06,
+      "loss": 0.395,
+      "step": 31397
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.510021051417366e-06,
+      "loss": 0.3571,
+      "step": 31398
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.509444300256654e-06,
+      "loss": 0.508,
+      "step": 31399
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.508867549095943e-06,
+      "loss": 0.3678,
+      "step": 31400
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5082907979352313e-06,
+      "loss": 0.4,
+      "step": 31401
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5077140467745194e-06,
+      "loss": 0.534,
+      "step": 31402
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5071372956138075e-06,
+      "loss": 0.3628,
+      "step": 31403
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.506560544453096e-06,
+      "loss": 0.3802,
+      "step": 31404
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.505983793292384e-06,
+      "loss": 0.3653,
+      "step": 31405
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5054070421316722e-06,
+      "loss": 0.4689,
+      "step": 31406
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.504830290970961e-06,
+      "loss": 0.3997,
+      "step": 31407
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5042535398102493e-06,
+      "loss": 0.4103,
+      "step": 31408
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5036767886495374e-06,
+      "loss": 0.3317,
+      "step": 31409
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.503100037488826e-06,
+      "loss": 0.4525,
+      "step": 31410
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.502523286328114e-06,
+      "loss": 0.3425,
+      "step": 31411
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.501946535167402e-06,
+      "loss": 0.4594,
+      "step": 31412
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.5013697840066903e-06,
+      "loss": 0.4072,
+      "step": 31413
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.500793032845979e-06,
+      "loss": 0.3455,
+      "step": 31414
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.500216281685267e-06,
+      "loss": 0.3924,
+      "step": 31415
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4996395305245554e-06,
+      "loss": 0.3578,
+      "step": 31416
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4990627793638435e-06,
+      "loss": 0.4048,
+      "step": 31417
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.498486028203132e-06,
+      "loss": 0.3383,
+      "step": 31418
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.49790927704242e-06,
+      "loss": 0.4395,
+      "step": 31419
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4973325258817087e-06,
+      "loss": 0.4869,
+      "step": 31420
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.496755774720997e-06,
+      "loss": 0.3471,
+      "step": 31421
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.496179023560285e-06,
+      "loss": 0.3082,
+      "step": 31422
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4956022723995735e-06,
+      "loss": 0.3476,
+      "step": 31423
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4950255212388616e-06,
+      "loss": 0.2946,
+      "step": 31424
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.49444877007815e-06,
+      "loss": 0.3278,
+      "step": 31425
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.493872018917438e-06,
+      "loss": 0.3044,
+      "step": 31426
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4932952677567267e-06,
+      "loss": 0.3415,
+      "step": 31427
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.492718516596015e-06,
+      "loss": 0.4256,
+      "step": 31428
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.492141765435303e-06,
+      "loss": 0.4182,
+      "step": 31429
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4915650142745915e-06,
+      "loss": 0.4271,
+      "step": 31430
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4909882631138796e-06,
+      "loss": 0.4497,
+      "step": 31431
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.490411511953168e-06,
+      "loss": 0.426,
+      "step": 31432
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4898347607924562e-06,
+      "loss": 0.4884,
+      "step": 31433
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4892580096317444e-06,
+      "loss": 0.4309,
+      "step": 31434
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.488681258471033e-06,
+      "loss": 0.4051,
+      "step": 31435
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.488104507310321e-06,
+      "loss": 0.4251,
+      "step": 31436
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4875277561496095e-06,
+      "loss": 0.4298,
+      "step": 31437
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.486951004988898e-06,
+      "loss": 0.3592,
+      "step": 31438
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.486374253828186e-06,
+      "loss": 0.3373,
+      "step": 31439
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4857975026674743e-06,
+      "loss": 0.4425,
+      "step": 31440
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4852207515067624e-06,
+      "loss": 0.3587,
+      "step": 31441
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.484644000346051e-06,
+      "loss": 0.3534,
+      "step": 31442
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4840672491853394e-06,
+      "loss": 0.3672,
+      "step": 31443
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4834904980246276e-06,
+      "loss": 0.3614,
+      "step": 31444
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4829137468639157e-06,
+      "loss": 0.348,
+      "step": 31445
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4823369957032038e-06,
+      "loss": 0.2969,
+      "step": 31446
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4817602445424923e-06,
+      "loss": 0.3463,
+      "step": 31447
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.481183493381781e-06,
+      "loss": 0.3316,
+      "step": 31448
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.480606742221069e-06,
+      "loss": 0.4496,
+      "step": 31449
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4800299910603575e-06,
+      "loss": 0.4051,
+      "step": 31450
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4794532398996456e-06,
+      "loss": 0.3981,
+      "step": 31451
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4788764887389337e-06,
+      "loss": 0.2968,
+      "step": 31452
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.478299737578222e-06,
+      "loss": 0.5385,
+      "step": 31453
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4777229864175103e-06,
+      "loss": 0.3812,
+      "step": 31454
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.477146235256799e-06,
+      "loss": 0.2745,
+      "step": 31455
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.476569484096087e-06,
+      "loss": 0.5058,
+      "step": 31456
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.475992732935375e-06,
+      "loss": 0.3832,
+      "step": 31457
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4754159817746636e-06,
+      "loss": 0.4007,
+      "step": 31458
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4748392306139517e-06,
+      "loss": 0.4036,
+      "step": 31459
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4742624794532403e-06,
+      "loss": 0.4265,
+      "step": 31460
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4736857282925284e-06,
+      "loss": 0.3572,
+      "step": 31461
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.473108977131817e-06,
+      "loss": 0.4063,
+      "step": 31462
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.472532225971105e-06,
+      "loss": 0.3584,
+      "step": 31463
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.471955474810393e-06,
+      "loss": 0.3434,
+      "step": 31464
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4713787236496816e-06,
+      "loss": 0.4428,
+      "step": 31465
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4708019724889698e-06,
+      "loss": 0.4125,
+      "step": 31466
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4702252213282583e-06,
+      "loss": 0.3974,
+      "step": 31467
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4696484701675464e-06,
+      "loss": 0.2753,
+      "step": 31468
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.469071719006835e-06,
+      "loss": 0.4369,
+      "step": 31469
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.468494967846123e-06,
+      "loss": 0.4008,
+      "step": 31470
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.467918216685411e-06,
+      "loss": 0.4287,
+      "step": 31471
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4673414655246997e-06,
+      "loss": 0.4775,
+      "step": 31472
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4667647143639878e-06,
+      "loss": 0.374,
+      "step": 31473
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4661879632032763e-06,
+      "loss": 0.5054,
+      "step": 31474
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4656112120425644e-06,
+      "loss": 0.3597,
+      "step": 31475
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4650344608818525e-06,
+      "loss": 0.4077,
+      "step": 31476
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.464457709721141e-06,
+      "loss": 0.47,
+      "step": 31477
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.463880958560429e-06,
+      "loss": 0.325,
+      "step": 31478
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4633042073997177e-06,
+      "loss": 0.4428,
+      "step": 31479
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.462727456239006e-06,
+      "loss": 0.3326,
+      "step": 31480
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4621507050782943e-06,
+      "loss": 0.414,
+      "step": 31481
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4615739539175825e-06,
+      "loss": 0.4014,
+      "step": 31482
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4609972027568706e-06,
+      "loss": 0.3887,
+      "step": 31483
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.460420451596159e-06,
+      "loss": 0.4008,
+      "step": 31484
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.459843700435447e-06,
+      "loss": 0.4252,
+      "step": 31485
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4592669492747357e-06,
+      "loss": 0.3072,
+      "step": 31486
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.458690198114024e-06,
+      "loss": 0.2731,
+      "step": 31487
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.458113446953312e-06,
+      "loss": 0.4012,
+      "step": 31488
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4575366957926005e-06,
+      "loss": 0.4158,
+      "step": 31489
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4569599446318886e-06,
+      "loss": 0.3686,
+      "step": 31490
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.456383193471177e-06,
+      "loss": 0.3777,
+      "step": 31491
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4558064423104652e-06,
+      "loss": 0.3778,
+      "step": 31492
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4552296911497538e-06,
+      "loss": 0.3559,
+      "step": 31493
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.454652939989042e-06,
+      "loss": 0.272,
+      "step": 31494
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.45407618882833e-06,
+      "loss": 0.3896,
+      "step": 31495
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4534994376676185e-06,
+      "loss": 0.3186,
+      "step": 31496
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4529226865069066e-06,
+      "loss": 0.3295,
+      "step": 31497
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.452345935346195e-06,
+      "loss": 0.3867,
+      "step": 31498
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4517691841854837e-06,
+      "loss": 0.4191,
+      "step": 31499
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.451192433024772e-06,
+      "loss": 0.3784,
+      "step": 31500
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.45061568186406e-06,
+      "loss": 0.435,
+      "step": 31501
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.450038930703348e-06,
+      "loss": 0.3969,
+      "step": 31502
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4494621795426366e-06,
+      "loss": 0.3815,
+      "step": 31503
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.448885428381925e-06,
+      "loss": 0.4004,
+      "step": 31504
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.448308677221213e-06,
+      "loss": 0.4208,
+      "step": 31505
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4477319260605013e-06,
+      "loss": 0.4584,
+      "step": 31506
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4471551748997894e-06,
+      "loss": 0.46,
+      "step": 31507
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.446578423739078e-06,
+      "loss": 0.4052,
+      "step": 31508
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.446001672578366e-06,
+      "loss": 0.3755,
+      "step": 31509
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4454249214176546e-06,
+      "loss": 0.3483,
+      "step": 31510
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.444848170256943e-06,
+      "loss": 0.3018,
+      "step": 31511
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4442714190962312e-06,
+      "loss": 0.4209,
+      "step": 31512
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4436946679355193e-06,
+      "loss": 0.3733,
+      "step": 31513
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4431179167748074e-06,
+      "loss": 0.5065,
+      "step": 31514
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.442541165614096e-06,
+      "loss": 0.2925,
+      "step": 31515
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4419644144533845e-06,
+      "loss": 0.436,
+      "step": 31516
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4413876632926726e-06,
+      "loss": 0.3568,
+      "step": 31517
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4408109121319607e-06,
+      "loss": 0.4103,
+      "step": 31518
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4402341609712493e-06,
+      "loss": 0.411,
+      "step": 31519
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4396574098105374e-06,
+      "loss": 0.4793,
+      "step": 31520
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.439080658649826e-06,
+      "loss": 0.3243,
+      "step": 31521
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.438503907489114e-06,
+      "loss": 0.4678,
+      "step": 31522
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4379271563284025e-06,
+      "loss": 0.4078,
+      "step": 31523
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4373504051676906e-06,
+      "loss": 0.4209,
+      "step": 31524
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4367736540069788e-06,
+      "loss": 0.5259,
+      "step": 31525
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4361969028462673e-06,
+      "loss": 0.4882,
+      "step": 31526
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4356201516855554e-06,
+      "loss": 0.4734,
+      "step": 31527
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.435043400524844e-06,
+      "loss": 0.4067,
+      "step": 31528
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.434466649364132e-06,
+      "loss": 0.3894,
+      "step": 31529
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4338898982034206e-06,
+      "loss": 0.4379,
+      "step": 31530
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4333131470427087e-06,
+      "loss": 0.3392,
+      "step": 31531
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4327363958819968e-06,
+      "loss": 0.4165,
+      "step": 31532
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4321596447212853e-06,
+      "loss": 0.3958,
+      "step": 31533
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4315828935605734e-06,
+      "loss": 0.3865,
+      "step": 31534
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.431006142399862e-06,
+      "loss": 0.3286,
+      "step": 31535
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.43042939123915e-06,
+      "loss": 0.3335,
+      "step": 31536
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.429852640078438e-06,
+      "loss": 0.4193,
+      "step": 31537
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4292758889177267e-06,
+      "loss": 0.2956,
+      "step": 31538
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.428699137757015e-06,
+      "loss": 0.353,
+      "step": 31539
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4281223865963033e-06,
+      "loss": 0.4411,
+      "step": 31540
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4275456354355915e-06,
+      "loss": 0.3046,
+      "step": 31541
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.42696888427488e-06,
+      "loss": 0.3488,
+      "step": 31542
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.426392133114168e-06,
+      "loss": 0.5298,
+      "step": 31543
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.425815381953456e-06,
+      "loss": 0.3285,
+      "step": 31544
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4252386307927447e-06,
+      "loss": 0.3438,
+      "step": 31545
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.424661879632033e-06,
+      "loss": 0.2476,
+      "step": 31546
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4240851284713214e-06,
+      "loss": 0.4164,
+      "step": 31547
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4235083773106095e-06,
+      "loss": 0.4451,
+      "step": 31548
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2.4229316261498976e-06,
+      "loss": 0.3547,
+      "step": 31549
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.422354874989186e-06,
+      "loss": 0.4005,
+      "step": 31550
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4217781238284742e-06,
+      "loss": 0.3431,
+      "step": 31551
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4212013726677628e-06,
+      "loss": 0.5039,
+      "step": 31552
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.420624621507051e-06,
+      "loss": 0.371,
+      "step": 31553
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4200478703463394e-06,
+      "loss": 0.5017,
+      "step": 31554
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4194711191856275e-06,
+      "loss": 0.4941,
+      "step": 31555
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4188943680249156e-06,
+      "loss": 0.4347,
+      "step": 31556
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.418317616864204e-06,
+      "loss": 0.3493,
+      "step": 31557
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4177408657034923e-06,
+      "loss": 0.3247,
+      "step": 31558
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.417164114542781e-06,
+      "loss": 0.2958,
+      "step": 31559
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.416587363382069e-06,
+      "loss": 0.3177,
+      "step": 31560
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4160106122213574e-06,
+      "loss": 0.3839,
+      "step": 31561
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4154338610606455e-06,
+      "loss": 0.4364,
+      "step": 31562
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4148571098999337e-06,
+      "loss": 0.3779,
+      "step": 31563
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.414280358739222e-06,
+      "loss": 0.355,
+      "step": 31564
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4137036075785107e-06,
+      "loss": 0.3845,
+      "step": 31565
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.413126856417799e-06,
+      "loss": 0.3289,
+      "step": 31566
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.412550105257087e-06,
+      "loss": 0.3975,
+      "step": 31567
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.411973354096375e-06,
+      "loss": 0.3762,
+      "step": 31568
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4113966029356636e-06,
+      "loss": 0.4168,
+      "step": 31569
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4108198517749517e-06,
+      "loss": 0.4811,
+      "step": 31570
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4102431006142402e-06,
+      "loss": 0.2698,
+      "step": 31571
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4096663494535288e-06,
+      "loss": 0.3099,
+      "step": 31572
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.409089598292817e-06,
+      "loss": 0.4953,
+      "step": 31573
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.408512847132105e-06,
+      "loss": 0.4417,
+      "step": 31574
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.407936095971393e-06,
+      "loss": 0.402,
+      "step": 31575
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4073593448106816e-06,
+      "loss": 0.4407,
+      "step": 31576
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.40678259364997e-06,
+      "loss": 0.3271,
+      "step": 31577
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4062058424892583e-06,
+      "loss": 0.394,
+      "step": 31578
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4056290913285464e-06,
+      "loss": 0.3529,
+      "step": 31579
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4050523401678345e-06,
+      "loss": 0.3716,
+      "step": 31580
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.404475589007123e-06,
+      "loss": 0.4251,
+      "step": 31581
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4038988378464115e-06,
+      "loss": 0.4511,
+      "step": 31582
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4033220866856996e-06,
+      "loss": 0.4389,
+      "step": 31583
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.402745335524988e-06,
+      "loss": 0.5226,
+      "step": 31584
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4021685843642763e-06,
+      "loss": 0.3873,
+      "step": 31585
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.4015918332035644e-06,
+      "loss": 0.3805,
+      "step": 31586
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.401015082042853e-06,
+      "loss": 0.3375,
+      "step": 31587
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.400438330882141e-06,
+      "loss": 0.3841,
+      "step": 31588
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3998615797214296e-06,
+      "loss": 0.3407,
+      "step": 31589
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3992848285607177e-06,
+      "loss": 0.4156,
+      "step": 31590
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.398708077400006e-06,
+      "loss": 0.385,
+      "step": 31591
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3981313262392943e-06,
+      "loss": 0.444,
+      "step": 31592
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3975545750785824e-06,
+      "loss": 0.3823,
+      "step": 31593
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.396977823917871e-06,
+      "loss": 0.4265,
+      "step": 31594
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.396401072757159e-06,
+      "loss": 0.3888,
+      "step": 31595
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3958243215964476e-06,
+      "loss": 0.2979,
+      "step": 31596
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3952475704357357e-06,
+      "loss": 0.4256,
+      "step": 31597
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.394670819275024e-06,
+      "loss": 0.3768,
+      "step": 31598
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3940940681143123e-06,
+      "loss": 0.4098,
+      "step": 31599
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3935173169536005e-06,
+      "loss": 0.3708,
+      "step": 31600
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.392940565792889e-06,
+      "loss": 0.3634,
+      "step": 31601
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.392363814632177e-06,
+      "loss": 0.3991,
+      "step": 31602
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3917870634714656e-06,
+      "loss": 0.36,
+      "step": 31603
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3912103123107537e-06,
+      "loss": 0.3573,
+      "step": 31604
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.390633561150042e-06,
+      "loss": 0.3758,
+      "step": 31605
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3900568099893304e-06,
+      "loss": 0.5065,
+      "step": 31606
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3894800588286185e-06,
+      "loss": 0.4835,
+      "step": 31607
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.388903307667907e-06,
+      "loss": 0.3609,
+      "step": 31608
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.388326556507195e-06,
+      "loss": 0.4348,
+      "step": 31609
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3877498053464832e-06,
+      "loss": 0.4266,
+      "step": 31610
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3871730541857718e-06,
+      "loss": 0.3498,
+      "step": 31611
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.38659630302506e-06,
+      "loss": 0.2527,
+      "step": 31612
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3860195518643484e-06,
+      "loss": 0.4666,
+      "step": 31613
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3854428007036365e-06,
+      "loss": 0.3472,
+      "step": 31614
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.384866049542925e-06,
+      "loss": 0.4296,
+      "step": 31615
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.384289298382213e-06,
+      "loss": 0.513,
+      "step": 31616
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3837125472215013e-06,
+      "loss": 0.4363,
+      "step": 31617
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.38313579606079e-06,
+      "loss": 0.4007,
+      "step": 31618
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.382559044900078e-06,
+      "loss": 0.4488,
+      "step": 31619
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3819822937393664e-06,
+      "loss": 0.3727,
+      "step": 31620
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3814055425786545e-06,
+      "loss": 0.4754,
+      "step": 31621
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.380828791417943e-06,
+      "loss": 0.373,
+      "step": 31622
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.380252040257231e-06,
+      "loss": 0.4608,
+      "step": 31623
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3796752890965193e-06,
+      "loss": 0.4117,
+      "step": 31624
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.379098537935808e-06,
+      "loss": 0.2489,
+      "step": 31625
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3785217867750964e-06,
+      "loss": 0.5317,
+      "step": 31626
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3779450356143845e-06,
+      "loss": 0.4406,
+      "step": 31627
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3773682844536726e-06,
+      "loss": 0.3147,
+      "step": 31628
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3767915332929607e-06,
+      "loss": 0.3217,
+      "step": 31629
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3762147821322492e-06,
+      "loss": 0.3635,
+      "step": 31630
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3756380309715373e-06,
+      "loss": 0.3703,
+      "step": 31631
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.375061279810826e-06,
+      "loss": 0.3292,
+      "step": 31632
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3744845286501144e-06,
+      "loss": 0.3004,
+      "step": 31633
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3739077774894025e-06,
+      "loss": 0.3178,
+      "step": 31634
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3733310263286906e-06,
+      "loss": 0.4489,
+      "step": 31635
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3727542751679787e-06,
+      "loss": 0.4539,
+      "step": 31636
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3721775240072672e-06,
+      "loss": 0.4507,
+      "step": 31637
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3716007728465558e-06,
+      "loss": 0.4185,
+      "step": 31638
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.371024021685844e-06,
+      "loss": 0.4216,
+      "step": 31639
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.370447270525132e-06,
+      "loss": 0.4061,
+      "step": 31640
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.36987051936442e-06,
+      "loss": 0.5634,
+      "step": 31641
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3692937682037086e-06,
+      "loss": 0.3494,
+      "step": 31642
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.368717017042997e-06,
+      "loss": 0.3975,
+      "step": 31643
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3681402658822853e-06,
+      "loss": 0.514,
+      "step": 31644
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.367563514721574e-06,
+      "loss": 0.3645,
+      "step": 31645
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.366986763560862e-06,
+      "loss": 0.3119,
+      "step": 31646
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.36641001240015e-06,
+      "loss": 0.3389,
+      "step": 31647
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3658332612394386e-06,
+      "loss": 0.3337,
+      "step": 31648
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3652565100787267e-06,
+      "loss": 0.3744,
+      "step": 31649
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.364679758918015e-06,
+      "loss": 0.2975,
+      "step": 31650
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3641030077573033e-06,
+      "loss": 0.341,
+      "step": 31651
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3635262565965914e-06,
+      "loss": 0.407,
+      "step": 31652
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.36294950543588e-06,
+      "loss": 0.3269,
+      "step": 31653
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.362372754275168e-06,
+      "loss": 0.358,
+      "step": 31654
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3617960031144566e-06,
+      "loss": 0.3496,
+      "step": 31655
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3612192519537447e-06,
+      "loss": 0.439,
+      "step": 31656
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3606425007930332e-06,
+      "loss": 0.4028,
+      "step": 31657
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3600657496323213e-06,
+      "loss": 0.3668,
+      "step": 31658
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3594889984716094e-06,
+      "loss": 0.354,
+      "step": 31659
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.358912247310898e-06,
+      "loss": 0.3932,
+      "step": 31660
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.358335496150186e-06,
+      "loss": 0.3686,
+      "step": 31661
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3577587449894746e-06,
+      "loss": 0.3204,
+      "step": 31662
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3571819938287627e-06,
+      "loss": 0.3983,
+      "step": 31663
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3566052426680513e-06,
+      "loss": 0.327,
+      "step": 31664
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3560284915073394e-06,
+      "loss": 0.3544,
+      "step": 31665
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3554517403466275e-06,
+      "loss": 0.3947,
+      "step": 31666
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.354874989185916e-06,
+      "loss": 0.3626,
+      "step": 31667
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.354298238025204e-06,
+      "loss": 0.3067,
+      "step": 31668
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3537214868644927e-06,
+      "loss": 0.4368,
+      "step": 31669
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3531447357037808e-06,
+      "loss": 0.4267,
+      "step": 31670
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.352567984543069e-06,
+      "loss": 0.3495,
+      "step": 31671
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3519912333823574e-06,
+      "loss": 0.3821,
+      "step": 31672
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3514144822216455e-06,
+      "loss": 0.3248,
+      "step": 31673
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.350837731060934e-06,
+      "loss": 0.3864,
+      "step": 31674
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.350260979900222e-06,
+      "loss": 0.4123,
+      "step": 31675
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3496842287395107e-06,
+      "loss": 0.3906,
+      "step": 31676
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.349107477578799e-06,
+      "loss": 0.5532,
+      "step": 31677
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.348530726418087e-06,
+      "loss": 0.4785,
+      "step": 31678
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3479539752573754e-06,
+      "loss": 0.3893,
+      "step": 31679
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3473772240966635e-06,
+      "loss": 0.4834,
+      "step": 31680
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.346800472935952e-06,
+      "loss": 0.4255,
+      "step": 31681
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.34622372177524e-06,
+      "loss": 0.3647,
+      "step": 31682
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3456469706145287e-06,
+      "loss": 0.3166,
+      "step": 31683
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.345070219453817e-06,
+      "loss": 0.4416,
+      "step": 31684
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.344493468293105e-06,
+      "loss": 0.4147,
+      "step": 31685
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3439167171323935e-06,
+      "loss": 0.4902,
+      "step": 31686
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.343339965971682e-06,
+      "loss": 0.3151,
+      "step": 31687
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.34276321481097e-06,
+      "loss": 0.4798,
+      "step": 31688
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3421864636502582e-06,
+      "loss": 0.4058,
+      "step": 31689
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3416097124895463e-06,
+      "loss": 0.4396,
+      "step": 31690
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.341032961328835e-06,
+      "loss": 0.3278,
+      "step": 31691
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.340456210168123e-06,
+      "loss": 0.3298,
+      "step": 31692
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3398794590074115e-06,
+      "loss": 0.3864,
+      "step": 31693
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3393027078467e-06,
+      "loss": 0.3129,
+      "step": 31694
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.338725956685988e-06,
+      "loss": 0.4729,
+      "step": 31695
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3381492055252762e-06,
+      "loss": 0.3868,
+      "step": 31696
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3375724543645644e-06,
+      "loss": 0.4177,
+      "step": 31697
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.336995703203853e-06,
+      "loss": 0.2965,
+      "step": 31698
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3364189520431414e-06,
+      "loss": 0.3217,
+      "step": 31699
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3358422008824295e-06,
+      "loss": 0.4192,
+      "step": 31700
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3352654497217176e-06,
+      "loss": 0.4453,
+      "step": 31701
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3346886985610057e-06,
+      "loss": 0.4503,
+      "step": 31702
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3341119474002943e-06,
+      "loss": 0.3455,
+      "step": 31703
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.333535196239583e-06,
+      "loss": 0.3757,
+      "step": 31704
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.332958445078871e-06,
+      "loss": 0.3622,
+      "step": 31705
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3323816939181594e-06,
+      "loss": 0.2895,
+      "step": 31706
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3318049427574476e-06,
+      "loss": 0.4093,
+      "step": 31707
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3312281915967357e-06,
+      "loss": 0.4031,
+      "step": 31708
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.330651440436024e-06,
+      "loss": 0.3255,
+      "step": 31709
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3300746892753123e-06,
+      "loss": 0.5029,
+      "step": 31710
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.329497938114601e-06,
+      "loss": 0.3719,
+      "step": 31711
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.328921186953889e-06,
+      "loss": 0.4521,
+      "step": 31712
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.328344435793177e-06,
+      "loss": 0.4169,
+      "step": 31713
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3277676846324656e-06,
+      "loss": 0.4813,
+      "step": 31714
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3271909334717537e-06,
+      "loss": 0.4583,
+      "step": 31715
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3266141823110422e-06,
+      "loss": 0.3647,
+      "step": 31716
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3260374311503303e-06,
+      "loss": 0.4492,
+      "step": 31717
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.325460679989619e-06,
+      "loss": 0.4107,
+      "step": 31718
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.324883928828907e-06,
+      "loss": 0.4557,
+      "step": 31719
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.324307177668195e-06,
+      "loss": 0.4042,
+      "step": 31720
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3237304265074836e-06,
+      "loss": 0.3927,
+      "step": 31721
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3231536753467717e-06,
+      "loss": 0.357,
+      "step": 31722
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3225769241860603e-06,
+      "loss": 0.3809,
+      "step": 31723
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3220001730253484e-06,
+      "loss": 0.4001,
+      "step": 31724
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.321423421864637e-06,
+      "loss": 0.3844,
+      "step": 31725
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.320846670703925e-06,
+      "loss": 0.3399,
+      "step": 31726
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.320269919543213e-06,
+      "loss": 0.4015,
+      "step": 31727
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2.3196931683825016e-06,
+      "loss": 0.358,
+      "step": 31728
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3191164172217898e-06,
+      "loss": 0.4526,
+      "step": 31729
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3185396660610783e-06,
+      "loss": 0.2776,
+      "step": 31730
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3179629149003664e-06,
+      "loss": 0.4119,
+      "step": 31731
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3173861637396545e-06,
+      "loss": 0.4497,
+      "step": 31732
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.316809412578943e-06,
+      "loss": 0.5238,
+      "step": 31733
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.316232661418231e-06,
+      "loss": 0.3624,
+      "step": 31734
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3156559102575197e-06,
+      "loss": 0.4119,
+      "step": 31735
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3150791590968078e-06,
+      "loss": 0.5301,
+      "step": 31736
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3145024079360963e-06,
+      "loss": 0.4273,
+      "step": 31737
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3139256567753844e-06,
+      "loss": 0.3762,
+      "step": 31738
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3133489056146725e-06,
+      "loss": 0.5463,
+      "step": 31739
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.312772154453961e-06,
+      "loss": 0.4327,
+      "step": 31740
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.312195403293249e-06,
+      "loss": 0.3557,
+      "step": 31741
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3116186521325377e-06,
+      "loss": 0.4057,
+      "step": 31742
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.311041900971826e-06,
+      "loss": 0.5728,
+      "step": 31743
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.310465149811114e-06,
+      "loss": 0.3064,
+      "step": 31744
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3098883986504025e-06,
+      "loss": 0.4922,
+      "step": 31745
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3093116474896906e-06,
+      "loss": 0.4422,
+      "step": 31746
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.308734896328979e-06,
+      "loss": 0.3701,
+      "step": 31747
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3081581451682676e-06,
+      "loss": 0.4075,
+      "step": 31748
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3075813940075557e-06,
+      "loss": 0.363,
+      "step": 31749
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.307004642846844e-06,
+      "loss": 0.3802,
+      "step": 31750
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.306427891686132e-06,
+      "loss": 0.3781,
+      "step": 31751
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3058511405254205e-06,
+      "loss": 0.5342,
+      "step": 31752
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3052743893647086e-06,
+      "loss": 0.4451,
+      "step": 31753
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.304697638203997e-06,
+      "loss": 0.3885,
+      "step": 31754
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3041208870432857e-06,
+      "loss": 0.4154,
+      "step": 31755
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3035441358825738e-06,
+      "loss": 0.392,
+      "step": 31756
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.302967384721862e-06,
+      "loss": 0.3207,
+      "step": 31757
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.30239063356115e-06,
+      "loss": 0.3155,
+      "step": 31758
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3018138824004385e-06,
+      "loss": 0.33,
+      "step": 31759
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.301237131239727e-06,
+      "loss": 0.3622,
+      "step": 31760
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.300660380079015e-06,
+      "loss": 0.4255,
+      "step": 31761
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.3000836289183033e-06,
+      "loss": 0.441,
+      "step": 31762
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2995068777575914e-06,
+      "loss": 0.3599,
+      "step": 31763
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.29893012659688e-06,
+      "loss": 0.2344,
+      "step": 31764
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2983533754361684e-06,
+      "loss": 0.3954,
+      "step": 31765
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2977766242754566e-06,
+      "loss": 0.3623,
+      "step": 31766
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.297199873114745e-06,
+      "loss": 0.2931,
+      "step": 31767
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.296623121954033e-06,
+      "loss": 0.5235,
+      "step": 31768
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2960463707933213e-06,
+      "loss": 0.3401,
+      "step": 31769
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.29546961963261e-06,
+      "loss": 0.4831,
+      "step": 31770
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.294892868471898e-06,
+      "loss": 0.5127,
+      "step": 31771
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2943161173111865e-06,
+      "loss": 0.3631,
+      "step": 31772
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2937393661504746e-06,
+      "loss": 0.456,
+      "step": 31773
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2931626149897627e-06,
+      "loss": 0.4278,
+      "step": 31774
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2925858638290512e-06,
+      "loss": 0.3776,
+      "step": 31775
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2920091126683393e-06,
+      "loss": 0.3588,
+      "step": 31776
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.291432361507628e-06,
+      "loss": 0.4205,
+      "step": 31777
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.290855610346916e-06,
+      "loss": 0.3987,
+      "step": 31778
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2902788591862045e-06,
+      "loss": 0.5026,
+      "step": 31779
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2897021080254926e-06,
+      "loss": 0.3235,
+      "step": 31780
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2891253568647807e-06,
+      "loss": 0.3584,
+      "step": 31781
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2885486057040693e-06,
+      "loss": 0.408,
+      "step": 31782
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2879718545433574e-06,
+      "loss": 0.4371,
+      "step": 31783
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.287395103382646e-06,
+      "loss": 0.4021,
+      "step": 31784
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.286818352221934e-06,
+      "loss": 0.3412,
+      "step": 31785
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2862416010612225e-06,
+      "loss": 0.4299,
+      "step": 31786
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2856648499005106e-06,
+      "loss": 0.3003,
+      "step": 31787
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2850880987397988e-06,
+      "loss": 0.3667,
+      "step": 31788
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2845113475790873e-06,
+      "loss": 0.3816,
+      "step": 31789
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2839345964183754e-06,
+      "loss": 0.463,
+      "step": 31790
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.283357845257664e-06,
+      "loss": 0.4421,
+      "step": 31791
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.282781094096952e-06,
+      "loss": 0.2872,
+      "step": 31792
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.28220434293624e-06,
+      "loss": 0.4523,
+      "step": 31793
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2816275917755287e-06,
+      "loss": 0.3992,
+      "step": 31794
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2810508406148168e-06,
+      "loss": 0.4545,
+      "step": 31795
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2804740894541053e-06,
+      "loss": 0.3249,
+      "step": 31796
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2798973382933934e-06,
+      "loss": 0.4153,
+      "step": 31797
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.279320587132682e-06,
+      "loss": 0.4339,
+      "step": 31798
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.27874383597197e-06,
+      "loss": 0.394,
+      "step": 31799
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.278167084811258e-06,
+      "loss": 0.354,
+      "step": 31800
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2775903336505467e-06,
+      "loss": 0.3775,
+      "step": 31801
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.277013582489835e-06,
+      "loss": 0.464,
+      "step": 31802
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2764368313291233e-06,
+      "loss": 0.2851,
+      "step": 31803
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2758600801684115e-06,
+      "loss": 0.4281,
+      "step": 31804
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2752833290076996e-06,
+      "loss": 0.273,
+      "step": 31805
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.274706577846988e-06,
+      "loss": 0.4792,
+      "step": 31806
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.274129826686276e-06,
+      "loss": 0.4751,
+      "step": 31807
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2735530755255647e-06,
+      "loss": 0.4159,
+      "step": 31808
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2729763243648533e-06,
+      "loss": 0.4084,
+      "step": 31809
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2723995732041414e-06,
+      "loss": 0.3811,
+      "step": 31810
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2718228220434295e-06,
+      "loss": 0.2477,
+      "step": 31811
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2712460708827176e-06,
+      "loss": 0.472,
+      "step": 31812
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.270669319722006e-06,
+      "loss": 0.395,
+      "step": 31813
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2700925685612942e-06,
+      "loss": 0.4218,
+      "step": 31814
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2695158174005828e-06,
+      "loss": 0.398,
+      "step": 31815
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2689390662398713e-06,
+      "loss": 0.442,
+      "step": 31816
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2683623150791594e-06,
+      "loss": 0.3404,
+      "step": 31817
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2677855639184475e-06,
+      "loss": 0.5147,
+      "step": 31818
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2672088127577356e-06,
+      "loss": 0.3188,
+      "step": 31819
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.266632061597024e-06,
+      "loss": 0.4494,
+      "step": 31820
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2660553104363127e-06,
+      "loss": 0.3829,
+      "step": 31821
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.265478559275601e-06,
+      "loss": 0.3718,
+      "step": 31822
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.264901808114889e-06,
+      "loss": 0.3563,
+      "step": 31823
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.264325056954177e-06,
+      "loss": 0.4012,
+      "step": 31824
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2637483057934655e-06,
+      "loss": 0.4349,
+      "step": 31825
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.263171554632754e-06,
+      "loss": 0.2787,
+      "step": 31826
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.262594803472042e-06,
+      "loss": 0.3677,
+      "step": 31827
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2620180523113307e-06,
+      "loss": 0.4066,
+      "step": 31828
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.261441301150619e-06,
+      "loss": 0.3692,
+      "step": 31829
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.260864549989907e-06,
+      "loss": 0.3408,
+      "step": 31830
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.260287798829195e-06,
+      "loss": 0.3758,
+      "step": 31831
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2597110476684836e-06,
+      "loss": 0.3393,
+      "step": 31832
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.259134296507772e-06,
+      "loss": 0.3386,
+      "step": 31833
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2585575453470602e-06,
+      "loss": 0.4123,
+      "step": 31834
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2579807941863483e-06,
+      "loss": 0.4008,
+      "step": 31835
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2574040430256364e-06,
+      "loss": 0.2908,
+      "step": 31836
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.256827291864925e-06,
+      "loss": 0.3019,
+      "step": 31837
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2562505407042135e-06,
+      "loss": 0.3099,
+      "step": 31838
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2556737895435016e-06,
+      "loss": 0.4109,
+      "step": 31839
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.25509703838279e-06,
+      "loss": 0.3415,
+      "step": 31840
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2545202872220783e-06,
+      "loss": 0.502,
+      "step": 31841
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2539435360613664e-06,
+      "loss": 0.4802,
+      "step": 31842
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.253366784900655e-06,
+      "loss": 0.4021,
+      "step": 31843
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.252790033739943e-06,
+      "loss": 0.3498,
+      "step": 31844
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2522132825792315e-06,
+      "loss": 0.3253,
+      "step": 31845
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2516365314185196e-06,
+      "loss": 0.4337,
+      "step": 31846
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.251059780257808e-06,
+      "loss": 0.3496,
+      "step": 31847
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2504830290970963e-06,
+      "loss": 0.3996,
+      "step": 31848
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2499062779363844e-06,
+      "loss": 0.4038,
+      "step": 31849
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.249329526775673e-06,
+      "loss": 0.4203,
+      "step": 31850
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.248752775614961e-06,
+      "loss": 0.3392,
+      "step": 31851
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2481760244542496e-06,
+      "loss": 0.4439,
+      "step": 31852
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2475992732935377e-06,
+      "loss": 0.2804,
+      "step": 31853
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2470225221328258e-06,
+      "loss": 0.4177,
+      "step": 31854
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2464457709721143e-06,
+      "loss": 0.4742,
+      "step": 31855
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2458690198114024e-06,
+      "loss": 0.411,
+      "step": 31856
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.245292268650691e-06,
+      "loss": 0.2962,
+      "step": 31857
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.244715517489979e-06,
+      "loss": 0.5043,
+      "step": 31858
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2441387663292676e-06,
+      "loss": 0.3903,
+      "step": 31859
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2435620151685557e-06,
+      "loss": 0.3878,
+      "step": 31860
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.242985264007844e-06,
+      "loss": 0.4014,
+      "step": 31861
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2424085128471323e-06,
+      "loss": 0.4833,
+      "step": 31862
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2418317616864205e-06,
+      "loss": 0.4369,
+      "step": 31863
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.241255010525709e-06,
+      "loss": 0.4989,
+      "step": 31864
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.240678259364997e-06,
+      "loss": 0.4929,
+      "step": 31865
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.240101508204285e-06,
+      "loss": 0.405,
+      "step": 31866
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2395247570435737e-06,
+      "loss": 0.3562,
+      "step": 31867
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.238948005882862e-06,
+      "loss": 0.5729,
+      "step": 31868
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2383712547221504e-06,
+      "loss": 0.4576,
+      "step": 31869
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.237794503561439e-06,
+      "loss": 0.3991,
+      "step": 31870
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.237217752400727e-06,
+      "loss": 0.3274,
+      "step": 31871
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.236641001240015e-06,
+      "loss": 0.4599,
+      "step": 31872
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2360642500793032e-06,
+      "loss": 0.4854,
+      "step": 31873
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2354874989185918e-06,
+      "loss": 0.3849,
+      "step": 31874
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.23491074775788e-06,
+      "loss": 0.3164,
+      "step": 31875
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2343339965971684e-06,
+      "loss": 0.3494,
+      "step": 31876
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2337572454364565e-06,
+      "loss": 0.3741,
+      "step": 31877
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.233180494275745e-06,
+      "loss": 0.4139,
+      "step": 31878
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.232603743115033e-06,
+      "loss": 0.3772,
+      "step": 31879
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2320269919543213e-06,
+      "loss": 0.3485,
+      "step": 31880
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.23145024079361e-06,
+      "loss": 0.2825,
+      "step": 31881
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2308734896328983e-06,
+      "loss": 0.3852,
+      "step": 31882
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2302967384721864e-06,
+      "loss": 0.4871,
+      "step": 31883
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2297199873114745e-06,
+      "loss": 0.3524,
+      "step": 31884
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2291432361507627e-06,
+      "loss": 0.4234,
+      "step": 31885
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.228566484990051e-06,
+      "loss": 0.2979,
+      "step": 31886
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2279897338293397e-06,
+      "loss": 0.5577,
+      "step": 31887
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.227412982668628e-06,
+      "loss": 0.4347,
+      "step": 31888
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2268362315079164e-06,
+      "loss": 0.383,
+      "step": 31889
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2262594803472045e-06,
+      "loss": 0.5247,
+      "step": 31890
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2256827291864926e-06,
+      "loss": 0.3591,
+      "step": 31891
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2251059780257807e-06,
+      "loss": 0.26,
+      "step": 31892
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2245292268650692e-06,
+      "loss": 0.3052,
+      "step": 31893
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2239524757043577e-06,
+      "loss": 0.4447,
+      "step": 31894
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.223375724543646e-06,
+      "loss": 0.315,
+      "step": 31895
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.222798973382934e-06,
+      "loss": 0.3081,
+      "step": 31896
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.222222222222222e-06,
+      "loss": 0.5086,
+      "step": 31897
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2216454710615106e-06,
+      "loss": 0.4021,
+      "step": 31898
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.221068719900799e-06,
+      "loss": 0.3825,
+      "step": 31899
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2204919687400872e-06,
+      "loss": 0.4279,
+      "step": 31900
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2199152175793758e-06,
+      "loss": 0.4287,
+      "step": 31901
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.219338466418664e-06,
+      "loss": 0.5216,
+      "step": 31902
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.218761715257952e-06,
+      "loss": 0.3588,
+      "step": 31903
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2181849640972405e-06,
+      "loss": 0.4523,
+      "step": 31904
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.2176082129365286e-06,
+      "loss": 0.4559,
+      "step": 31905
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2.217031461775817e-06,
+      "loss": 0.3804,
+      "step": 31906
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2164547106151053e-06,
+      "loss": 0.3991,
+      "step": 31907
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.215877959454394e-06,
+      "loss": 0.4547,
+      "step": 31908
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.215301208293682e-06,
+      "loss": 0.3689,
+      "step": 31909
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.21472445713297e-06,
+      "loss": 0.4397,
+      "step": 31910
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2141477059722586e-06,
+      "loss": 0.4755,
+      "step": 31911
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2135709548115467e-06,
+      "loss": 0.381,
+      "step": 31912
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.212994203650835e-06,
+      "loss": 0.3818,
+      "step": 31913
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2124174524901233e-06,
+      "loss": 0.4997,
+      "step": 31914
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2118407013294114e-06,
+      "loss": 0.3205,
+      "step": 31915
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2112639501687e-06,
+      "loss": 0.33,
+      "step": 31916
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.210687199007988e-06,
+      "loss": 0.5179,
+      "step": 31917
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2101104478472766e-06,
+      "loss": 0.3966,
+      "step": 31918
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2095336966865647e-06,
+      "loss": 0.359,
+      "step": 31919
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2089569455258532e-06,
+      "loss": 0.4325,
+      "step": 31920
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2083801943651413e-06,
+      "loss": 0.4203,
+      "step": 31921
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2078034432044294e-06,
+      "loss": 0.3477,
+      "step": 31922
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.207226692043718e-06,
+      "loss": 0.3833,
+      "step": 31923
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.206649940883006e-06,
+      "loss": 0.4261,
+      "step": 31924
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2060731897222946e-06,
+      "loss": 0.2629,
+      "step": 31925
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2054964385615827e-06,
+      "loss": 0.3558,
+      "step": 31926
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.204919687400871e-06,
+      "loss": 0.3,
+      "step": 31927
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2043429362401594e-06,
+      "loss": 0.4595,
+      "step": 31928
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2037661850794475e-06,
+      "loss": 0.3297,
+      "step": 31929
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.203189433918736e-06,
+      "loss": 0.4426,
+      "step": 31930
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2026126827580245e-06,
+      "loss": 0.3813,
+      "step": 31931
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2020359315973127e-06,
+      "loss": 0.4539,
+      "step": 31932
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2014591804366008e-06,
+      "loss": 0.3953,
+      "step": 31933
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.200882429275889e-06,
+      "loss": 0.5263,
+      "step": 31934
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.2003056781151774e-06,
+      "loss": 0.4395,
+      "step": 31935
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1997289269544655e-06,
+      "loss": 0.4581,
+      "step": 31936
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.199152175793754e-06,
+      "loss": 0.3964,
+      "step": 31937
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.198575424633042e-06,
+      "loss": 0.4314,
+      "step": 31938
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1979986734723307e-06,
+      "loss": 0.402,
+      "step": 31939
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.197421922311619e-06,
+      "loss": 0.3456,
+      "step": 31940
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.196845171150907e-06,
+      "loss": 0.4328,
+      "step": 31941
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1962684199901954e-06,
+      "loss": 0.3972,
+      "step": 31942
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.195691668829484e-06,
+      "loss": 0.5811,
+      "step": 31943
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.195114917668772e-06,
+      "loss": 0.3994,
+      "step": 31944
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.19453816650806e-06,
+      "loss": 0.3717,
+      "step": 31945
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1939614153473483e-06,
+      "loss": 0.4192,
+      "step": 31946
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.193384664186637e-06,
+      "loss": 0.3685,
+      "step": 31947
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1928079130259254e-06,
+      "loss": 0.3244,
+      "step": 31948
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1922311618652135e-06,
+      "loss": 0.3771,
+      "step": 31949
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.191654410704502e-06,
+      "loss": 0.3714,
+      "step": 31950
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.19107765954379e-06,
+      "loss": 0.3461,
+      "step": 31951
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1905009083830782e-06,
+      "loss": 0.4155,
+      "step": 31952
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1899241572223663e-06,
+      "loss": 0.5638,
+      "step": 31953
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.189347406061655e-06,
+      "loss": 0.4126,
+      "step": 31954
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1887706549009434e-06,
+      "loss": 0.3496,
+      "step": 31955
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1881939037402315e-06,
+      "loss": 0.4255,
+      "step": 31956
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1876171525795196e-06,
+      "loss": 0.2746,
+      "step": 31957
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1870404014188077e-06,
+      "loss": 0.4588,
+      "step": 31958
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1864636502580962e-06,
+      "loss": 0.283,
+      "step": 31959
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1858868990973848e-06,
+      "loss": 0.4777,
+      "step": 31960
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.185310147936673e-06,
+      "loss": 0.325,
+      "step": 31961
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1847333967759614e-06,
+      "loss": 0.2964,
+      "step": 31962
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1841566456152495e-06,
+      "loss": 0.3302,
+      "step": 31963
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1835798944545376e-06,
+      "loss": 0.4135,
+      "step": 31964
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.183003143293826e-06,
+      "loss": 0.3874,
+      "step": 31965
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1824263921331143e-06,
+      "loss": 0.5648,
+      "step": 31966
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.181849640972403e-06,
+      "loss": 0.4088,
+      "step": 31967
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.181272889811691e-06,
+      "loss": 0.3304,
+      "step": 31968
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.180696138650979e-06,
+      "loss": 0.2498,
+      "step": 31969
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1801193874902676e-06,
+      "loss": 0.3245,
+      "step": 31970
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1795426363295557e-06,
+      "loss": 0.3917,
+      "step": 31971
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.178965885168844e-06,
+      "loss": 0.3796,
+      "step": 31972
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1783891340081323e-06,
+      "loss": 0.4048,
+      "step": 31973
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.177812382847421e-06,
+      "loss": 0.3752,
+      "step": 31974
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.177235631686709e-06,
+      "loss": 0.5228,
+      "step": 31975
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.176658880525997e-06,
+      "loss": 0.4661,
+      "step": 31976
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1760821293652856e-06,
+      "loss": 0.3816,
+      "step": 31977
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1755053782045737e-06,
+      "loss": 0.4058,
+      "step": 31978
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1749286270438622e-06,
+      "loss": 0.3461,
+      "step": 31979
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1743518758831503e-06,
+      "loss": 0.3999,
+      "step": 31980
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.173775124722439e-06,
+      "loss": 0.4438,
+      "step": 31981
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.173198373561727e-06,
+      "loss": 0.4995,
+      "step": 31982
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.172621622401015e-06,
+      "loss": 0.4034,
+      "step": 31983
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1720448712403036e-06,
+      "loss": 0.4595,
+      "step": 31984
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1714681200795917e-06,
+      "loss": 0.3392,
+      "step": 31985
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1708913689188803e-06,
+      "loss": 0.3166,
+      "step": 31986
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1703146177581684e-06,
+      "loss": 0.2823,
+      "step": 31987
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1697378665974565e-06,
+      "loss": 0.352,
+      "step": 31988
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.169161115436745e-06,
+      "loss": 0.2885,
+      "step": 31989
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.168584364276033e-06,
+      "loss": 0.3544,
+      "step": 31990
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1680076131153216e-06,
+      "loss": 0.3627,
+      "step": 31991
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.16743086195461e-06,
+      "loss": 0.3014,
+      "step": 31992
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1668541107938983e-06,
+      "loss": 0.3707,
+      "step": 31993
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1662773596331864e-06,
+      "loss": 0.378,
+      "step": 31994
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1657006084724745e-06,
+      "loss": 0.4332,
+      "step": 31995
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.165123857311763e-06,
+      "loss": 0.3433,
+      "step": 31996
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.164547106151051e-06,
+      "loss": 0.3667,
+      "step": 31997
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1639703549903397e-06,
+      "loss": 0.4167,
+      "step": 31998
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.163393603829628e-06,
+      "loss": 0.4946,
+      "step": 31999
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1628168526689163e-06,
+      "loss": 0.3735,
+      "step": 32000
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1622401015082044e-06,
+      "loss": 0.3934,
+      "step": 32001
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1616633503474925e-06,
+      "loss": 0.4321,
+      "step": 32002
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.161086599186781e-06,
+      "loss": 0.4446,
+      "step": 32003
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1605098480260696e-06,
+      "loss": 0.4237,
+      "step": 32004
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1599330968653577e-06,
+      "loss": 0.2925,
+      "step": 32005
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.159356345704646e-06,
+      "loss": 0.2834,
+      "step": 32006
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.158779594543934e-06,
+      "loss": 0.3078,
+      "step": 32007
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1582028433832225e-06,
+      "loss": 0.4765,
+      "step": 32008
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.157626092222511e-06,
+      "loss": 0.4377,
+      "step": 32009
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.157049341061799e-06,
+      "loss": 0.3902,
+      "step": 32010
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1564725899010876e-06,
+      "loss": 0.409,
+      "step": 32011
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1558958387403757e-06,
+      "loss": 0.44,
+      "step": 32012
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.155319087579664e-06,
+      "loss": 0.3834,
+      "step": 32013
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.154742336418952e-06,
+      "loss": 0.3926,
+      "step": 32014
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1541655852582405e-06,
+      "loss": 0.4249,
+      "step": 32015
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.153588834097529e-06,
+      "loss": 0.3097,
+      "step": 32016
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.153012082936817e-06,
+      "loss": 0.3856,
+      "step": 32017
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1524353317761052e-06,
+      "loss": 0.4136,
+      "step": 32018
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1518585806153933e-06,
+      "loss": 0.4816,
+      "step": 32019
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.151281829454682e-06,
+      "loss": 0.3824,
+      "step": 32020
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1507050782939704e-06,
+      "loss": 0.3416,
+      "step": 32021
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1501283271332585e-06,
+      "loss": 0.3813,
+      "step": 32022
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.149551575972547e-06,
+      "loss": 0.3737,
+      "step": 32023
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.148974824811835e-06,
+      "loss": 0.3385,
+      "step": 32024
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1483980736511233e-06,
+      "loss": 0.3623,
+      "step": 32025
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.147821322490412e-06,
+      "loss": 0.2794,
+      "step": 32026
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1472445713297e-06,
+      "loss": 0.3837,
+      "step": 32027
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1466678201689884e-06,
+      "loss": 0.3866,
+      "step": 32028
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1460910690082766e-06,
+      "loss": 0.393,
+      "step": 32029
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1455143178475647e-06,
+      "loss": 0.4127,
+      "step": 32030
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.144937566686853e-06,
+      "loss": 0.475,
+      "step": 32031
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1443608155261413e-06,
+      "loss": 0.3733,
+      "step": 32032
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.14378406436543e-06,
+      "loss": 0.4016,
+      "step": 32033
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.143207313204718e-06,
+      "loss": 0.3734,
+      "step": 32034
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1426305620440065e-06,
+      "loss": 0.3225,
+      "step": 32035
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1420538108832946e-06,
+      "loss": 0.3483,
+      "step": 32036
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1414770597225827e-06,
+      "loss": 0.4154,
+      "step": 32037
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1409003085618712e-06,
+      "loss": 0.4462,
+      "step": 32038
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1403235574011593e-06,
+      "loss": 0.3907,
+      "step": 32039
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.139746806240448e-06,
+      "loss": 0.4026,
+      "step": 32040
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.139170055079736e-06,
+      "loss": 0.4023,
+      "step": 32041
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1385933039190245e-06,
+      "loss": 0.3512,
+      "step": 32042
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1380165527583126e-06,
+      "loss": 0.3671,
+      "step": 32043
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1374398015976007e-06,
+      "loss": 0.3788,
+      "step": 32044
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1368630504368893e-06,
+      "loss": 0.4424,
+      "step": 32045
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1362862992761774e-06,
+      "loss": 0.3452,
+      "step": 32046
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.135709548115466e-06,
+      "loss": 0.4915,
+      "step": 32047
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.135132796954754e-06,
+      "loss": 0.4423,
+      "step": 32048
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.134556045794042e-06,
+      "loss": 0.3931,
+      "step": 32049
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1339792946333306e-06,
+      "loss": 0.4883,
+      "step": 32050
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1334025434726188e-06,
+      "loss": 0.429,
+      "step": 32051
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1328257923119073e-06,
+      "loss": 0.3938,
+      "step": 32052
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1322490411511954e-06,
+      "loss": 0.4091,
+      "step": 32053
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.131672289990484e-06,
+      "loss": 0.4039,
+      "step": 32054
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.131095538829772e-06,
+      "loss": 0.4377,
+      "step": 32055
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.13051878766906e-06,
+      "loss": 0.3707,
+      "step": 32056
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1299420365083487e-06,
+      "loss": 0.4106,
+      "step": 32057
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1293652853476368e-06,
+      "loss": 0.4552,
+      "step": 32058
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1287885341869253e-06,
+      "loss": 0.477,
+      "step": 32059
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1282117830262134e-06,
+      "loss": 0.4138,
+      "step": 32060
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1276350318655015e-06,
+      "loss": 0.289,
+      "step": 32061
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.12705828070479e-06,
+      "loss": 0.3818,
+      "step": 32062
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.126481529544078e-06,
+      "loss": 0.5103,
+      "step": 32063
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1259047783833667e-06,
+      "loss": 0.3433,
+      "step": 32064
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1253280272226552e-06,
+      "loss": 0.2832,
+      "step": 32065
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1247512760619433e-06,
+      "loss": 0.3075,
+      "step": 32066
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1241745249012315e-06,
+      "loss": 0.4443,
+      "step": 32067
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1235977737405196e-06,
+      "loss": 0.3585,
+      "step": 32068
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.123021022579808e-06,
+      "loss": 0.3941,
+      "step": 32069
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1224442714190966e-06,
+      "loss": 0.3019,
+      "step": 32070
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1218675202583847e-06,
+      "loss": 0.3608,
+      "step": 32071
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1212907690976733e-06,
+      "loss": 0.4589,
+      "step": 32072
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1207140179369614e-06,
+      "loss": 0.455,
+      "step": 32073
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1201372667762495e-06,
+      "loss": 0.4373,
+      "step": 32074
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1195605156155376e-06,
+      "loss": 0.3505,
+      "step": 32075
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.118983764454826e-06,
+      "loss": 0.4794,
+      "step": 32076
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1184070132941147e-06,
+      "loss": 0.4925,
+      "step": 32077
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1178302621334028e-06,
+      "loss": 0.4432,
+      "step": 32078
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.117253510972691e-06,
+      "loss": 0.4467,
+      "step": 32079
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.116676759811979e-06,
+      "loss": 0.2931,
+      "step": 32080
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1161000086512675e-06,
+      "loss": 0.3754,
+      "step": 32081
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.115523257490556e-06,
+      "loss": 0.2737,
+      "step": 32082
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.114946506329844e-06,
+      "loss": 0.4664,
+      "step": 32083
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.1143697551691327e-06,
+      "loss": 0.5122,
+      "step": 32084
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2.113793004008421e-06,
+      "loss": 0.4042,
+      "step": 32085
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.113216252847709e-06,
+      "loss": 0.3932,
+      "step": 32086
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1126395016869974e-06,
+      "loss": 0.3825,
+      "step": 32087
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1120627505262855e-06,
+      "loss": 0.4245,
+      "step": 32088
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.111485999365574e-06,
+      "loss": 0.3617,
+      "step": 32089
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.110909248204862e-06,
+      "loss": 0.4024,
+      "step": 32090
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1103324970441503e-06,
+      "loss": 0.3956,
+      "step": 32091
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.109755745883439e-06,
+      "loss": 0.3698,
+      "step": 32092
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.109178994722727e-06,
+      "loss": 0.39,
+      "step": 32093
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1086022435620155e-06,
+      "loss": 0.4217,
+      "step": 32094
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1080254924013036e-06,
+      "loss": 0.326,
+      "step": 32095
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.107448741240592e-06,
+      "loss": 0.3846,
+      "step": 32096
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1068719900798802e-06,
+      "loss": 0.3923,
+      "step": 32097
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1062952389191683e-06,
+      "loss": 0.3389,
+      "step": 32098
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.105718487758457e-06,
+      "loss": 0.4027,
+      "step": 32099
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.105141736597745e-06,
+      "loss": 0.3679,
+      "step": 32100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1045649854370335e-06,
+      "loss": 0.2743,
+      "step": 32101
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1039882342763216e-06,
+      "loss": 0.341,
+      "step": 32102
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.10341148311561e-06,
+      "loss": 0.4651,
+      "step": 32103
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1028347319548983e-06,
+      "loss": 0.4132,
+      "step": 32104
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1022579807941864e-06,
+      "loss": 0.318,
+      "step": 32105
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.101681229633475e-06,
+      "loss": 0.4207,
+      "step": 32106
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.101104478472763e-06,
+      "loss": 0.3777,
+      "step": 32107
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.1005277273120515e-06,
+      "loss": 0.4151,
+      "step": 32108
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0999509761513396e-06,
+      "loss": 0.3693,
+      "step": 32109
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0993742249906278e-06,
+      "loss": 0.4144,
+      "step": 32110
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0987974738299163e-06,
+      "loss": 0.4356,
+      "step": 32111
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0982207226692044e-06,
+      "loss": 0.3628,
+      "step": 32112
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.097643971508493e-06,
+      "loss": 0.3501,
+      "step": 32113
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.097067220347781e-06,
+      "loss": 0.4512,
+      "step": 32114
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0964904691870696e-06,
+      "loss": 0.3642,
+      "step": 32115
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0959137180263577e-06,
+      "loss": 0.3015,
+      "step": 32116
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0953369668656458e-06,
+      "loss": 0.3268,
+      "step": 32117
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0947602157049343e-06,
+      "loss": 0.497,
+      "step": 32118
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0941834645442224e-06,
+      "loss": 0.3294,
+      "step": 32119
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.093606713383511e-06,
+      "loss": 0.4393,
+      "step": 32120
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.093029962222799e-06,
+      "loss": 0.3816,
+      "step": 32121
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.092453211062087e-06,
+      "loss": 0.5169,
+      "step": 32122
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0918764599013757e-06,
+      "loss": 0.3159,
+      "step": 32123
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.091299708740664e-06,
+      "loss": 0.4351,
+      "step": 32124
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0907229575799523e-06,
+      "loss": 0.3963,
+      "step": 32125
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.090146206419241e-06,
+      "loss": 0.4321,
+      "step": 32126
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.089569455258529e-06,
+      "loss": 0.4466,
+      "step": 32127
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.088992704097817e-06,
+      "loss": 0.435,
+      "step": 32128
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.088415952937105e-06,
+      "loss": 0.3352,
+      "step": 32129
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0878392017763937e-06,
+      "loss": 0.4342,
+      "step": 32130
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0872624506156823e-06,
+      "loss": 0.4259,
+      "step": 32131
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0866856994549704e-06,
+      "loss": 0.4881,
+      "step": 32132
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0861089482942585e-06,
+      "loss": 0.3918,
+      "step": 32133
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.085532197133547e-06,
+      "loss": 0.4233,
+      "step": 32134
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.084955445972835e-06,
+      "loss": 0.5356,
+      "step": 32135
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0843786948121232e-06,
+      "loss": 0.464,
+      "step": 32136
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0838019436514118e-06,
+      "loss": 0.363,
+      "step": 32137
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0832251924907003e-06,
+      "loss": 0.5161,
+      "step": 32138
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0826484413299884e-06,
+      "loss": 0.3254,
+      "step": 32139
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0820716901692765e-06,
+      "loss": 0.394,
+      "step": 32140
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0814949390085646e-06,
+      "loss": 0.3813,
+      "step": 32141
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.080918187847853e-06,
+      "loss": 0.4425,
+      "step": 32142
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0803414366871417e-06,
+      "loss": 0.4565,
+      "step": 32143
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.07976468552643e-06,
+      "loss": 0.3727,
+      "step": 32144
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0791879343657183e-06,
+      "loss": 0.3947,
+      "step": 32145
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0786111832050064e-06,
+      "loss": 0.3019,
+      "step": 32146
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0780344320442945e-06,
+      "loss": 0.4474,
+      "step": 32147
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.077457680883583e-06,
+      "loss": 0.3345,
+      "step": 32148
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.076880929722871e-06,
+      "loss": 0.4751,
+      "step": 32149
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0763041785621597e-06,
+      "loss": 0.399,
+      "step": 32150
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.075727427401448e-06,
+      "loss": 0.4155,
+      "step": 32151
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.075150676240736e-06,
+      "loss": 0.4265,
+      "step": 32152
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0745739250800245e-06,
+      "loss": 0.3317,
+      "step": 32153
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0739971739193126e-06,
+      "loss": 0.4645,
+      "step": 32154
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.073420422758601e-06,
+      "loss": 0.4966,
+      "step": 32155
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0728436715978892e-06,
+      "loss": 0.3205,
+      "step": 32156
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0722669204371777e-06,
+      "loss": 0.2899,
+      "step": 32157
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.071690169276466e-06,
+      "loss": 0.3133,
+      "step": 32158
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.071113418115754e-06,
+      "loss": 0.3885,
+      "step": 32159
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0705366669550425e-06,
+      "loss": 0.5909,
+      "step": 32160
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0699599157943306e-06,
+      "loss": 0.3682,
+      "step": 32161
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.069383164633619e-06,
+      "loss": 0.4282,
+      "step": 32162
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0688064134729072e-06,
+      "loss": 0.4411,
+      "step": 32163
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0682296623121958e-06,
+      "loss": 0.3801,
+      "step": 32164
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.067652911151484e-06,
+      "loss": 0.4408,
+      "step": 32165
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.067076159990772e-06,
+      "loss": 0.4461,
+      "step": 32166
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0664994088300605e-06,
+      "loss": 0.3737,
+      "step": 32167
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0659226576693486e-06,
+      "loss": 0.3506,
+      "step": 32168
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.065345906508637e-06,
+      "loss": 0.4522,
+      "step": 32169
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0647691553479253e-06,
+      "loss": 0.4889,
+      "step": 32170
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0641924041872134e-06,
+      "loss": 0.4822,
+      "step": 32171
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.063615653026502e-06,
+      "loss": 0.4023,
+      "step": 32172
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.06303890186579e-06,
+      "loss": 0.2862,
+      "step": 32173
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0624621507050786e-06,
+      "loss": 0.4035,
+      "step": 32174
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0618853995443667e-06,
+      "loss": 0.4307,
+      "step": 32175
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.061308648383655e-06,
+      "loss": 0.3679,
+      "step": 32176
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0607318972229433e-06,
+      "loss": 0.3661,
+      "step": 32177
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0601551460622314e-06,
+      "loss": 0.3704,
+      "step": 32178
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.05957839490152e-06,
+      "loss": 0.4066,
+      "step": 32179
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.059001643740808e-06,
+      "loss": 0.3619,
+      "step": 32180
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0584248925800966e-06,
+      "loss": 0.3952,
+      "step": 32181
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0578481414193847e-06,
+      "loss": 0.3195,
+      "step": 32182
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.057271390258673e-06,
+      "loss": 0.301,
+      "step": 32183
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0566946390979613e-06,
+      "loss": 0.3864,
+      "step": 32184
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0561178879372494e-06,
+      "loss": 0.4493,
+      "step": 32185
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.055541136776538e-06,
+      "loss": 0.3843,
+      "step": 32186
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0549643856158265e-06,
+      "loss": 0.4234,
+      "step": 32187
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0543876344551146e-06,
+      "loss": 0.3454,
+      "step": 32188
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0538108832944027e-06,
+      "loss": 0.3718,
+      "step": 32189
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.053234132133691e-06,
+      "loss": 0.319,
+      "step": 32190
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0526573809729794e-06,
+      "loss": 0.2447,
+      "step": 32191
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.052080629812268e-06,
+      "loss": 0.3142,
+      "step": 32192
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.051503878651556e-06,
+      "loss": 0.4846,
+      "step": 32193
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.050927127490844e-06,
+      "loss": 0.4176,
+      "step": 32194
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0503503763301327e-06,
+      "loss": 0.4321,
+      "step": 32195
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0497736251694208e-06,
+      "loss": 0.4331,
+      "step": 32196
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.049196874008709e-06,
+      "loss": 0.402,
+      "step": 32197
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0486201228479974e-06,
+      "loss": 0.4222,
+      "step": 32198
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.048043371687286e-06,
+      "loss": 0.3554,
+      "step": 32199
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.047466620526574e-06,
+      "loss": 0.4324,
+      "step": 32200
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.046889869365862e-06,
+      "loss": 0.3392,
+      "step": 32201
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0463131182051503e-06,
+      "loss": 0.4409,
+      "step": 32202
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.045736367044439e-06,
+      "loss": 0.389,
+      "step": 32203
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0451596158837273e-06,
+      "loss": 0.5423,
+      "step": 32204
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0445828647230154e-06,
+      "loss": 0.4568,
+      "step": 32205
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.044006113562304e-06,
+      "loss": 0.4606,
+      "step": 32206
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.043429362401592e-06,
+      "loss": 0.317,
+      "step": 32207
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.04285261124088e-06,
+      "loss": 0.3571,
+      "step": 32208
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0422758600801687e-06,
+      "loss": 0.3651,
+      "step": 32209
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.041699108919457e-06,
+      "loss": 0.4308,
+      "step": 32210
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0411223577587454e-06,
+      "loss": 0.2943,
+      "step": 32211
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0405456065980335e-06,
+      "loss": 0.3631,
+      "step": 32212
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0399688554373216e-06,
+      "loss": 0.316,
+      "step": 32213
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0393921042766097e-06,
+      "loss": 0.3131,
+      "step": 32214
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0388153531158982e-06,
+      "loss": 0.4679,
+      "step": 32215
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0382386019551867e-06,
+      "loss": 0.4455,
+      "step": 32216
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.037661850794475e-06,
+      "loss": 0.5143,
+      "step": 32217
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0370850996337634e-06,
+      "loss": 0.3644,
+      "step": 32218
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0365083484730515e-06,
+      "loss": 0.4482,
+      "step": 32219
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0359315973123396e-06,
+      "loss": 0.377,
+      "step": 32220
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.035354846151628e-06,
+      "loss": 0.4127,
+      "step": 32221
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0347780949909162e-06,
+      "loss": 0.3092,
+      "step": 32222
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0342013438302048e-06,
+      "loss": 0.4075,
+      "step": 32223
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.033624592669493e-06,
+      "loss": 0.451,
+      "step": 32224
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.033047841508781e-06,
+      "loss": 0.2991,
+      "step": 32225
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0324710903480695e-06,
+      "loss": 0.4479,
+      "step": 32226
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0318943391873576e-06,
+      "loss": 0.4412,
+      "step": 32227
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.031317588026646e-06,
+      "loss": 0.3586,
+      "step": 32228
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0307408368659343e-06,
+      "loss": 0.4243,
+      "step": 32229
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.030164085705223e-06,
+      "loss": 0.3675,
+      "step": 32230
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.029587334544511e-06,
+      "loss": 0.4785,
+      "step": 32231
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.029010583383799e-06,
+      "loss": 0.3634,
+      "step": 32232
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0284338322230876e-06,
+      "loss": 0.5035,
+      "step": 32233
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0278570810623757e-06,
+      "loss": 0.3098,
+      "step": 32234
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.027280329901664e-06,
+      "loss": 0.3072,
+      "step": 32235
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0267035787409523e-06,
+      "loss": 0.3453,
+      "step": 32236
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.026126827580241e-06,
+      "loss": 0.4434,
+      "step": 32237
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.025550076419529e-06,
+      "loss": 0.4427,
+      "step": 32238
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.024973325258817e-06,
+      "loss": 0.461,
+      "step": 32239
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0243965740981056e-06,
+      "loss": 0.3555,
+      "step": 32240
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0238198229373937e-06,
+      "loss": 0.5204,
+      "step": 32241
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0232430717766822e-06,
+      "loss": 0.5066,
+      "step": 32242
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0226663206159703e-06,
+      "loss": 0.3502,
+      "step": 32243
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0220895694552584e-06,
+      "loss": 0.4008,
+      "step": 32244
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.021512818294547e-06,
+      "loss": 0.3697,
+      "step": 32245
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.020936067133835e-06,
+      "loss": 0.3826,
+      "step": 32246
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0203593159731236e-06,
+      "loss": 0.3827,
+      "step": 32247
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.019782564812412e-06,
+      "loss": 0.3572,
+      "step": 32248
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0192058136517003e-06,
+      "loss": 0.3356,
+      "step": 32249
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0186290624909884e-06,
+      "loss": 0.5311,
+      "step": 32250
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0180523113302765e-06,
+      "loss": 0.4038,
+      "step": 32251
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.017475560169565e-06,
+      "loss": 0.3664,
+      "step": 32252
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0168988090088535e-06,
+      "loss": 0.4017,
+      "step": 32253
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0163220578481416e-06,
+      "loss": 0.4503,
+      "step": 32254
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0157453066874298e-06,
+      "loss": 0.2833,
+      "step": 32255
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0151685555267183e-06,
+      "loss": 0.3868,
+      "step": 32256
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0145918043660064e-06,
+      "loss": 0.4965,
+      "step": 32257
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0140150532052945e-06,
+      "loss": 0.3282,
+      "step": 32258
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.013438302044583e-06,
+      "loss": 0.3273,
+      "step": 32259
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0128615508838716e-06,
+      "loss": 0.2731,
+      "step": 32260
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0122847997231597e-06,
+      "loss": 0.3418,
+      "step": 32261
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.011708048562448e-06,
+      "loss": 0.4633,
+      "step": 32262
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.011131297401736e-06,
+      "loss": 0.3903,
+      "step": 32263
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2.0105545462410244e-06,
+      "loss": 0.3521,
+      "step": 32264
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.009977795080313e-06,
+      "loss": 0.4648,
+      "step": 32265
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.009401043919601e-06,
+      "loss": 0.3101,
+      "step": 32266
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0088242927588896e-06,
+      "loss": 0.3814,
+      "step": 32267
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0082475415981777e-06,
+      "loss": 0.3485,
+      "step": 32268
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.007670790437466e-06,
+      "loss": 0.4522,
+      "step": 32269
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0070940392767544e-06,
+      "loss": 0.3437,
+      "step": 32270
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0065172881160425e-06,
+      "loss": 0.3727,
+      "step": 32271
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.005940536955331e-06,
+      "loss": 0.391,
+      "step": 32272
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.005363785794619e-06,
+      "loss": 0.4029,
+      "step": 32273
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.004787034633907e-06,
+      "loss": 0.5364,
+      "step": 32274
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0042102834731953e-06,
+      "loss": 0.4121,
+      "step": 32275
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.003633532312484e-06,
+      "loss": 0.3286,
+      "step": 32276
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0030567811517724e-06,
+      "loss": 0.4455,
+      "step": 32277
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0024800299910605e-06,
+      "loss": 0.5135,
+      "step": 32278
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.001903278830349e-06,
+      "loss": 0.4238,
+      "step": 32279
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.001326527669637e-06,
+      "loss": 0.432,
+      "step": 32280
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0007497765089252e-06,
+      "loss": 0.3785,
+      "step": 32281
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2.0001730253482138e-06,
+      "loss": 0.3897,
+      "step": 32282
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.999596274187502e-06,
+      "loss": 0.4041,
+      "step": 32283
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9990195230267904e-06,
+      "loss": 0.3835,
+      "step": 32284
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9984427718660785e-06,
+      "loss": 0.379,
+      "step": 32285
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9978660207053666e-06,
+      "loss": 0.4025,
+      "step": 32286
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.997289269544655e-06,
+      "loss": 0.4257,
+      "step": 32287
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9967125183839433e-06,
+      "loss": 0.4831,
+      "step": 32288
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.996135767223232e-06,
+      "loss": 0.3811,
+      "step": 32289
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.99555901606252e-06,
+      "loss": 0.3773,
+      "step": 32290
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9949822649018084e-06,
+      "loss": 0.33,
+      "step": 32291
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9944055137410966e-06,
+      "loss": 0.5582,
+      "step": 32292
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9938287625803847e-06,
+      "loss": 0.425,
+      "step": 32293
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.993252011419673e-06,
+      "loss": 0.4795,
+      "step": 32294
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9926752602589613e-06,
+      "loss": 0.398,
+      "step": 32295
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.99209850909825e-06,
+      "loss": 0.3124,
+      "step": 32296
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.991521757937538e-06,
+      "loss": 0.3452,
+      "step": 32297
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9909450067768265e-06,
+      "loss": 0.3346,
+      "step": 32298
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9903682556161146e-06,
+      "loss": 0.4894,
+      "step": 32299
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9897915044554027e-06,
+      "loss": 0.4762,
+      "step": 32300
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9892147532946912e-06,
+      "loss": 0.519,
+      "step": 32301
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9886380021339793e-06,
+      "loss": 0.3722,
+      "step": 32302
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.988061250973268e-06,
+      "loss": 0.3893,
+      "step": 32303
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.987484499812556e-06,
+      "loss": 0.3703,
+      "step": 32304
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.986907748651844e-06,
+      "loss": 0.3918,
+      "step": 32305
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9863309974911326e-06,
+      "loss": 0.4351,
+      "step": 32306
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9857542463304207e-06,
+      "loss": 0.3503,
+      "step": 32307
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9851774951697093e-06,
+      "loss": 0.4,
+      "step": 32308
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.984600744008998e-06,
+      "loss": 0.345,
+      "step": 32309
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.984023992848286e-06,
+      "loss": 0.3112,
+      "step": 32310
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.983447241687574e-06,
+      "loss": 0.4484,
+      "step": 32311
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.982870490526862e-06,
+      "loss": 0.3351,
+      "step": 32312
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9822937393661506e-06,
+      "loss": 0.3768,
+      "step": 32313
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.981716988205439e-06,
+      "loss": 0.4032,
+      "step": 32314
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9811402370447273e-06,
+      "loss": 0.3669,
+      "step": 32315
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9805634858840154e-06,
+      "loss": 0.3586,
+      "step": 32316
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.979986734723304e-06,
+      "loss": 0.3864,
+      "step": 32317
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.979409983562592e-06,
+      "loss": 0.4471,
+      "step": 32318
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.97883323240188e-06,
+      "loss": 0.3689,
+      "step": 32319
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9782564812411687e-06,
+      "loss": 0.5038,
+      "step": 32320
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.977679730080457e-06,
+      "loss": 0.4468,
+      "step": 32321
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9771029789197453e-06,
+      "loss": 0.3553,
+      "step": 32322
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9765262277590334e-06,
+      "loss": 0.3935,
+      "step": 32323
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9759494765983215e-06,
+      "loss": 0.4086,
+      "step": 32324
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.97537272543761e-06,
+      "loss": 0.4098,
+      "step": 32325
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9747959742768986e-06,
+      "loss": 0.4667,
+      "step": 32326
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9742192231161867e-06,
+      "loss": 0.3167,
+      "step": 32327
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9736424719554752e-06,
+      "loss": 0.3218,
+      "step": 32328
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9730657207947633e-06,
+      "loss": 0.3903,
+      "step": 32329
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9724889696340515e-06,
+      "loss": 0.4787,
+      "step": 32330
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.97191221847334e-06,
+      "loss": 0.5391,
+      "step": 32331
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.971335467312628e-06,
+      "loss": 0.2785,
+      "step": 32332
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9707587161519166e-06,
+      "loss": 0.3895,
+      "step": 32333
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9701819649912047e-06,
+      "loss": 0.3304,
+      "step": 32334
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.969605213830493e-06,
+      "loss": 0.3816,
+      "step": 32335
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.969028462669781e-06,
+      "loss": 0.4239,
+      "step": 32336
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9684517115090695e-06,
+      "loss": 0.4205,
+      "step": 32337
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.967874960348358e-06,
+      "loss": 0.295,
+      "step": 32338
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.967298209187646e-06,
+      "loss": 0.4802,
+      "step": 32339
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9667214580269347e-06,
+      "loss": 0.3227,
+      "step": 32340
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9661447068662228e-06,
+      "loss": 0.4921,
+      "step": 32341
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.965567955705511e-06,
+      "loss": 0.4187,
+      "step": 32342
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9649912045447994e-06,
+      "loss": 0.4036,
+      "step": 32343
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9644144533840875e-06,
+      "loss": 0.4307,
+      "step": 32344
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.963837702223376e-06,
+      "loss": 0.3972,
+      "step": 32345
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.963260951062664e-06,
+      "loss": 0.437,
+      "step": 32346
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9626841999019523e-06,
+      "loss": 0.4685,
+      "step": 32347
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.962107448741241e-06,
+      "loss": 0.3734,
+      "step": 32348
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.961530697580529e-06,
+      "loss": 0.363,
+      "step": 32349
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9609539464198174e-06,
+      "loss": 0.3812,
+      "step": 32350
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9603771952591056e-06,
+      "loss": 0.4078,
+      "step": 32351
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.959800444098394e-06,
+      "loss": 0.3625,
+      "step": 32352
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.959223692937682e-06,
+      "loss": 0.376,
+      "step": 32353
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9586469417769703e-06,
+      "loss": 0.4419,
+      "step": 32354
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.958070190616259e-06,
+      "loss": 0.3613,
+      "step": 32355
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.957493439455547e-06,
+      "loss": 0.3735,
+      "step": 32356
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9569166882948355e-06,
+      "loss": 0.4785,
+      "step": 32357
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9563399371341236e-06,
+      "loss": 0.3193,
+      "step": 32358
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.955763185973412e-06,
+      "loss": 0.4264,
+      "step": 32359
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9551864348127002e-06,
+      "loss": 0.2629,
+      "step": 32360
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9546096836519883e-06,
+      "loss": 0.3489,
+      "step": 32361
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.954032932491277e-06,
+      "loss": 0.4758,
+      "step": 32362
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.953456181330565e-06,
+      "loss": 0.4606,
+      "step": 32363
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9528794301698535e-06,
+      "loss": 0.3813,
+      "step": 32364
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9523026790091416e-06,
+      "loss": 0.4411,
+      "step": 32365
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9517259278484297e-06,
+      "loss": 0.3525,
+      "step": 32366
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9511491766877183e-06,
+      "loss": 0.3542,
+      "step": 32367
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9505724255270064e-06,
+      "loss": 0.3833,
+      "step": 32368
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.949995674366295e-06,
+      "loss": 0.3903,
+      "step": 32369
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9494189232055834e-06,
+      "loss": 0.3211,
+      "step": 32370
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9488421720448715e-06,
+      "loss": 0.3555,
+      "step": 32371
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9482654208841596e-06,
+      "loss": 0.4934,
+      "step": 32372
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9476886697234478e-06,
+      "loss": 0.3605,
+      "step": 32373
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9471119185627363e-06,
+      "loss": 0.44,
+      "step": 32374
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9465351674020244e-06,
+      "loss": 0.3808,
+      "step": 32375
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.945958416241313e-06,
+      "loss": 0.3833,
+      "step": 32376
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.945381665080601e-06,
+      "loss": 0.3345,
+      "step": 32377
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.944804913919889e-06,
+      "loss": 0.3428,
+      "step": 32378
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9442281627591777e-06,
+      "loss": 0.4527,
+      "step": 32379
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9436514115984658e-06,
+      "loss": 0.4755,
+      "step": 32380
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9430746604377543e-06,
+      "loss": 0.4627,
+      "step": 32381
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.942497909277043e-06,
+      "loss": 0.4646,
+      "step": 32382
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.941921158116331e-06,
+      "loss": 0.3868,
+      "step": 32383
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.941344406955619e-06,
+      "loss": 0.3007,
+      "step": 32384
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.940767655794907e-06,
+      "loss": 0.4362,
+      "step": 32385
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9401909046341957e-06,
+      "loss": 0.2784,
+      "step": 32386
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9396141534734842e-06,
+      "loss": 0.4506,
+      "step": 32387
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9390374023127723e-06,
+      "loss": 0.464,
+      "step": 32388
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.938460651152061e-06,
+      "loss": 0.4324,
+      "step": 32389
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.937883899991349e-06,
+      "loss": 0.4069,
+      "step": 32390
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.937307148830637e-06,
+      "loss": 0.3972,
+      "step": 32391
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9367303976699256e-06,
+      "loss": 0.435,
+      "step": 32392
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9361536465092137e-06,
+      "loss": 0.4204,
+      "step": 32393
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9355768953485023e-06,
+      "loss": 0.4426,
+      "step": 32394
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9350001441877904e-06,
+      "loss": 0.4945,
+      "step": 32395
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9344233930270785e-06,
+      "loss": 0.3595,
+      "step": 32396
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9338466418663666e-06,
+      "loss": 0.4319,
+      "step": 32397
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.933269890705655e-06,
+      "loss": 0.4891,
+      "step": 32398
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9326931395449437e-06,
+      "loss": 0.4735,
+      "step": 32399
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9321163883842318e-06,
+      "loss": 0.4943,
+      "step": 32400
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9315396372235203e-06,
+      "loss": 0.4115,
+      "step": 32401
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9309628860628084e-06,
+      "loss": 0.5318,
+      "step": 32402
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9303861349020965e-06,
+      "loss": 0.4675,
+      "step": 32403
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.929809383741385e-06,
+      "loss": 0.3815,
+      "step": 32404
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.929232632580673e-06,
+      "loss": 0.4217,
+      "step": 32405
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9286558814199617e-06,
+      "loss": 0.3278,
+      "step": 32406
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.92807913025925e-06,
+      "loss": 0.4217,
+      "step": 32407
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.927502379098538e-06,
+      "loss": 0.4204,
+      "step": 32408
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9269256279378264e-06,
+      "loss": 0.3579,
+      "step": 32409
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9263488767771145e-06,
+      "loss": 0.4498,
+      "step": 32410
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.925772125616403e-06,
+      "loss": 0.3371,
+      "step": 32411
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.925195374455691e-06,
+      "loss": 0.4834,
+      "step": 32412
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9246186232949797e-06,
+      "loss": 0.3212,
+      "step": 32413
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.924041872134268e-06,
+      "loss": 0.4152,
+      "step": 32414
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.923465120973556e-06,
+      "loss": 0.2831,
+      "step": 32415
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9228883698128445e-06,
+      "loss": 0.3555,
+      "step": 32416
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9223116186521326e-06,
+      "loss": 0.3916,
+      "step": 32417
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.921734867491421e-06,
+      "loss": 0.3595,
+      "step": 32418
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9211581163307092e-06,
+      "loss": 0.3382,
+      "step": 32419
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9205813651699978e-06,
+      "loss": 0.4095,
+      "step": 32420
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.920004614009286e-06,
+      "loss": 0.4307,
+      "step": 32421
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.919427862848574e-06,
+      "loss": 0.4594,
+      "step": 32422
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9188511116878625e-06,
+      "loss": 0.4467,
+      "step": 32423
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9182743605271506e-06,
+      "loss": 0.3734,
+      "step": 32424
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.917697609366439e-06,
+      "loss": 0.3705,
+      "step": 32425
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9171208582057272e-06,
+      "loss": 0.4231,
+      "step": 32426
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9165441070450154e-06,
+      "loss": 0.5922,
+      "step": 32427
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.915967355884304e-06,
+      "loss": 0.3888,
+      "step": 32428
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.915390604723592e-06,
+      "loss": 0.3672,
+      "step": 32429
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9148138535628805e-06,
+      "loss": 0.5981,
+      "step": 32430
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.914237102402169e-06,
+      "loss": 0.4285,
+      "step": 32431
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.913660351241457e-06,
+      "loss": 0.3661,
+      "step": 32432
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9130836000807453e-06,
+      "loss": 0.3114,
+      "step": 32433
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9125068489200334e-06,
+      "loss": 0.5233,
+      "step": 32434
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.911930097759322e-06,
+      "loss": 0.354,
+      "step": 32435
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.91135334659861e-06,
+      "loss": 0.3829,
+      "step": 32436
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9107765954378986e-06,
+      "loss": 0.2865,
+      "step": 32437
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9101998442771867e-06,
+      "loss": 0.3599,
+      "step": 32438
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9096230931164748e-06,
+      "loss": 0.2785,
+      "step": 32439
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9090463419557633e-06,
+      "loss": 0.3821,
+      "step": 32440
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9084695907950514e-06,
+      "loss": 0.3981,
+      "step": 32441
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.90789283963434e-06,
+      "loss": 0.362,
+      "step": 32442
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 1.9073160884736283e-06,
+      "loss": 0.3692,
+      "step": 32443
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9067393373129164e-06,
+      "loss": 0.4717,
+      "step": 32444
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.906162586152205e-06,
+      "loss": 0.3927,
+      "step": 32445
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.905585834991493e-06,
+      "loss": 0.3335,
+      "step": 32446
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9050090838307813e-06,
+      "loss": 0.4183,
+      "step": 32447
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9044323326700699e-06,
+      "loss": 0.4211,
+      "step": 32448
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.903855581509358e-06,
+      "loss": 0.4312,
+      "step": 32449
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9032788303486463e-06,
+      "loss": 0.4061,
+      "step": 32450
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9027020791879344e-06,
+      "loss": 0.4077,
+      "step": 32451
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9021253280272227e-06,
+      "loss": 0.3603,
+      "step": 32452
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9015485768665113e-06,
+      "loss": 0.3348,
+      "step": 32453
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9009718257057994e-06,
+      "loss": 0.4301,
+      "step": 32454
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.9003950745450877e-06,
+      "loss": 0.389,
+      "step": 32455
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8998183233843758e-06,
+      "loss": 0.5384,
+      "step": 32456
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8992415722236643e-06,
+      "loss": 0.4271,
+      "step": 32457
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8986648210629524e-06,
+      "loss": 0.4053,
+      "step": 32458
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8980880699022408e-06,
+      "loss": 0.5227,
+      "step": 32459
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8975113187415293e-06,
+      "loss": 0.4339,
+      "step": 32460
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8969345675808174e-06,
+      "loss": 0.4589,
+      "step": 32461
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8963578164201057e-06,
+      "loss": 0.3176,
+      "step": 32462
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8957810652593938e-06,
+      "loss": 0.45,
+      "step": 32463
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8952043140986824e-06,
+      "loss": 0.3661,
+      "step": 32464
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8946275629379707e-06,
+      "loss": 0.3717,
+      "step": 32465
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8940508117772588e-06,
+      "loss": 0.413,
+      "step": 32466
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8934740606165471e-06,
+      "loss": 0.3751,
+      "step": 32467
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8928973094558354e-06,
+      "loss": 0.4802,
+      "step": 32468
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8923205582951238e-06,
+      "loss": 0.3573,
+      "step": 32469
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.891743807134412e-06,
+      "loss": 0.4301,
+      "step": 32470
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8911670559737002e-06,
+      "loss": 0.4493,
+      "step": 32471
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8905903048129887e-06,
+      "loss": 0.3797,
+      "step": 32472
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8900135536522768e-06,
+      "loss": 0.402,
+      "step": 32473
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8894368024915651e-06,
+      "loss": 0.4102,
+      "step": 32474
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8888600513308537e-06,
+      "loss": 0.396,
+      "step": 32475
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8882833001701418e-06,
+      "loss": 0.3438,
+      "step": 32476
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8877065490094301e-06,
+      "loss": 0.3185,
+      "step": 32477
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8871297978487182e-06,
+      "loss": 0.4771,
+      "step": 32478
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8865530466880067e-06,
+      "loss": 0.4742,
+      "step": 32479
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8859762955272949e-06,
+      "loss": 0.3286,
+      "step": 32480
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8853995443665832e-06,
+      "loss": 0.4718,
+      "step": 32481
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8848227932058715e-06,
+      "loss": 0.3765,
+      "step": 32482
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8842460420451596e-06,
+      "loss": 0.4666,
+      "step": 32483
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8836692908844481e-06,
+      "loss": 0.345,
+      "step": 32484
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8830925397237362e-06,
+      "loss": 0.4616,
+      "step": 32485
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8825157885630246e-06,
+      "loss": 0.3555,
+      "step": 32486
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.881939037402313e-06,
+      "loss": 0.2657,
+      "step": 32487
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8813622862416012e-06,
+      "loss": 0.3652,
+      "step": 32488
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8807855350808895e-06,
+      "loss": 0.3545,
+      "step": 32489
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8802087839201776e-06,
+      "loss": 0.446,
+      "step": 32490
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8796320327594662e-06,
+      "loss": 0.3363,
+      "step": 32491
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8790552815987545e-06,
+      "loss": 0.3445,
+      "step": 32492
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8784785304380426e-06,
+      "loss": 0.4624,
+      "step": 32493
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8779017792773311e-06,
+      "loss": 0.4759,
+      "step": 32494
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8773250281166192e-06,
+      "loss": 0.4573,
+      "step": 32495
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8767482769559076e-06,
+      "loss": 0.4566,
+      "step": 32496
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8761715257951957e-06,
+      "loss": 0.3824,
+      "step": 32497
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.875594774634484e-06,
+      "loss": 0.3675,
+      "step": 32498
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8750180234737725e-06,
+      "loss": 0.4155,
+      "step": 32499
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8744412723130606e-06,
+      "loss": 0.3411,
+      "step": 32500
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.873864521152349e-06,
+      "loss": 0.4141,
+      "step": 32501
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.873287769991637e-06,
+      "loss": 0.2871,
+      "step": 32502
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8727110188309256e-06,
+      "loss": 0.2572,
+      "step": 32503
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.872134267670214e-06,
+      "loss": 0.3297,
+      "step": 32504
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.871557516509502e-06,
+      "loss": 0.3102,
+      "step": 32505
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8709807653487905e-06,
+      "loss": 0.4382,
+      "step": 32506
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8704040141880787e-06,
+      "loss": 0.4187,
+      "step": 32507
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.869827263027367e-06,
+      "loss": 0.3177,
+      "step": 32508
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8692505118666553e-06,
+      "loss": 0.4499,
+      "step": 32509
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8686737607059436e-06,
+      "loss": 0.3482,
+      "step": 32510
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.868097009545232e-06,
+      "loss": 0.4089,
+      "step": 32511
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.86752025838452e-06,
+      "loss": 0.4219,
+      "step": 32512
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8669435072238084e-06,
+      "loss": 0.3471,
+      "step": 32513
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.866366756063097e-06,
+      "loss": 0.3962,
+      "step": 32514
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.865790004902385e-06,
+      "loss": 0.4483,
+      "step": 32515
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8652132537416733e-06,
+      "loss": 0.4001,
+      "step": 32516
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8646365025809614e-06,
+      "loss": 0.3976,
+      "step": 32517
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.86405975142025e-06,
+      "loss": 0.3831,
+      "step": 32518
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.863483000259538e-06,
+      "loss": 0.3883,
+      "step": 32519
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8629062490988264e-06,
+      "loss": 0.3525,
+      "step": 32520
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.862329497938115e-06,
+      "loss": 0.2982,
+      "step": 32521
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.861752746777403e-06,
+      "loss": 0.3751,
+      "step": 32522
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8611759956166914e-06,
+      "loss": 0.4016,
+      "step": 32523
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8605992444559795e-06,
+      "loss": 0.4326,
+      "step": 32524
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.860022493295268e-06,
+      "loss": 0.3791,
+      "step": 32525
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8594457421345563e-06,
+      "loss": 0.3952,
+      "step": 32526
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8588689909738444e-06,
+      "loss": 0.4168,
+      "step": 32527
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8582922398131328e-06,
+      "loss": 0.4086,
+      "step": 32528
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8577154886524209e-06,
+      "loss": 0.4182,
+      "step": 32529
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8571387374917094e-06,
+      "loss": 0.4639,
+      "step": 32530
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8565619863309977e-06,
+      "loss": 0.3062,
+      "step": 32531
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8559852351702858e-06,
+      "loss": 0.4399,
+      "step": 32532
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8554084840095744e-06,
+      "loss": 0.4771,
+      "step": 32533
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8548317328488625e-06,
+      "loss": 0.4571,
+      "step": 32534
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8542549816881508e-06,
+      "loss": 0.4352,
+      "step": 32535
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8536782305274393e-06,
+      "loss": 0.4811,
+      "step": 32536
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8531014793667274e-06,
+      "loss": 0.3737,
+      "step": 32537
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8525247282060157e-06,
+      "loss": 0.4435,
+      "step": 32538
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8519479770453039e-06,
+      "loss": 0.3161,
+      "step": 32539
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8513712258845924e-06,
+      "loss": 0.4252,
+      "step": 32540
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8507944747238805e-06,
+      "loss": 0.287,
+      "step": 32541
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8502177235631688e-06,
+      "loss": 0.449,
+      "step": 32542
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8496409724024571e-06,
+      "loss": 0.3733,
+      "step": 32543
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8490642212417452e-06,
+      "loss": 0.3761,
+      "step": 32544
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8484874700810338e-06,
+      "loss": 0.3538,
+      "step": 32545
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8479107189203219e-06,
+      "loss": 0.3968,
+      "step": 32546
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8473339677596102e-06,
+      "loss": 0.3462,
+      "step": 32547
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8467572165988987e-06,
+      "loss": 0.3715,
+      "step": 32548
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8461804654381868e-06,
+      "loss": 0.4162,
+      "step": 32549
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8456037142774752e-06,
+      "loss": 0.3458,
+      "step": 32550
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8450269631167633e-06,
+      "loss": 0.5302,
+      "step": 32551
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8444502119560518e-06,
+      "loss": 0.39,
+      "step": 32552
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8438734607953401e-06,
+      "loss": 0.4066,
+      "step": 32553
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8432967096346282e-06,
+      "loss": 0.3593,
+      "step": 32554
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8427199584739166e-06,
+      "loss": 0.541,
+      "step": 32555
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8421432073132049e-06,
+      "loss": 0.4621,
+      "step": 32556
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8415664561524932e-06,
+      "loss": 0.4045,
+      "step": 32557
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8409897049917813e-06,
+      "loss": 0.3465,
+      "step": 32558
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8404129538310696e-06,
+      "loss": 0.4708,
+      "step": 32559
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8398362026703582e-06,
+      "loss": 0.4376,
+      "step": 32560
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8392594515096463e-06,
+      "loss": 0.396,
+      "step": 32561
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8386827003489346e-06,
+      "loss": 0.2946,
+      "step": 32562
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8381059491882227e-06,
+      "loss": 0.4345,
+      "step": 32563
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8375291980275112e-06,
+      "loss": 0.4372,
+      "step": 32564
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8369524468667995e-06,
+      "loss": 0.4325,
+      "step": 32565
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8363756957060877e-06,
+      "loss": 0.4893,
+      "step": 32566
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8357989445453762e-06,
+      "loss": 0.3824,
+      "step": 32567
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8352221933846643e-06,
+      "loss": 0.3399,
+      "step": 32568
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8346454422239526e-06,
+      "loss": 0.4312,
+      "step": 32569
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.834068691063241e-06,
+      "loss": 0.4094,
+      "step": 32570
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8334919399025293e-06,
+      "loss": 0.4766,
+      "step": 32571
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8329151887418176e-06,
+      "loss": 0.3544,
+      "step": 32572
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8323384375811057e-06,
+      "loss": 0.4484,
+      "step": 32573
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.831761686420394e-06,
+      "loss": 0.3407,
+      "step": 32574
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8311849352596825e-06,
+      "loss": 0.5195,
+      "step": 32575
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8306081840989706e-06,
+      "loss": 0.417,
+      "step": 32576
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.830031432938259e-06,
+      "loss": 0.4223,
+      "step": 32577
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.829454681777547e-06,
+      "loss": 0.3133,
+      "step": 32578
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8288779306168356e-06,
+      "loss": 0.4551,
+      "step": 32579
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8283011794561237e-06,
+      "loss": 0.447,
+      "step": 32580
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.827724428295412e-06,
+      "loss": 0.3368,
+      "step": 32581
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8271476771347006e-06,
+      "loss": 0.452,
+      "step": 32582
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8265709259739887e-06,
+      "loss": 0.5211,
+      "step": 32583
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.825994174813277e-06,
+      "loss": 0.3454,
+      "step": 32584
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8254174236525651e-06,
+      "loss": 0.4756,
+      "step": 32585
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8248406724918536e-06,
+      "loss": 0.2536,
+      "step": 32586
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.824263921331142e-06,
+      "loss": 0.4239,
+      "step": 32587
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.82368717017043e-06,
+      "loss": 0.4731,
+      "step": 32588
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8231104190097184e-06,
+      "loss": 0.3646,
+      "step": 32589
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8225336678490065e-06,
+      "loss": 0.395,
+      "step": 32590
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.821956916688295e-06,
+      "loss": 0.3911,
+      "step": 32591
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8213801655275833e-06,
+      "loss": 0.3774,
+      "step": 32592
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8208034143668715e-06,
+      "loss": 0.3123,
+      "step": 32593
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.82022666320616e-06,
+      "loss": 0.4248,
+      "step": 32594
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.819649912045448e-06,
+      "loss": 0.357,
+      "step": 32595
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8190731608847364e-06,
+      "loss": 0.3124,
+      "step": 32596
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8184964097240245e-06,
+      "loss": 0.3404,
+      "step": 32597
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.817919658563313e-06,
+      "loss": 0.3849,
+      "step": 32598
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8173429074026014e-06,
+      "loss": 0.495,
+      "step": 32599
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8167661562418895e-06,
+      "loss": 0.4628,
+      "step": 32600
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8161894050811778e-06,
+      "loss": 0.4744,
+      "step": 32601
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8156126539204661e-06,
+      "loss": 0.3805,
+      "step": 32602
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8150359027597545e-06,
+      "loss": 0.3657,
+      "step": 32603
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8144591515990428e-06,
+      "loss": 0.3633,
+      "step": 32604
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8138824004383309e-06,
+      "loss": 0.5453,
+      "step": 32605
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8133056492776194e-06,
+      "loss": 0.4236,
+      "step": 32606
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8127288981169075e-06,
+      "loss": 0.3139,
+      "step": 32607
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8121521469561958e-06,
+      "loss": 0.3996,
+      "step": 32608
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8115753957954844e-06,
+      "loss": 0.3834,
+      "step": 32609
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8109986446347725e-06,
+      "loss": 0.4388,
+      "step": 32610
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8104218934740608e-06,
+      "loss": 0.4544,
+      "step": 32611
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.809845142313349e-06,
+      "loss": 0.4325,
+      "step": 32612
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8092683911526374e-06,
+      "loss": 0.4802,
+      "step": 32613
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8086916399919258e-06,
+      "loss": 0.4291,
+      "step": 32614
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8081148888312139e-06,
+      "loss": 0.4045,
+      "step": 32615
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8075381376705022e-06,
+      "loss": 0.4515,
+      "step": 32616
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8069613865097905e-06,
+      "loss": 0.4531,
+      "step": 32617
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8063846353490788e-06,
+      "loss": 0.353,
+      "step": 32618
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.805807884188367e-06,
+      "loss": 0.3811,
+      "step": 32619
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8052311330276553e-06,
+      "loss": 0.4589,
+      "step": 32620
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.8046543818669438e-06,
+      "loss": 0.4514,
+      "step": 32621
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.804077630706232e-06,
+      "loss": 0.5025,
+      "step": 32622
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8035008795455202e-06,
+      "loss": 0.3586,
+      "step": 32623
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8029241283848083e-06,
+      "loss": 0.4616,
+      "step": 32624
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8023473772240969e-06,
+      "loss": 0.3689,
+      "step": 32625
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8017706260633852e-06,
+      "loss": 0.3793,
+      "step": 32626
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8011938749026733e-06,
+      "loss": 0.3929,
+      "step": 32627
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.8006171237419618e-06,
+      "loss": 0.3999,
+      "step": 32628
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.80004037258125e-06,
+      "loss": 0.4759,
+      "step": 32629
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7994636214205383e-06,
+      "loss": 0.3198,
+      "step": 32630
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7988868702598266e-06,
+      "loss": 0.4391,
+      "step": 32631
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.798310119099115e-06,
+      "loss": 0.4464,
+      "step": 32632
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7977333679384032e-06,
+      "loss": 0.3625,
+      "step": 32633
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7971566167776913e-06,
+      "loss": 0.3403,
+      "step": 32634
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7965798656169796e-06,
+      "loss": 0.4413,
+      "step": 32635
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7960031144562682e-06,
+      "loss": 0.372,
+      "step": 32636
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7954263632955563e-06,
+      "loss": 0.5112,
+      "step": 32637
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7948496121348446e-06,
+      "loss": 0.3382,
+      "step": 32638
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7942728609741327e-06,
+      "loss": 0.3311,
+      "step": 32639
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7936961098134212e-06,
+      "loss": 0.4446,
+      "step": 32640
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7931193586527094e-06,
+      "loss": 0.4954,
+      "step": 32641
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7925426074919977e-06,
+      "loss": 0.3631,
+      "step": 32642
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7919658563312862e-06,
+      "loss": 0.5334,
+      "step": 32643
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7913891051705743e-06,
+      "loss": 0.4045,
+      "step": 32644
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7908123540098626e-06,
+      "loss": 0.3693,
+      "step": 32645
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7902356028491507e-06,
+      "loss": 0.4392,
+      "step": 32646
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.789658851688439e-06,
+      "loss": 0.2783,
+      "step": 32647
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7890821005277276e-06,
+      "loss": 0.4214,
+      "step": 32648
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7885053493670157e-06,
+      "loss": 0.3507,
+      "step": 32649
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.787928598206304e-06,
+      "loss": 0.4284,
+      "step": 32650
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7873518470455921e-06,
+      "loss": 0.4382,
+      "step": 32651
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7867750958848807e-06,
+      "loss": 0.5195,
+      "step": 32652
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.786198344724169e-06,
+      "loss": 0.4065,
+      "step": 32653
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.785621593563457e-06,
+      "loss": 0.41,
+      "step": 32654
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7850448424027456e-06,
+      "loss": 0.3411,
+      "step": 32655
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7844680912420337e-06,
+      "loss": 0.3891,
+      "step": 32656
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.783891340081322e-06,
+      "loss": 0.4961,
+      "step": 32657
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7833145889206102e-06,
+      "loss": 0.2723,
+      "step": 32658
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7827378377598987e-06,
+      "loss": 0.3572,
+      "step": 32659
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.782161086599187e-06,
+      "loss": 0.4541,
+      "step": 32660
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7815843354384751e-06,
+      "loss": 0.4702,
+      "step": 32661
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7810075842777634e-06,
+      "loss": 0.2831,
+      "step": 32662
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7804308331170518e-06,
+      "loss": 0.3938,
+      "step": 32663
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.77985408195634e-06,
+      "loss": 0.359,
+      "step": 32664
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7792773307956284e-06,
+      "loss": 0.3523,
+      "step": 32665
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7787005796349165e-06,
+      "loss": 0.4323,
+      "step": 32666
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.778123828474205e-06,
+      "loss": 0.4147,
+      "step": 32667
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7775470773134932e-06,
+      "loss": 0.3515,
+      "step": 32668
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7769703261527815e-06,
+      "loss": 0.4445,
+      "step": 32669
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.77639357499207e-06,
+      "loss": 0.3332,
+      "step": 32670
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7758168238313581e-06,
+      "loss": 0.4693,
+      "step": 32671
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7752400726706464e-06,
+      "loss": 0.4006,
+      "step": 32672
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7746633215099345e-06,
+      "loss": 0.3908,
+      "step": 32673
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.774086570349223e-06,
+      "loss": 0.4228,
+      "step": 32674
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7735098191885114e-06,
+      "loss": 0.4997,
+      "step": 32675
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7729330680277995e-06,
+      "loss": 0.3375,
+      "step": 32676
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7723563168670878e-06,
+      "loss": 0.4366,
+      "step": 32677
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7717795657063761e-06,
+      "loss": 0.3653,
+      "step": 32678
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7712028145456645e-06,
+      "loss": 0.3921,
+      "step": 32679
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7706260633849526e-06,
+      "loss": 0.368,
+      "step": 32680
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.770049312224241e-06,
+      "loss": 0.4896,
+      "step": 32681
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7694725610635294e-06,
+      "loss": 0.4767,
+      "step": 32682
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7688958099028175e-06,
+      "loss": 0.4346,
+      "step": 32683
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7683190587421059e-06,
+      "loss": 0.315,
+      "step": 32684
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.767742307581394e-06,
+      "loss": 0.4305,
+      "step": 32685
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7671655564206825e-06,
+      "loss": 0.3621,
+      "step": 32686
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7665888052599708e-06,
+      "loss": 0.4691,
+      "step": 32687
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.766012054099259e-06,
+      "loss": 0.5016,
+      "step": 32688
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7654353029385475e-06,
+      "loss": 0.4196,
+      "step": 32689
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7648585517778356e-06,
+      "loss": 0.3112,
+      "step": 32690
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7642818006171239e-06,
+      "loss": 0.3787,
+      "step": 32691
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7637050494564122e-06,
+      "loss": 0.3785,
+      "step": 32692
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7631282982957003e-06,
+      "loss": 0.482,
+      "step": 32693
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7625515471349889e-06,
+      "loss": 0.4599,
+      "step": 32694
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.761974795974277e-06,
+      "loss": 0.4328,
+      "step": 32695
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7613980448135653e-06,
+      "loss": 0.3603,
+      "step": 32696
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7608212936528538e-06,
+      "loss": 0.3971,
+      "step": 32697
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.760244542492142e-06,
+      "loss": 0.4314,
+      "step": 32698
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7596677913314302e-06,
+      "loss": 0.4004,
+      "step": 32699
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7590910401707184e-06,
+      "loss": 0.4632,
+      "step": 32700
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7585142890100069e-06,
+      "loss": 0.5102,
+      "step": 32701
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.757937537849295e-06,
+      "loss": 0.4137,
+      "step": 32702
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7573607866885833e-06,
+      "loss": 0.5944,
+      "step": 32703
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7567840355278718e-06,
+      "loss": 0.4167,
+      "step": 32704
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.75620728436716e-06,
+      "loss": 0.4176,
+      "step": 32705
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7556305332064483e-06,
+      "loss": 0.4296,
+      "step": 32706
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7550537820457364e-06,
+      "loss": 0.2951,
+      "step": 32707
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7544770308850247e-06,
+      "loss": 0.3355,
+      "step": 32708
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7539002797243132e-06,
+      "loss": 0.4735,
+      "step": 32709
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7533235285636013e-06,
+      "loss": 0.4035,
+      "step": 32710
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7527467774028897e-06,
+      "loss": 0.3507,
+      "step": 32711
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7521700262421778e-06,
+      "loss": 0.3952,
+      "step": 32712
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7515932750814663e-06,
+      "loss": 0.4077,
+      "step": 32713
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7510165239207546e-06,
+      "loss": 0.3406,
+      "step": 32714
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7504397727600427e-06,
+      "loss": 0.4052,
+      "step": 32715
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7498630215993313e-06,
+      "loss": 0.5483,
+      "step": 32716
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7492862704386194e-06,
+      "loss": 0.2912,
+      "step": 32717
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7487095192779077e-06,
+      "loss": 0.3015,
+      "step": 32718
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7481327681171958e-06,
+      "loss": 0.415,
+      "step": 32719
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7475560169564843e-06,
+      "loss": 0.398,
+      "step": 32720
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7469792657957727e-06,
+      "loss": 0.2231,
+      "step": 32721
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7464025146350608e-06,
+      "loss": 0.3949,
+      "step": 32722
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.745825763474349e-06,
+      "loss": 0.4816,
+      "step": 32723
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7452490123136374e-06,
+      "loss": 0.2979,
+      "step": 32724
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7446722611529257e-06,
+      "loss": 0.5016,
+      "step": 32725
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.744095509992214e-06,
+      "loss": 0.306,
+      "step": 32726
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7435187588315022e-06,
+      "loss": 0.3467,
+      "step": 32727
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7429420076707907e-06,
+      "loss": 0.3442,
+      "step": 32728
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7423652565100788e-06,
+      "loss": 0.3405,
+      "step": 32729
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7417885053493671e-06,
+      "loss": 0.4253,
+      "step": 32730
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7412117541886556e-06,
+      "loss": 0.3977,
+      "step": 32731
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7406350030279438e-06,
+      "loss": 0.3306,
+      "step": 32732
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.740058251867232e-06,
+      "loss": 0.3833,
+      "step": 32733
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7394815007065202e-06,
+      "loss": 0.5395,
+      "step": 32734
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7389047495458087e-06,
+      "loss": 0.4518,
+      "step": 32735
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.738327998385097e-06,
+      "loss": 0.3354,
+      "step": 32736
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7377512472243851e-06,
+      "loss": 0.3448,
+      "step": 32737
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7371744960636735e-06,
+      "loss": 0.3112,
+      "step": 32738
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7365977449029616e-06,
+      "loss": 0.4275,
+      "step": 32739
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7360209937422501e-06,
+      "loss": 0.2825,
+      "step": 32740
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7354442425815382e-06,
+      "loss": 0.4517,
+      "step": 32741
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7348674914208265e-06,
+      "loss": 0.4618,
+      "step": 32742
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.734290740260115e-06,
+      "loss": 0.437,
+      "step": 32743
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7337139890994032e-06,
+      "loss": 0.3816,
+      "step": 32744
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7331372379386915e-06,
+      "loss": 0.4378,
+      "step": 32745
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7325604867779796e-06,
+      "loss": 0.3454,
+      "step": 32746
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7319837356172681e-06,
+      "loss": 0.3182,
+      "step": 32747
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7314069844565565e-06,
+      "loss": 0.3288,
+      "step": 32748
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7308302332958446e-06,
+      "loss": 0.3674,
+      "step": 32749
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.730253482135133e-06,
+      "loss": 0.346,
+      "step": 32750
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7296767309744212e-06,
+      "loss": 0.5488,
+      "step": 32751
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7290999798137095e-06,
+      "loss": 0.3512,
+      "step": 32752
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7285232286529978e-06,
+      "loss": 0.441,
+      "step": 32753
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.727946477492286e-06,
+      "loss": 0.3694,
+      "step": 32754
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7273697263315745e-06,
+      "loss": 0.3479,
+      "step": 32755
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7267929751708626e-06,
+      "loss": 0.3346,
+      "step": 32756
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.726216224010151e-06,
+      "loss": 0.3746,
+      "step": 32757
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.725639472849439e-06,
+      "loss": 0.3211,
+      "step": 32758
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7250627216887276e-06,
+      "loss": 0.3915,
+      "step": 32759
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7244859705280159e-06,
+      "loss": 0.4442,
+      "step": 32760
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.723909219367304e-06,
+      "loss": 0.2685,
+      "step": 32761
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7233324682065925e-06,
+      "loss": 0.3334,
+      "step": 32762
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7227557170458806e-06,
+      "loss": 0.416,
+      "step": 32763
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.722178965885169e-06,
+      "loss": 0.3715,
+      "step": 32764
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7216022147244575e-06,
+      "loss": 0.5504,
+      "step": 32765
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7210254635637456e-06,
+      "loss": 0.3779,
+      "step": 32766
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.720448712403034e-06,
+      "loss": 0.4154,
+      "step": 32767
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.719871961242322e-06,
+      "loss": 0.3373,
+      "step": 32768
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7192952100816103e-06,
+      "loss": 0.4878,
+      "step": 32769
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7187184589208989e-06,
+      "loss": 0.3878,
+      "step": 32770
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.718141707760187e-06,
+      "loss": 0.3807,
+      "step": 32771
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7175649565994753e-06,
+      "loss": 0.3082,
+      "step": 32772
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7169882054387634e-06,
+      "loss": 0.3884,
+      "step": 32773
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.716411454278052e-06,
+      "loss": 0.3771,
+      "step": 32774
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7158347031173403e-06,
+      "loss": 0.4945,
+      "step": 32775
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7152579519566284e-06,
+      "loss": 0.4283,
+      "step": 32776
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.714681200795917e-06,
+      "loss": 0.4324,
+      "step": 32777
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.714104449635205e-06,
+      "loss": 0.2737,
+      "step": 32778
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7135276984744933e-06,
+      "loss": 0.3305,
+      "step": 32779
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7129509473137814e-06,
+      "loss": 0.4343,
+      "step": 32780
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.71237419615307e-06,
+      "loss": 0.362,
+      "step": 32781
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7117974449923583e-06,
+      "loss": 0.3454,
+      "step": 32782
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7112206938316464e-06,
+      "loss": 0.3709,
+      "step": 32783
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7106439426709347e-06,
+      "loss": 0.3781,
+      "step": 32784
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7100671915102228e-06,
+      "loss": 0.3297,
+      "step": 32785
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7094904403495114e-06,
+      "loss": 0.2871,
+      "step": 32786
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7089136891887997e-06,
+      "loss": 0.3807,
+      "step": 32787
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7083369380280878e-06,
+      "loss": 0.3951,
+      "step": 32788
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7077601868673763e-06,
+      "loss": 0.414,
+      "step": 32789
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7071834357066644e-06,
+      "loss": 0.4225,
+      "step": 32790
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7066066845459528e-06,
+      "loss": 0.4302,
+      "step": 32791
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7060299333852413e-06,
+      "loss": 0.4062,
+      "step": 32792
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7054531822245294e-06,
+      "loss": 0.3459,
+      "step": 32793
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7048764310638177e-06,
+      "loss": 0.4422,
+      "step": 32794
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7042996799031058e-06,
+      "loss": 0.3757,
+      "step": 32795
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7037229287423944e-06,
+      "loss": 0.3777,
+      "step": 32796
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7031461775816827e-06,
+      "loss": 0.4923,
+      "step": 32797
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7025694264209708e-06,
+      "loss": 0.502,
+      "step": 32798
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.701992675260259e-06,
+      "loss": 0.43,
+      "step": 32799
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 1.7014159240995472e-06,
+      "loss": 0.3126,
+      "step": 32800
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.7008391729388357e-06,
+      "loss": 0.402,
+      "step": 32801
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.7002624217781239e-06,
+      "loss": 0.4772,
+      "step": 32802
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6996856706174122e-06,
+      "loss": 0.3889,
+      "step": 32803
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6991089194567007e-06,
+      "loss": 0.3831,
+      "step": 32804
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6985321682959888e-06,
+      "loss": 0.338,
+      "step": 32805
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6979554171352771e-06,
+      "loss": 0.4513,
+      "step": 32806
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6973786659745652e-06,
+      "loss": 0.4376,
+      "step": 32807
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6968019148138538e-06,
+      "loss": 0.4108,
+      "step": 32808
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.696225163653142e-06,
+      "loss": 0.5235,
+      "step": 32809
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6956484124924302e-06,
+      "loss": 0.3245,
+      "step": 32810
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6950716613317187e-06,
+      "loss": 0.3623,
+      "step": 32811
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6944949101710068e-06,
+      "loss": 0.2938,
+      "step": 32812
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6939181590102952e-06,
+      "loss": 0.4058,
+      "step": 32813
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6933414078495835e-06,
+      "loss": 0.4948,
+      "step": 32814
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6927646566888716e-06,
+      "loss": 0.4734,
+      "step": 32815
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6921879055281601e-06,
+      "loss": 0.3831,
+      "step": 32816
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6916111543674482e-06,
+      "loss": 0.3759,
+      "step": 32817
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6910344032067366e-06,
+      "loss": 0.4224,
+      "step": 32818
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6904576520460247e-06,
+      "loss": 0.4885,
+      "step": 32819
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6898809008853132e-06,
+      "loss": 0.309,
+      "step": 32820
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6893041497246015e-06,
+      "loss": 0.3339,
+      "step": 32821
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6887273985638896e-06,
+      "loss": 0.5525,
+      "step": 32822
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6881506474031782e-06,
+      "loss": 0.4314,
+      "step": 32823
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6875738962424663e-06,
+      "loss": 0.493,
+      "step": 32824
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6869971450817546e-06,
+      "loss": 0.3053,
+      "step": 32825
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.686420393921043e-06,
+      "loss": 0.4023,
+      "step": 32826
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6858436427603312e-06,
+      "loss": 0.4614,
+      "step": 32827
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6852668915996195e-06,
+      "loss": 0.3733,
+      "step": 32828
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6846901404389077e-06,
+      "loss": 0.4152,
+      "step": 32829
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.684113389278196e-06,
+      "loss": 0.4499,
+      "step": 32830
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6835366381174845e-06,
+      "loss": 0.4113,
+      "step": 32831
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6829598869567726e-06,
+      "loss": 0.3469,
+      "step": 32832
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.682383135796061e-06,
+      "loss": 0.4534,
+      "step": 32833
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.681806384635349e-06,
+      "loss": 0.4424,
+      "step": 32834
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6812296334746376e-06,
+      "loss": 0.3772,
+      "step": 32835
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.680652882313926e-06,
+      "loss": 0.3946,
+      "step": 32836
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.680076131153214e-06,
+      "loss": 0.4031,
+      "step": 32837
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6794993799925025e-06,
+      "loss": 0.4186,
+      "step": 32838
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6789226288317906e-06,
+      "loss": 0.4208,
+      "step": 32839
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.678345877671079e-06,
+      "loss": 0.3268,
+      "step": 32840
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.677769126510367e-06,
+      "loss": 0.3373,
+      "step": 32841
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6771923753496556e-06,
+      "loss": 0.3744,
+      "step": 32842
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.676615624188944e-06,
+      "loss": 0.44,
+      "step": 32843
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.676038873028232e-06,
+      "loss": 0.3224,
+      "step": 32844
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6754621218675204e-06,
+      "loss": 0.379,
+      "step": 32845
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6748853707068085e-06,
+      "loss": 0.5225,
+      "step": 32846
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.674308619546097e-06,
+      "loss": 0.2948,
+      "step": 32847
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6737318683853853e-06,
+      "loss": 0.493,
+      "step": 32848
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6731551172246734e-06,
+      "loss": 0.4211,
+      "step": 32849
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.672578366063962e-06,
+      "loss": 0.3889,
+      "step": 32850
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.67200161490325e-06,
+      "loss": 0.5626,
+      "step": 32851
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6714248637425384e-06,
+      "loss": 0.4711,
+      "step": 32852
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.670848112581827e-06,
+      "loss": 0.2904,
+      "step": 32853
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.670271361421115e-06,
+      "loss": 0.4106,
+      "step": 32854
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6696946102604034e-06,
+      "loss": 0.3313,
+      "step": 32855
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6691178590996915e-06,
+      "loss": 0.3533,
+      "step": 32856
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.66854110793898e-06,
+      "loss": 0.2622,
+      "step": 32857
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6679643567782683e-06,
+      "loss": 0.3703,
+      "step": 32858
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6673876056175564e-06,
+      "loss": 0.338,
+      "step": 32859
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6668108544568447e-06,
+      "loss": 0.4194,
+      "step": 32860
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6662341032961328e-06,
+      "loss": 0.3148,
+      "step": 32861
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6656573521354214e-06,
+      "loss": 0.3689,
+      "step": 32862
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6650806009747095e-06,
+      "loss": 0.4078,
+      "step": 32863
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6645038498139978e-06,
+      "loss": 0.4164,
+      "step": 32864
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6639270986532863e-06,
+      "loss": 0.2769,
+      "step": 32865
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6633503474925745e-06,
+      "loss": 0.492,
+      "step": 32866
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6627735963318628e-06,
+      "loss": 0.4365,
+      "step": 32867
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6621968451711509e-06,
+      "loss": 0.4347,
+      "step": 32868
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6616200940104394e-06,
+      "loss": 0.2453,
+      "step": 32869
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6610433428497277e-06,
+      "loss": 0.4923,
+      "step": 32870
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6604665916890158e-06,
+      "loss": 0.3646,
+      "step": 32871
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6598898405283042e-06,
+      "loss": 0.4525,
+      "step": 32872
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6593130893675925e-06,
+      "loss": 0.4381,
+      "step": 32873
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6587363382068808e-06,
+      "loss": 0.419,
+      "step": 32874
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6581595870461691e-06,
+      "loss": 0.3335,
+      "step": 32875
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6575828358854572e-06,
+      "loss": 0.4595,
+      "step": 32876
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6570060847247458e-06,
+      "loss": 0.3898,
+      "step": 32877
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6564293335640339e-06,
+      "loss": 0.4274,
+      "step": 32878
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6558525824033222e-06,
+      "loss": 0.5468,
+      "step": 32879
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6552758312426103e-06,
+      "loss": 0.3068,
+      "step": 32880
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6546990800818988e-06,
+      "loss": 0.3016,
+      "step": 32881
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6541223289211872e-06,
+      "loss": 0.3855,
+      "step": 32882
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6535455777604753e-06,
+      "loss": 0.3417,
+      "step": 32883
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6529688265997638e-06,
+      "loss": 0.3528,
+      "step": 32884
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.652392075439052e-06,
+      "loss": 0.4823,
+      "step": 32885
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6518153242783402e-06,
+      "loss": 0.4704,
+      "step": 32886
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6512385731176285e-06,
+      "loss": 0.3591,
+      "step": 32887
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6506618219569169e-06,
+      "loss": 0.3923,
+      "step": 32888
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6500850707962052e-06,
+      "loss": 0.4339,
+      "step": 32889
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6495083196354933e-06,
+      "loss": 0.3833,
+      "step": 32890
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6489315684747816e-06,
+      "loss": 0.4518,
+      "step": 32891
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6483548173140701e-06,
+      "loss": 0.4525,
+      "step": 32892
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6477780661533583e-06,
+      "loss": 0.4641,
+      "step": 32893
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6472013149926466e-06,
+      "loss": 0.4272,
+      "step": 32894
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6466245638319347e-06,
+      "loss": 0.4579,
+      "step": 32895
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6460478126712232e-06,
+      "loss": 0.3421,
+      "step": 32896
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6454710615105115e-06,
+      "loss": 0.3453,
+      "step": 32897
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6448943103497996e-06,
+      "loss": 0.4637,
+      "step": 32898
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6443175591890882e-06,
+      "loss": 0.4388,
+      "step": 32899
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6437408080283763e-06,
+      "loss": 0.3308,
+      "step": 32900
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6431640568676646e-06,
+      "loss": 0.3056,
+      "step": 32901
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6425873057069527e-06,
+      "loss": 0.4173,
+      "step": 32902
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6420105545462412e-06,
+      "loss": 0.3563,
+      "step": 32903
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6414338033855296e-06,
+      "loss": 0.3948,
+      "step": 32904
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6408570522248177e-06,
+      "loss": 0.339,
+      "step": 32905
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.640280301064106e-06,
+      "loss": 0.3037,
+      "step": 32906
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.639703549903394e-06,
+      "loss": 0.3537,
+      "step": 32907
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6391267987426826e-06,
+      "loss": 0.4394,
+      "step": 32908
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.638550047581971e-06,
+      "loss": 0.4178,
+      "step": 32909
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.637973296421259e-06,
+      "loss": 0.4641,
+      "step": 32910
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6373965452605476e-06,
+      "loss": 0.3827,
+      "step": 32911
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6368197940998357e-06,
+      "loss": 0.3412,
+      "step": 32912
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.636243042939124e-06,
+      "loss": 0.3972,
+      "step": 32913
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6356662917784126e-06,
+      "loss": 0.4304,
+      "step": 32914
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6350895406177007e-06,
+      "loss": 0.3537,
+      "step": 32915
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.634512789456989e-06,
+      "loss": 0.4078,
+      "step": 32916
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.633936038296277e-06,
+      "loss": 0.431,
+      "step": 32917
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6333592871355654e-06,
+      "loss": 0.4532,
+      "step": 32918
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.632782535974854e-06,
+      "loss": 0.4619,
+      "step": 32919
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.632205784814142e-06,
+      "loss": 0.3021,
+      "step": 32920
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6316290336534304e-06,
+      "loss": 0.4786,
+      "step": 32921
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6310522824927185e-06,
+      "loss": 0.356,
+      "step": 32922
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.630475531332007e-06,
+      "loss": 0.3123,
+      "step": 32923
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6298987801712951e-06,
+      "loss": 0.3012,
+      "step": 32924
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6293220290105834e-06,
+      "loss": 0.3884,
+      "step": 32925
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.628745277849872e-06,
+      "loss": 0.4451,
+      "step": 32926
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.62816852668916e-06,
+      "loss": 0.4066,
+      "step": 32927
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6275917755284484e-06,
+      "loss": 0.4227,
+      "step": 32928
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6270150243677365e-06,
+      "loss": 0.4424,
+      "step": 32929
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.626438273207025e-06,
+      "loss": 0.4058,
+      "step": 32930
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6258615220463134e-06,
+      "loss": 0.3884,
+      "step": 32931
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6252847708856015e-06,
+      "loss": 0.404,
+      "step": 32932
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6247080197248898e-06,
+      "loss": 0.2526,
+      "step": 32933
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6241312685641781e-06,
+      "loss": 0.4277,
+      "step": 32934
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6235545174034664e-06,
+      "loss": 0.4102,
+      "step": 32935
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6229777662427548e-06,
+      "loss": 0.3342,
+      "step": 32936
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6224010150820429e-06,
+      "loss": 0.4375,
+      "step": 32937
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6218242639213314e-06,
+      "loss": 0.2708,
+      "step": 32938
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6212475127606195e-06,
+      "loss": 0.3023,
+      "step": 32939
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6206707615999078e-06,
+      "loss": 0.2897,
+      "step": 32940
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.620094010439196e-06,
+      "loss": 0.367,
+      "step": 32941
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6195172592784845e-06,
+      "loss": 0.4652,
+      "step": 32942
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6189405081177728e-06,
+      "loss": 0.3594,
+      "step": 32943
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.618363756957061e-06,
+      "loss": 0.4318,
+      "step": 32944
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6177870057963494e-06,
+      "loss": 0.3706,
+      "step": 32945
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6172102546356375e-06,
+      "loss": 0.5096,
+      "step": 32946
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6166335034749259e-06,
+      "loss": 0.313,
+      "step": 32947
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6160567523142142e-06,
+      "loss": 0.3105,
+      "step": 32948
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6154800011535025e-06,
+      "loss": 0.3835,
+      "step": 32949
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6149032499927908e-06,
+      "loss": 0.3785,
+      "step": 32950
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.614326498832079e-06,
+      "loss": 0.4272,
+      "step": 32951
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6137497476713673e-06,
+      "loss": 0.3882,
+      "step": 32952
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6131729965106558e-06,
+      "loss": 0.4786,
+      "step": 32953
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6125962453499439e-06,
+      "loss": 0.3981,
+      "step": 32954
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6120194941892322e-06,
+      "loss": 0.492,
+      "step": 32955
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6114427430285203e-06,
+      "loss": 0.3272,
+      "step": 32956
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6108659918678089e-06,
+      "loss": 0.3709,
+      "step": 32957
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6102892407070972e-06,
+      "loss": 0.3878,
+      "step": 32958
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6097124895463853e-06,
+      "loss": 0.3957,
+      "step": 32959
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6091357383856738e-06,
+      "loss": 0.3547,
+      "step": 32960
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.608558987224962e-06,
+      "loss": 0.3186,
+      "step": 32961
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6079822360642502e-06,
+      "loss": 0.4846,
+      "step": 32962
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6074054849035384e-06,
+      "loss": 0.3805,
+      "step": 32963
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6068287337428267e-06,
+      "loss": 0.4139,
+      "step": 32964
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6062519825821152e-06,
+      "loss": 0.3862,
+      "step": 32965
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6056752314214033e-06,
+      "loss": 0.383,
+      "step": 32966
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6050984802606916e-06,
+      "loss": 0.3592,
+      "step": 32967
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6045217290999797e-06,
+      "loss": 0.3761,
+      "step": 32968
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6039449779392683e-06,
+      "loss": 0.3804,
+      "step": 32969
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6033682267785566e-06,
+      "loss": 0.385,
+      "step": 32970
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6027914756178447e-06,
+      "loss": 0.4079,
+      "step": 32971
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6022147244571332e-06,
+      "loss": 0.4268,
+      "step": 32972
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6016379732964213e-06,
+      "loss": 0.4355,
+      "step": 32973
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6010612221357097e-06,
+      "loss": 0.4551,
+      "step": 32974
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.6004844709749982e-06,
+      "loss": 0.45,
+      "step": 32975
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.5999077198142863e-06,
+      "loss": 0.4085,
+      "step": 32976
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.5993309686535746e-06,
+      "loss": 0.3851,
+      "step": 32977
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.5987542174928627e-06,
+      "loss": 0.4279,
+      "step": 32978
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 1.598177466332151e-06,
+      "loss": 0.4396,
+      "step": 32979
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5976007151714394e-06,
+      "loss": 0.3936,
+      "step": 32980
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5970239640107277e-06,
+      "loss": 0.3931,
+      "step": 32981
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.596447212850016e-06,
+      "loss": 0.4387,
+      "step": 32982
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5958704616893041e-06,
+      "loss": 0.317,
+      "step": 32983
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5952937105285927e-06,
+      "loss": 0.4983,
+      "step": 32984
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5947169593678808e-06,
+      "loss": 0.4055,
+      "step": 32985
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.594140208207169e-06,
+      "loss": 0.5314,
+      "step": 32986
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5935634570464576e-06,
+      "loss": 0.3577,
+      "step": 32987
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5929867058857457e-06,
+      "loss": 0.4243,
+      "step": 32988
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.592409954725034e-06,
+      "loss": 0.3195,
+      "step": 32989
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5918332035643222e-06,
+      "loss": 0.4059,
+      "step": 32990
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5912564524036107e-06,
+      "loss": 0.2908,
+      "step": 32991
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.590679701242899e-06,
+      "loss": 0.3993,
+      "step": 32992
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5901029500821871e-06,
+      "loss": 0.3316,
+      "step": 32993
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5895261989214754e-06,
+      "loss": 0.5365,
+      "step": 32994
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5889494477607638e-06,
+      "loss": 0.4123,
+      "step": 32995
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.588372696600052e-06,
+      "loss": 0.4277,
+      "step": 32996
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5877959454393404e-06,
+      "loss": 0.3903,
+      "step": 32997
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5872191942786285e-06,
+      "loss": 0.4907,
+      "step": 32998
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.586642443117917e-06,
+      "loss": 0.461,
+      "step": 32999
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5860656919572051e-06,
+      "loss": 0.3476,
+      "step": 33000
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5854889407964935e-06,
+      "loss": 0.3686,
+      "step": 33001
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5849121896357816e-06,
+      "loss": 0.4377,
+      "step": 33002
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5843354384750701e-06,
+      "loss": 0.3504,
+      "step": 33003
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5837586873143584e-06,
+      "loss": 0.2361,
+      "step": 33004
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5831819361536465e-06,
+      "loss": 0.4301,
+      "step": 33005
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.582605184992935e-06,
+      "loss": 0.3733,
+      "step": 33006
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5820284338322232e-06,
+      "loss": 0.2701,
+      "step": 33007
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5814516826715115e-06,
+      "loss": 0.3131,
+      "step": 33008
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5808749315107998e-06,
+      "loss": 0.3363,
+      "step": 33009
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.580298180350088e-06,
+      "loss": 0.4785,
+      "step": 33010
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5797214291893765e-06,
+      "loss": 0.3563,
+      "step": 33011
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5791446780286646e-06,
+      "loss": 0.4966,
+      "step": 33012
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5785679268679529e-06,
+      "loss": 0.3999,
+      "step": 33013
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5779911757072414e-06,
+      "loss": 0.3391,
+      "step": 33014
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5774144245465295e-06,
+      "loss": 0.3947,
+      "step": 33015
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5768376733858178e-06,
+      "loss": 0.3653,
+      "step": 33016
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.576260922225106e-06,
+      "loss": 0.3951,
+      "step": 33017
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5756841710643945e-06,
+      "loss": 0.5214,
+      "step": 33018
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5751074199036828e-06,
+      "loss": 0.2657,
+      "step": 33019
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.574530668742971e-06,
+      "loss": 0.3897,
+      "step": 33020
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5739539175822595e-06,
+      "loss": 0.5471,
+      "step": 33021
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5733771664215476e-06,
+      "loss": 0.2821,
+      "step": 33022
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5728004152608359e-06,
+      "loss": 0.4135,
+      "step": 33023
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.572223664100124e-06,
+      "loss": 0.3751,
+      "step": 33024
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5716469129394123e-06,
+      "loss": 0.3247,
+      "step": 33025
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5710701617787008e-06,
+      "loss": 0.3647,
+      "step": 33026
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.570493410617989e-06,
+      "loss": 0.4671,
+      "step": 33027
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5699166594572773e-06,
+      "loss": 0.5257,
+      "step": 33028
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5693399082965654e-06,
+      "loss": 0.4959,
+      "step": 33029
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.568763157135854e-06,
+      "loss": 0.5426,
+      "step": 33030
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5681864059751422e-06,
+      "loss": 0.4284,
+      "step": 33031
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5676096548144303e-06,
+      "loss": 0.4322,
+      "step": 33032
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5670329036537189e-06,
+      "loss": 0.3068,
+      "step": 33033
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.566456152493007e-06,
+      "loss": 0.2817,
+      "step": 33034
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5658794013322953e-06,
+      "loss": 0.4002,
+      "step": 33035
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5653026501715836e-06,
+      "loss": 0.328,
+      "step": 33036
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.564725899010872e-06,
+      "loss": 0.3458,
+      "step": 33037
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5641491478501603e-06,
+      "loss": 0.3787,
+      "step": 33038
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5635723966894484e-06,
+      "loss": 0.4161,
+      "step": 33039
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5629956455287367e-06,
+      "loss": 0.4244,
+      "step": 33040
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.562418894368025e-06,
+      "loss": 0.3833,
+      "step": 33041
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5618421432073133e-06,
+      "loss": 0.4106,
+      "step": 33042
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5612653920466017e-06,
+      "loss": 0.3875,
+      "step": 33043
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5606886408858898e-06,
+      "loss": 0.5433,
+      "step": 33044
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5601118897251783e-06,
+      "loss": 0.4183,
+      "step": 33045
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5595351385644664e-06,
+      "loss": 0.3923,
+      "step": 33046
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5589583874037547e-06,
+      "loss": 0.4117,
+      "step": 33047
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5583816362430433e-06,
+      "loss": 0.4759,
+      "step": 33048
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5578048850823314e-06,
+      "loss": 0.4118,
+      "step": 33049
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5572281339216197e-06,
+      "loss": 0.4006,
+      "step": 33050
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5566513827609078e-06,
+      "loss": 0.3373,
+      "step": 33051
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5560746316001963e-06,
+      "loss": 0.4694,
+      "step": 33052
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5554978804394846e-06,
+      "loss": 0.5175,
+      "step": 33053
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5549211292787728e-06,
+      "loss": 0.3524,
+      "step": 33054
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.554344378118061e-06,
+      "loss": 0.2685,
+      "step": 33055
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5537676269573492e-06,
+      "loss": 0.344,
+      "step": 33056
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5531908757966377e-06,
+      "loss": 0.3124,
+      "step": 33057
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.552614124635926e-06,
+      "loss": 0.3968,
+      "step": 33058
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5520373734752141e-06,
+      "loss": 0.3544,
+      "step": 33059
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5514606223145027e-06,
+      "loss": 0.4738,
+      "step": 33060
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5508838711537908e-06,
+      "loss": 0.2489,
+      "step": 33061
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.550307119993079e-06,
+      "loss": 0.4053,
+      "step": 33062
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5497303688323672e-06,
+      "loss": 0.3897,
+      "step": 33063
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5491536176716557e-06,
+      "loss": 0.4149,
+      "step": 33064
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.548576866510944e-06,
+      "loss": 0.386,
+      "step": 33065
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5480001153502322e-06,
+      "loss": 0.3935,
+      "step": 33066
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5474233641895207e-06,
+      "loss": 0.3656,
+      "step": 33067
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5468466130288088e-06,
+      "loss": 0.3671,
+      "step": 33068
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5462698618680971e-06,
+      "loss": 0.272,
+      "step": 33069
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5456931107073855e-06,
+      "loss": 0.4789,
+      "step": 33070
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5451163595466736e-06,
+      "loss": 0.4421,
+      "step": 33071
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.544539608385962e-06,
+      "loss": 0.4492,
+      "step": 33072
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5439628572252502e-06,
+      "loss": 0.3999,
+      "step": 33073
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5433861060645385e-06,
+      "loss": 0.3808,
+      "step": 33074
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.542809354903827e-06,
+      "loss": 0.4294,
+      "step": 33075
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5422326037431152e-06,
+      "loss": 0.4123,
+      "step": 33076
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5416558525824035e-06,
+      "loss": 0.3223,
+      "step": 33077
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5410791014216916e-06,
+      "loss": 0.5109,
+      "step": 33078
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5405023502609801e-06,
+      "loss": 0.4666,
+      "step": 33079
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5399255991002684e-06,
+      "loss": 0.4733,
+      "step": 33080
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5393488479395566e-06,
+      "loss": 0.4366,
+      "step": 33081
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5387720967788449e-06,
+      "loss": 0.4559,
+      "step": 33082
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5381953456181332e-06,
+      "loss": 0.4287,
+      "step": 33083
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5376185944574215e-06,
+      "loss": 0.5535,
+      "step": 33084
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5370418432967096e-06,
+      "loss": 0.3009,
+      "step": 33085
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.536465092135998e-06,
+      "loss": 0.3223,
+      "step": 33086
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5358883409752865e-06,
+      "loss": 0.2841,
+      "step": 33087
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5353115898145746e-06,
+      "loss": 0.3543,
+      "step": 33088
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.534734838653863e-06,
+      "loss": 0.3336,
+      "step": 33089
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.534158087493151e-06,
+      "loss": 0.4526,
+      "step": 33090
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5335813363324395e-06,
+      "loss": 0.3189,
+      "step": 33091
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5330045851717279e-06,
+      "loss": 0.3885,
+      "step": 33092
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.532427834011016e-06,
+      "loss": 0.3793,
+      "step": 33093
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5318510828503045e-06,
+      "loss": 0.3274,
+      "step": 33094
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5312743316895926e-06,
+      "loss": 0.4362,
+      "step": 33095
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.530697580528881e-06,
+      "loss": 0.3372,
+      "step": 33096
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5301208293681693e-06,
+      "loss": 0.2662,
+      "step": 33097
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5295440782074576e-06,
+      "loss": 0.4865,
+      "step": 33098
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.528967327046746e-06,
+      "loss": 0.3413,
+      "step": 33099
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.528390575886034e-06,
+      "loss": 0.3126,
+      "step": 33100
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5278138247253223e-06,
+      "loss": 0.4686,
+      "step": 33101
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5272370735646104e-06,
+      "loss": 0.4277,
+      "step": 33102
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.526660322403899e-06,
+      "loss": 0.3634,
+      "step": 33103
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5260835712431873e-06,
+      "loss": 0.4143,
+      "step": 33104
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5255068200824754e-06,
+      "loss": 0.3563,
+      "step": 33105
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.524930068921764e-06,
+      "loss": 0.4594,
+      "step": 33106
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.524353317761052e-06,
+      "loss": 0.38,
+      "step": 33107
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5237765666003404e-06,
+      "loss": 0.5069,
+      "step": 33108
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5231998154396289e-06,
+      "loss": 0.4185,
+      "step": 33109
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.522623064278917e-06,
+      "loss": 0.4183,
+      "step": 33110
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5220463131182053e-06,
+      "loss": 0.3611,
+      "step": 33111
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5214695619574934e-06,
+      "loss": 0.3542,
+      "step": 33112
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.520892810796782e-06,
+      "loss": 0.3985,
+      "step": 33113
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5203160596360703e-06,
+      "loss": 0.3523,
+      "step": 33114
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5197393084753584e-06,
+      "loss": 0.4591,
+      "step": 33115
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5191625573146467e-06,
+      "loss": 0.3201,
+      "step": 33116
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5185858061539348e-06,
+      "loss": 0.3232,
+      "step": 33117
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5180090549932234e-06,
+      "loss": 0.4874,
+      "step": 33118
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5174323038325117e-06,
+      "loss": 0.5387,
+      "step": 33119
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5168555526717998e-06,
+      "loss": 0.3923,
+      "step": 33120
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5162788015110883e-06,
+      "loss": 0.4428,
+      "step": 33121
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5157020503503764e-06,
+      "loss": 0.3836,
+      "step": 33122
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5151252991896647e-06,
+      "loss": 0.5282,
+      "step": 33123
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5145485480289528e-06,
+      "loss": 0.4624,
+      "step": 33124
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5139717968682414e-06,
+      "loss": 0.3655,
+      "step": 33125
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5133950457075297e-06,
+      "loss": 0.4042,
+      "step": 33126
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5128182945468178e-06,
+      "loss": 0.3171,
+      "step": 33127
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5122415433861061e-06,
+      "loss": 0.2744,
+      "step": 33128
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5116647922253945e-06,
+      "loss": 0.4309,
+      "step": 33129
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5110880410646828e-06,
+      "loss": 0.3181,
+      "step": 33130
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.510511289903971e-06,
+      "loss": 0.5126,
+      "step": 33131
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5099345387432592e-06,
+      "loss": 0.453,
+      "step": 33132
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5093577875825477e-06,
+      "loss": 0.4394,
+      "step": 33133
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5087810364218358e-06,
+      "loss": 0.2665,
+      "step": 33134
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5082042852611242e-06,
+      "loss": 0.3507,
+      "step": 33135
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5076275341004127e-06,
+      "loss": 0.4323,
+      "step": 33136
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5070507829397008e-06,
+      "loss": 0.3968,
+      "step": 33137
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5064740317789891e-06,
+      "loss": 0.4376,
+      "step": 33138
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5058972806182772e-06,
+      "loss": 0.5291,
+      "step": 33139
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5053205294575658e-06,
+      "loss": 0.3563,
+      "step": 33140
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5047437782968539e-06,
+      "loss": 0.4965,
+      "step": 33141
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5041670271361422e-06,
+      "loss": 0.4047,
+      "step": 33142
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5035902759754305e-06,
+      "loss": 0.4055,
+      "step": 33143
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5030135248147188e-06,
+      "loss": 0.3136,
+      "step": 33144
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5024367736540072e-06,
+      "loss": 0.3581,
+      "step": 33145
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5018600224932953e-06,
+      "loss": 0.3174,
+      "step": 33146
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5012832713325836e-06,
+      "loss": 0.4781,
+      "step": 33147
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5007065201718721e-06,
+      "loss": 0.5023,
+      "step": 33148
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.5001297690111602e-06,
+      "loss": 0.3687,
+      "step": 33149
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4995530178504485e-06,
+      "loss": 0.2674,
+      "step": 33150
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4989762666897367e-06,
+      "loss": 0.3657,
+      "step": 33151
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4983995155290252e-06,
+      "loss": 0.2551,
+      "step": 33152
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4978227643683135e-06,
+      "loss": 0.4437,
+      "step": 33153
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4972460132076016e-06,
+      "loss": 0.3739,
+      "step": 33154
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4966692620468901e-06,
+      "loss": 0.4611,
+      "step": 33155
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4960925108861783e-06,
+      "loss": 0.3001,
+      "step": 33156
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.4955157597254666e-06,
+      "loss": 0.3026,
+      "step": 33157
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.494939008564755e-06,
+      "loss": 0.297,
+      "step": 33158
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4943622574040432e-06,
+      "loss": 0.3311,
+      "step": 33159
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4937855062433315e-06,
+      "loss": 0.472,
+      "step": 33160
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4932087550826196e-06,
+      "loss": 0.4104,
+      "step": 33161
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.492632003921908e-06,
+      "loss": 0.5596,
+      "step": 33162
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.492055252761196e-06,
+      "loss": 0.2137,
+      "step": 33163
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4914785016004846e-06,
+      "loss": 0.3425,
+      "step": 33164
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.490901750439773e-06,
+      "loss": 0.3497,
+      "step": 33165
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.490324999279061e-06,
+      "loss": 0.3551,
+      "step": 33166
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4897482481183496e-06,
+      "loss": 0.342,
+      "step": 33167
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4891714969576377e-06,
+      "loss": 0.3696,
+      "step": 33168
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.488594745796926e-06,
+      "loss": 0.3856,
+      "step": 33169
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4880179946362145e-06,
+      "loss": 0.3353,
+      "step": 33170
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4874412434755026e-06,
+      "loss": 0.4468,
+      "step": 33171
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.486864492314791e-06,
+      "loss": 0.2862,
+      "step": 33172
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.486287741154079e-06,
+      "loss": 0.3033,
+      "step": 33173
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4857109899933674e-06,
+      "loss": 0.3607,
+      "step": 33174
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.485134238832656e-06,
+      "loss": 0.3145,
+      "step": 33175
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.484557487671944e-06,
+      "loss": 0.4465,
+      "step": 33176
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4839807365112323e-06,
+      "loss": 0.4277,
+      "step": 33177
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4834039853505205e-06,
+      "loss": 0.5383,
+      "step": 33178
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.482827234189809e-06,
+      "loss": 0.2369,
+      "step": 33179
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4822504830290973e-06,
+      "loss": 0.2758,
+      "step": 33180
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4816737318683854e-06,
+      "loss": 0.3244,
+      "step": 33181
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.481096980707674e-06,
+      "loss": 0.4443,
+      "step": 33182
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.480520229546962e-06,
+      "loss": 0.3041,
+      "step": 33183
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4799434783862504e-06,
+      "loss": 0.524,
+      "step": 33184
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4793667272255385e-06,
+      "loss": 0.3996,
+      "step": 33185
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.478789976064827e-06,
+      "loss": 0.3665,
+      "step": 33186
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4782132249041153e-06,
+      "loss": 0.3152,
+      "step": 33187
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4776364737434034e-06,
+      "loss": 0.3353,
+      "step": 33188
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4770597225826918e-06,
+      "loss": 0.4097,
+      "step": 33189
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.47648297142198e-06,
+      "loss": 0.3757,
+      "step": 33190
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4759062202612684e-06,
+      "loss": 0.3231,
+      "step": 33191
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4753294691005567e-06,
+      "loss": 0.2534,
+      "step": 33192
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4747527179398448e-06,
+      "loss": 0.383,
+      "step": 33193
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4741759667791334e-06,
+      "loss": 0.5766,
+      "step": 33194
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4735992156184215e-06,
+      "loss": 0.4888,
+      "step": 33195
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4730224644577098e-06,
+      "loss": 0.4545,
+      "step": 33196
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4724457132969983e-06,
+      "loss": 0.3117,
+      "step": 33197
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4718689621362864e-06,
+      "loss": 0.4329,
+      "step": 33198
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4712922109755748e-06,
+      "loss": 0.3463,
+      "step": 33199
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4707154598148629e-06,
+      "loss": 0.4334,
+      "step": 33200
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4701387086541514e-06,
+      "loss": 0.3521,
+      "step": 33201
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4695619574934395e-06,
+      "loss": 0.3717,
+      "step": 33202
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4689852063327278e-06,
+      "loss": 0.3672,
+      "step": 33203
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4684084551720162e-06,
+      "loss": 0.4217,
+      "step": 33204
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4678317040113045e-06,
+      "loss": 0.4668,
+      "step": 33205
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4672549528505928e-06,
+      "loss": 0.346,
+      "step": 33206
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.466678201689881e-06,
+      "loss": 0.4721,
+      "step": 33207
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4661014505291692e-06,
+      "loss": 0.4527,
+      "step": 33208
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4655246993684578e-06,
+      "loss": 0.4422,
+      "step": 33209
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4649479482077459e-06,
+      "loss": 0.4352,
+      "step": 33210
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4643711970470342e-06,
+      "loss": 0.3625,
+      "step": 33211
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4637944458863223e-06,
+      "loss": 0.2649,
+      "step": 33212
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4632176947256108e-06,
+      "loss": 0.2925,
+      "step": 33213
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4626409435648991e-06,
+      "loss": 0.3388,
+      "step": 33214
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4620641924041873e-06,
+      "loss": 0.4443,
+      "step": 33215
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4614874412434758e-06,
+      "loss": 0.4598,
+      "step": 33216
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4609106900827639e-06,
+      "loss": 0.4377,
+      "step": 33217
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4603339389220522e-06,
+      "loss": 0.3365,
+      "step": 33218
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4597571877613405e-06,
+      "loss": 0.3519,
+      "step": 33219
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4591804366006286e-06,
+      "loss": 0.4978,
+      "step": 33220
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4586036854399172e-06,
+      "loss": 0.4377,
+      "step": 33221
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4580269342792053e-06,
+      "loss": 0.3801,
+      "step": 33222
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4574501831184936e-06,
+      "loss": 0.4639,
+      "step": 33223
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4568734319577817e-06,
+      "loss": 0.3806,
+      "step": 33224
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4562966807970702e-06,
+      "loss": 0.4436,
+      "step": 33225
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4557199296363586e-06,
+      "loss": 0.432,
+      "step": 33226
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4551431784756467e-06,
+      "loss": 0.3517,
+      "step": 33227
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4545664273149352e-06,
+      "loss": 0.5119,
+      "step": 33228
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4539896761542233e-06,
+      "loss": 0.2709,
+      "step": 33229
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4534129249935116e-06,
+      "loss": 0.3798,
+      "step": 33230
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4528361738328002e-06,
+      "loss": 0.4311,
+      "step": 33231
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4522594226720883e-06,
+      "loss": 0.3895,
+      "step": 33232
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4516826715113766e-06,
+      "loss": 0.5155,
+      "step": 33233
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4511059203506647e-06,
+      "loss": 0.4248,
+      "step": 33234
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.450529169189953e-06,
+      "loss": 0.513,
+      "step": 33235
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4499524180292416e-06,
+      "loss": 0.3362,
+      "step": 33236
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4493756668685297e-06,
+      "loss": 0.3618,
+      "step": 33237
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.448798915707818e-06,
+      "loss": 0.3693,
+      "step": 33238
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.448222164547106e-06,
+      "loss": 0.4399,
+      "step": 33239
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4476454133863946e-06,
+      "loss": 0.4725,
+      "step": 33240
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.447068662225683e-06,
+      "loss": 0.4244,
+      "step": 33241
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.446491911064971e-06,
+      "loss": 0.3256,
+      "step": 33242
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4459151599042596e-06,
+      "loss": 0.382,
+      "step": 33243
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4453384087435477e-06,
+      "loss": 0.4712,
+      "step": 33244
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.444761657582836e-06,
+      "loss": 0.3136,
+      "step": 33245
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4441849064221241e-06,
+      "loss": 0.2817,
+      "step": 33246
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4436081552614127e-06,
+      "loss": 0.4918,
+      "step": 33247
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.443031404100701e-06,
+      "loss": 0.4341,
+      "step": 33248
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.442454652939989e-06,
+      "loss": 0.4388,
+      "step": 33249
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4418779017792774e-06,
+      "loss": 0.3328,
+      "step": 33250
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4413011506185657e-06,
+      "loss": 0.2918,
+      "step": 33251
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.440724399457854e-06,
+      "loss": 0.4604,
+      "step": 33252
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4401476482971424e-06,
+      "loss": 0.3163,
+      "step": 33253
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4395708971364305e-06,
+      "loss": 0.3522,
+      "step": 33254
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.438994145975719e-06,
+      "loss": 0.3511,
+      "step": 33255
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4384173948150071e-06,
+      "loss": 0.3237,
+      "step": 33256
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4378406436542954e-06,
+      "loss": 0.4521,
+      "step": 33257
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.437263892493584e-06,
+      "loss": 0.2506,
+      "step": 33258
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.436687141332872e-06,
+      "loss": 0.3426,
+      "step": 33259
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4361103901721604e-06,
+      "loss": 0.3752,
+      "step": 33260
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4355336390114485e-06,
+      "loss": 0.4181,
+      "step": 33261
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.434956887850737e-06,
+      "loss": 0.4184,
+      "step": 33262
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4343801366900251e-06,
+      "loss": 0.3734,
+      "step": 33263
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4338033855293135e-06,
+      "loss": 0.4175,
+      "step": 33264
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4332266343686018e-06,
+      "loss": 0.2869,
+      "step": 33265
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.43264988320789e-06,
+      "loss": 0.4325,
+      "step": 33266
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4320731320471784e-06,
+      "loss": 0.3603,
+      "step": 33267
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4314963808864665e-06,
+      "loss": 0.3864,
+      "step": 33268
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4309196297257549e-06,
+      "loss": 0.4577,
+      "step": 33269
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4303428785650434e-06,
+      "loss": 0.366,
+      "step": 33270
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4297661274043315e-06,
+      "loss": 0.4509,
+      "step": 33271
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4291893762436198e-06,
+      "loss": 0.3126,
+      "step": 33272
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.428612625082908e-06,
+      "loss": 0.4359,
+      "step": 33273
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4280358739221965e-06,
+      "loss": 0.406,
+      "step": 33274
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4274591227614848e-06,
+      "loss": 0.4973,
+      "step": 33275
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4268823716007729e-06,
+      "loss": 0.3339,
+      "step": 33276
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4263056204400614e-06,
+      "loss": 0.3143,
+      "step": 33277
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4257288692793495e-06,
+      "loss": 0.3998,
+      "step": 33278
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4251521181186378e-06,
+      "loss": 0.3624,
+      "step": 33279
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4245753669579262e-06,
+      "loss": 0.5227,
+      "step": 33280
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4239986157972143e-06,
+      "loss": 0.374,
+      "step": 33281
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4234218646365028e-06,
+      "loss": 0.3643,
+      "step": 33282
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.422845113475791e-06,
+      "loss": 0.4829,
+      "step": 33283
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4222683623150792e-06,
+      "loss": 0.37,
+      "step": 33284
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4216916111543673e-06,
+      "loss": 0.4209,
+      "step": 33285
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4211148599936559e-06,
+      "loss": 0.3939,
+      "step": 33286
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4205381088329442e-06,
+      "loss": 0.3881,
+      "step": 33287
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4199613576722323e-06,
+      "loss": 0.3722,
+      "step": 33288
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4193846065115208e-06,
+      "loss": 0.4013,
+      "step": 33289
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.418807855350809e-06,
+      "loss": 0.2872,
+      "step": 33290
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4182311041900973e-06,
+      "loss": 0.4205,
+      "step": 33291
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4176543530293858e-06,
+      "loss": 0.4883,
+      "step": 33292
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.417077601868674e-06,
+      "loss": 0.4063,
+      "step": 33293
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4165008507079622e-06,
+      "loss": 0.4245,
+      "step": 33294
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4159240995472503e-06,
+      "loss": 0.283,
+      "step": 33295
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4153473483865387e-06,
+      "loss": 0.3255,
+      "step": 33296
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4147705972258272e-06,
+      "loss": 0.4293,
+      "step": 33297
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4141938460651153e-06,
+      "loss": 0.3783,
+      "step": 33298
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4136170949044036e-06,
+      "loss": 0.427,
+      "step": 33299
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4130403437436917e-06,
+      "loss": 0.3735,
+      "step": 33300
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4124635925829803e-06,
+      "loss": 0.4131,
+      "step": 33301
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4118868414222684e-06,
+      "loss": 0.3605,
+      "step": 33302
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4113100902615567e-06,
+      "loss": 0.5294,
+      "step": 33303
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4107333391008452e-06,
+      "loss": 0.285,
+      "step": 33304
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4101565879401333e-06,
+      "loss": 0.3169,
+      "step": 33305
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4095798367794217e-06,
+      "loss": 0.4071,
+      "step": 33306
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4090030856187098e-06,
+      "loss": 0.4247,
+      "step": 33307
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4084263344579983e-06,
+      "loss": 0.2898,
+      "step": 33308
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4078495832972866e-06,
+      "loss": 0.4008,
+      "step": 33309
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4072728321365747e-06,
+      "loss": 0.4536,
+      "step": 33310
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.406696080975863e-06,
+      "loss": 0.4512,
+      "step": 33311
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4061193298151512e-06,
+      "loss": 0.3623,
+      "step": 33312
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4055425786544397e-06,
+      "loss": 0.3647,
+      "step": 33313
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.404965827493728e-06,
+      "loss": 0.4055,
+      "step": 33314
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4043890763330161e-06,
+      "loss": 0.3974,
+      "step": 33315
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4038123251723046e-06,
+      "loss": 0.4705,
+      "step": 33316
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4032355740115928e-06,
+      "loss": 0.344,
+      "step": 33317
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.402658822850881e-06,
+      "loss": 0.3219,
+      "step": 33318
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4020820716901696e-06,
+      "loss": 0.3497,
+      "step": 33319
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4015053205294577e-06,
+      "loss": 0.3291,
+      "step": 33320
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.400928569368746e-06,
+      "loss": 0.4703,
+      "step": 33321
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.4003518182080341e-06,
+      "loss": 0.2737,
+      "step": 33322
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3997750670473227e-06,
+      "loss": 0.3248,
+      "step": 33323
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3991983158866108e-06,
+      "loss": 0.4041,
+      "step": 33324
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.398621564725899e-06,
+      "loss": 0.4872,
+      "step": 33325
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3980448135651874e-06,
+      "loss": 0.3149,
+      "step": 33326
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3974680624044755e-06,
+      "loss": 0.4176,
+      "step": 33327
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.396891311243764e-06,
+      "loss": 0.3961,
+      "step": 33328
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3963145600830522e-06,
+      "loss": 0.3613,
+      "step": 33329
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3957378089223405e-06,
+      "loss": 0.3492,
+      "step": 33330
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.395161057761629e-06,
+      "loss": 0.4364,
+      "step": 33331
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3945843066009171e-06,
+      "loss": 0.3464,
+      "step": 33332
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3940075554402055e-06,
+      "loss": 0.4073,
+      "step": 33333
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3934308042794936e-06,
+      "loss": 0.3707,
+      "step": 33334
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.392854053118782e-06,
+      "loss": 0.5057,
+      "step": 33335
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 1.3922773019580704e-06,
+      "loss": 0.4203,
+      "step": 33336
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3917005507973585e-06,
+      "loss": 0.3337,
+      "step": 33337
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.391123799636647e-06,
+      "loss": 0.3248,
+      "step": 33338
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3905470484759352e-06,
+      "loss": 0.4476,
+      "step": 33339
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3899702973152235e-06,
+      "loss": 0.3293,
+      "step": 33340
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3893935461545118e-06,
+      "loss": 0.2981,
+      "step": 33341
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3888167949938e-06,
+      "loss": 0.3966,
+      "step": 33342
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3882400438330884e-06,
+      "loss": 0.307,
+      "step": 33343
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3876632926723766e-06,
+      "loss": 0.4913,
+      "step": 33344
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3870865415116649e-06,
+      "loss": 0.278,
+      "step": 33345
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.386509790350953e-06,
+      "loss": 0.3679,
+      "step": 33346
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3859330391902415e-06,
+      "loss": 0.4343,
+      "step": 33347
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3853562880295298e-06,
+      "loss": 0.3259,
+      "step": 33348
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.384779536868818e-06,
+      "loss": 0.3515,
+      "step": 33349
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3842027857081065e-06,
+      "loss": 0.4132,
+      "step": 33350
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3836260345473946e-06,
+      "loss": 0.4501,
+      "step": 33351
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.383049283386683e-06,
+      "loss": 0.4335,
+      "step": 33352
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3824725322259712e-06,
+      "loss": 0.4008,
+      "step": 33353
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3818957810652595e-06,
+      "loss": 0.3132,
+      "step": 33354
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3813190299045479e-06,
+      "loss": 0.3888,
+      "step": 33355
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.380742278743836e-06,
+      "loss": 0.3047,
+      "step": 33356
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3801655275831243e-06,
+      "loss": 0.3075,
+      "step": 33357
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3795887764224128e-06,
+      "loss": 0.3877,
+      "step": 33358
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.379012025261701e-06,
+      "loss": 0.3433,
+      "step": 33359
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3784352741009893e-06,
+      "loss": 0.4663,
+      "step": 33360
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3778585229402774e-06,
+      "loss": 0.4714,
+      "step": 33361
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.377281771779566e-06,
+      "loss": 0.4703,
+      "step": 33362
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.376705020618854e-06,
+      "loss": 0.3112,
+      "step": 33363
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3761282694581423e-06,
+      "loss": 0.4159,
+      "step": 33364
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3755515182974309e-06,
+      "loss": 0.4915,
+      "step": 33365
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.374974767136719e-06,
+      "loss": 0.3222,
+      "step": 33366
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3743980159760073e-06,
+      "loss": 0.4312,
+      "step": 33367
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3738212648152954e-06,
+      "loss": 0.2549,
+      "step": 33368
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.373244513654584e-06,
+      "loss": 0.3242,
+      "step": 33369
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3726677624938723e-06,
+      "loss": 0.4288,
+      "step": 33370
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3720910113331604e-06,
+      "loss": 0.4341,
+      "step": 33371
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3715142601724487e-06,
+      "loss": 0.2806,
+      "step": 33372
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3709375090117368e-06,
+      "loss": 0.5095,
+      "step": 33373
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3703607578510253e-06,
+      "loss": 0.4812,
+      "step": 33374
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3697840066903136e-06,
+      "loss": 0.3733,
+      "step": 33375
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3692072555296018e-06,
+      "loss": 0.3717,
+      "step": 33376
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3686305043688903e-06,
+      "loss": 0.3256,
+      "step": 33377
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3680537532081784e-06,
+      "loss": 0.389,
+      "step": 33378
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3674770020474667e-06,
+      "loss": 0.4276,
+      "step": 33379
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3669002508867552e-06,
+      "loss": 0.508,
+      "step": 33380
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3663234997260434e-06,
+      "loss": 0.5194,
+      "step": 33381
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3657467485653317e-06,
+      "loss": 0.3548,
+      "step": 33382
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3651699974046198e-06,
+      "loss": 0.3839,
+      "step": 33383
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3645932462439083e-06,
+      "loss": 0.4845,
+      "step": 33384
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3640164950831964e-06,
+      "loss": 0.4701,
+      "step": 33385
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3634397439224847e-06,
+      "loss": 0.3776,
+      "step": 33386
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.362862992761773e-06,
+      "loss": 0.3465,
+      "step": 33387
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3622862416010612e-06,
+      "loss": 0.3439,
+      "step": 33388
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3617094904403497e-06,
+      "loss": 0.3969,
+      "step": 33389
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3611327392796378e-06,
+      "loss": 0.3931,
+      "step": 33390
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3605559881189261e-06,
+      "loss": 0.2735,
+      "step": 33391
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3599792369582147e-06,
+      "loss": 0.445,
+      "step": 33392
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3594024857975028e-06,
+      "loss": 0.365,
+      "step": 33393
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.358825734636791e-06,
+      "loss": 0.3452,
+      "step": 33394
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3582489834760792e-06,
+      "loss": 0.3842,
+      "step": 33395
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3576722323153677e-06,
+      "loss": 0.3858,
+      "step": 33396
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.357095481154656e-06,
+      "loss": 0.4614,
+      "step": 33397
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3565187299939442e-06,
+      "loss": 0.412,
+      "step": 33398
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3559419788332325e-06,
+      "loss": 0.4689,
+      "step": 33399
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3553652276725208e-06,
+      "loss": 0.3038,
+      "step": 33400
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3547884765118091e-06,
+      "loss": 0.2997,
+      "step": 33401
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3542117253510974e-06,
+      "loss": 0.5073,
+      "step": 33402
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3536349741903856e-06,
+      "loss": 0.4554,
+      "step": 33403
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.353058223029674e-06,
+      "loss": 0.447,
+      "step": 33404
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3524814718689622e-06,
+      "loss": 0.3575,
+      "step": 33405
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3519047207082505e-06,
+      "loss": 0.3744,
+      "step": 33406
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3513279695475386e-06,
+      "loss": 0.3517,
+      "step": 33407
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3507512183868272e-06,
+      "loss": 0.3382,
+      "step": 33408
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3501744672261155e-06,
+      "loss": 0.4684,
+      "step": 33409
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3495977160654036e-06,
+      "loss": 0.5241,
+      "step": 33410
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3490209649046921e-06,
+      "loss": 0.4493,
+      "step": 33411
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3484442137439802e-06,
+      "loss": 0.511,
+      "step": 33412
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3478674625832685e-06,
+      "loss": 0.4626,
+      "step": 33413
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3472907114225569e-06,
+      "loss": 0.3507,
+      "step": 33414
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3467139602618452e-06,
+      "loss": 0.3551,
+      "step": 33415
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3461372091011335e-06,
+      "loss": 0.3203,
+      "step": 33416
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3455604579404216e-06,
+      "loss": 0.3376,
+      "step": 33417
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.34498370677971e-06,
+      "loss": 0.3782,
+      "step": 33418
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3444069556189985e-06,
+      "loss": 0.4526,
+      "step": 33419
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3438302044582866e-06,
+      "loss": 0.4714,
+      "step": 33420
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.343253453297575e-06,
+      "loss": 0.3538,
+      "step": 33421
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.342676702136863e-06,
+      "loss": 0.3392,
+      "step": 33422
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3420999509761515e-06,
+      "loss": 0.4024,
+      "step": 33423
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3415231998154396e-06,
+      "loss": 0.4446,
+      "step": 33424
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.340946448654728e-06,
+      "loss": 0.3067,
+      "step": 33425
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3403696974940165e-06,
+      "loss": 0.4614,
+      "step": 33426
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3397929463333046e-06,
+      "loss": 0.4059,
+      "step": 33427
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.339216195172593e-06,
+      "loss": 0.3996,
+      "step": 33428
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.338639444011881e-06,
+      "loss": 0.3743,
+      "step": 33429
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3380626928511696e-06,
+      "loss": 0.3933,
+      "step": 33430
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3374859416904579e-06,
+      "loss": 0.4369,
+      "step": 33431
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.336909190529746e-06,
+      "loss": 0.3879,
+      "step": 33432
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3363324393690343e-06,
+      "loss": 0.3498,
+      "step": 33433
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3357556882083224e-06,
+      "loss": 0.3211,
+      "step": 33434
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.335178937047611e-06,
+      "loss": 0.3527,
+      "step": 33435
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3346021858868993e-06,
+      "loss": 0.4357,
+      "step": 33436
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3340254347261874e-06,
+      "loss": 0.4611,
+      "step": 33437
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.333448683565476e-06,
+      "loss": 0.4835,
+      "step": 33438
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.332871932404764e-06,
+      "loss": 0.3176,
+      "step": 33439
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3322951812440523e-06,
+      "loss": 0.5116,
+      "step": 33440
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3317184300833409e-06,
+      "loss": 0.2723,
+      "step": 33441
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.331141678922629e-06,
+      "loss": 0.4182,
+      "step": 33442
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3305649277619173e-06,
+      "loss": 0.3941,
+      "step": 33443
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3299881766012054e-06,
+      "loss": 0.4523,
+      "step": 33444
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3294114254404937e-06,
+      "loss": 0.2985,
+      "step": 33445
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.328834674279782e-06,
+      "loss": 0.3762,
+      "step": 33446
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3282579231190704e-06,
+      "loss": 0.4037,
+      "step": 33447
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3276811719583587e-06,
+      "loss": 0.3048,
+      "step": 33448
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3271044207976468e-06,
+      "loss": 0.4052,
+      "step": 33449
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3265276696369353e-06,
+      "loss": 0.3001,
+      "step": 33450
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3259509184762234e-06,
+      "loss": 0.3607,
+      "step": 33451
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3253741673155118e-06,
+      "loss": 0.4644,
+      "step": 33452
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3247974161548003e-06,
+      "loss": 0.3679,
+      "step": 33453
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3242206649940884e-06,
+      "loss": 0.3779,
+      "step": 33454
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3236439138333767e-06,
+      "loss": 0.3612,
+      "step": 33455
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3230671626726648e-06,
+      "loss": 0.3209,
+      "step": 33456
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3224904115119534e-06,
+      "loss": 0.3902,
+      "step": 33457
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3219136603512417e-06,
+      "loss": 0.3566,
+      "step": 33458
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3213369091905298e-06,
+      "loss": 0.3804,
+      "step": 33459
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3207601580298181e-06,
+      "loss": 0.377,
+      "step": 33460
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3201834068691064e-06,
+      "loss": 0.2871,
+      "step": 33461
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3196066557083948e-06,
+      "loss": 0.3875,
+      "step": 33462
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.319029904547683e-06,
+      "loss": 0.5123,
+      "step": 33463
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3184531533869712e-06,
+      "loss": 0.3321,
+      "step": 33464
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3178764022262597e-06,
+      "loss": 0.4996,
+      "step": 33465
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3172996510655478e-06,
+      "loss": 0.3934,
+      "step": 33466
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3167228999048362e-06,
+      "loss": 0.4463,
+      "step": 33467
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3161461487441243e-06,
+      "loss": 0.3254,
+      "step": 33468
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3155693975834128e-06,
+      "loss": 0.3215,
+      "step": 33469
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3149926464227011e-06,
+      "loss": 0.3746,
+      "step": 33470
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3144158952619892e-06,
+      "loss": 0.404,
+      "step": 33471
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3138391441012778e-06,
+      "loss": 0.3008,
+      "step": 33472
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3132623929405659e-06,
+      "loss": 0.4154,
+      "step": 33473
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3126856417798542e-06,
+      "loss": 0.3816,
+      "step": 33474
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3121088906191425e-06,
+      "loss": 0.4732,
+      "step": 33475
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3115321394584308e-06,
+      "loss": 0.4746,
+      "step": 33476
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3109553882977191e-06,
+      "loss": 0.4082,
+      "step": 33477
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3103786371370073e-06,
+      "loss": 0.4835,
+      "step": 33478
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3098018859762956e-06,
+      "loss": 0.3448,
+      "step": 33479
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.309225134815584e-06,
+      "loss": 0.3287,
+      "step": 33480
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3086483836548722e-06,
+      "loss": 0.4385,
+      "step": 33481
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3080716324941605e-06,
+      "loss": 0.3487,
+      "step": 33482
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3074948813334486e-06,
+      "loss": 0.2891,
+      "step": 33483
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3069181301727372e-06,
+      "loss": 0.3579,
+      "step": 33484
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3063413790120253e-06,
+      "loss": 0.3564,
+      "step": 33485
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3057646278513136e-06,
+      "loss": 0.3697,
+      "step": 33486
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3051878766906021e-06,
+      "loss": 0.3495,
+      "step": 33487
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3046111255298902e-06,
+      "loss": 0.248,
+      "step": 33488
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3040343743691786e-06,
+      "loss": 0.4607,
+      "step": 33489
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3034576232084667e-06,
+      "loss": 0.5335,
+      "step": 33490
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.302880872047755e-06,
+      "loss": 0.3522,
+      "step": 33491
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3023041208870435e-06,
+      "loss": 0.4606,
+      "step": 33492
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.3017273697263316e-06,
+      "loss": 0.3758,
+      "step": 33493
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.30115061856562e-06,
+      "loss": 0.3768,
+      "step": 33494
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.300573867404908e-06,
+      "loss": 0.4164,
+      "step": 33495
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2999971162441966e-06,
+      "loss": 0.3618,
+      "step": 33496
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.299420365083485e-06,
+      "loss": 0.3434,
+      "step": 33497
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.298843613922773e-06,
+      "loss": 0.4117,
+      "step": 33498
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2982668627620616e-06,
+      "loss": 0.3542,
+      "step": 33499
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2976901116013497e-06,
+      "loss": 0.3896,
+      "step": 33500
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.297113360440638e-06,
+      "loss": 0.4894,
+      "step": 33501
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2965366092799265e-06,
+      "loss": 0.3601,
+      "step": 33502
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2959598581192146e-06,
+      "loss": 0.4783,
+      "step": 33503
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.295383106958503e-06,
+      "loss": 0.3232,
+      "step": 33504
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.294806355797791e-06,
+      "loss": 0.3955,
+      "step": 33505
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2942296046370794e-06,
+      "loss": 0.3832,
+      "step": 33506
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2936528534763677e-06,
+      "loss": 0.4011,
+      "step": 33507
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.293076102315656e-06,
+      "loss": 0.3734,
+      "step": 33508
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2924993511549443e-06,
+      "loss": 0.3473,
+      "step": 33509
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2919225999942324e-06,
+      "loss": 0.4027,
+      "step": 33510
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.291345848833521e-06,
+      "loss": 0.439,
+      "step": 33511
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.290769097672809e-06,
+      "loss": 0.301,
+      "step": 33512
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.2901923465120974e-06,
+      "loss": 0.366,
+      "step": 33513
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.289615595351386e-06,
+      "loss": 0.3521,
+      "step": 33514
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 1.289038844190674e-06,
+      "loss": 0.359,
+      "step": 33515
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2884620930299624e-06,
+      "loss": 0.3529,
+      "step": 33516
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2878853418692505e-06,
+      "loss": 0.3933,
+      "step": 33517
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.287308590708539e-06,
+      "loss": 0.3064,
+      "step": 33518
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2867318395478273e-06,
+      "loss": 0.4409,
+      "step": 33519
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2861550883871154e-06,
+      "loss": 0.3651,
+      "step": 33520
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2855783372264038e-06,
+      "loss": 0.3872,
+      "step": 33521
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.285001586065692e-06,
+      "loss": 0.3534,
+      "step": 33522
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2844248349049804e-06,
+      "loss": 0.5081,
+      "step": 33523
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2838480837442685e-06,
+      "loss": 0.2718,
+      "step": 33524
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2832713325835568e-06,
+      "loss": 0.4008,
+      "step": 33525
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2826945814228454e-06,
+      "loss": 0.3522,
+      "step": 33526
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2821178302621335e-06,
+      "loss": 0.338,
+      "step": 33527
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2815410791014218e-06,
+      "loss": 0.4105,
+      "step": 33528
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.28096432794071e-06,
+      "loss": 0.3827,
+      "step": 33529
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2803875767799984e-06,
+      "loss": 0.4261,
+      "step": 33530
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2798108256192868e-06,
+      "loss": 0.3921,
+      "step": 33531
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2792340744585749e-06,
+      "loss": 0.3556,
+      "step": 33532
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2786573232978634e-06,
+      "loss": 0.3898,
+      "step": 33533
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2780805721371515e-06,
+      "loss": 0.4093,
+      "step": 33534
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2775038209764398e-06,
+      "loss": 0.2885,
+      "step": 33535
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2769270698157281e-06,
+      "loss": 0.3334,
+      "step": 33536
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2763503186550162e-06,
+      "loss": 0.3727,
+      "step": 33537
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2757735674943048e-06,
+      "loss": 0.3555,
+      "step": 33538
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2751968163335929e-06,
+      "loss": 0.3596,
+      "step": 33539
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2746200651728812e-06,
+      "loss": 0.3144,
+      "step": 33540
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2740433140121697e-06,
+      "loss": 0.431,
+      "step": 33541
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2734665628514579e-06,
+      "loss": 0.3145,
+      "step": 33542
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2728898116907462e-06,
+      "loss": 0.5208,
+      "step": 33543
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2723130605300343e-06,
+      "loss": 0.4504,
+      "step": 33544
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2717363093693228e-06,
+      "loss": 0.347,
+      "step": 33545
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.271159558208611e-06,
+      "loss": 0.5848,
+      "step": 33546
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2705828070478992e-06,
+      "loss": 0.35,
+      "step": 33547
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2700060558871878e-06,
+      "loss": 0.3568,
+      "step": 33548
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2694293047264759e-06,
+      "loss": 0.3513,
+      "step": 33549
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2688525535657642e-06,
+      "loss": 0.4377,
+      "step": 33550
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2682758024050523e-06,
+      "loss": 0.4831,
+      "step": 33551
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2676990512443406e-06,
+      "loss": 0.3334,
+      "step": 33552
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2671223000836292e-06,
+      "loss": 0.2641,
+      "step": 33553
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2665455489229173e-06,
+      "loss": 0.2901,
+      "step": 33554
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2659687977622056e-06,
+      "loss": 0.3946,
+      "step": 33555
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2653920466014937e-06,
+      "loss": 0.2738,
+      "step": 33556
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2648152954407822e-06,
+      "loss": 0.398,
+      "step": 33557
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2642385442800706e-06,
+      "loss": 0.3391,
+      "step": 33558
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2636617931193587e-06,
+      "loss": 0.4118,
+      "step": 33559
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2630850419586472e-06,
+      "loss": 0.3496,
+      "step": 33560
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2625082907979353e-06,
+      "loss": 0.3889,
+      "step": 33561
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2619315396372236e-06,
+      "loss": 0.3746,
+      "step": 33562
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.261354788476512e-06,
+      "loss": 0.4942,
+      "step": 33563
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2607780373158003e-06,
+      "loss": 0.4471,
+      "step": 33564
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2602012861550886e-06,
+      "loss": 0.3549,
+      "step": 33565
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2596245349943767e-06,
+      "loss": 0.3833,
+      "step": 33566
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.259047783833665e-06,
+      "loss": 0.4741,
+      "step": 33567
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2584710326729533e-06,
+      "loss": 0.5146,
+      "step": 33568
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2578942815122417e-06,
+      "loss": 0.387,
+      "step": 33569
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.25731753035153e-06,
+      "loss": 0.4318,
+      "step": 33570
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.256740779190818e-06,
+      "loss": 0.4125,
+      "step": 33571
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2561640280301066e-06,
+      "loss": 0.4396,
+      "step": 33572
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2555872768693947e-06,
+      "loss": 0.3902,
+      "step": 33573
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.255010525708683e-06,
+      "loss": 0.4518,
+      "step": 33574
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2544337745479716e-06,
+      "loss": 0.4125,
+      "step": 33575
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2538570233872597e-06,
+      "loss": 0.3881,
+      "step": 33576
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.253280272226548e-06,
+      "loss": 0.4861,
+      "step": 33577
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2527035210658361e-06,
+      "loss": 0.4097,
+      "step": 33578
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2521267699051246e-06,
+      "loss": 0.4196,
+      "step": 33579
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.251550018744413e-06,
+      "loss": 0.4273,
+      "step": 33580
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.250973267583701e-06,
+      "loss": 0.3322,
+      "step": 33581
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2503965164229894e-06,
+      "loss": 0.3578,
+      "step": 33582
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2498197652622777e-06,
+      "loss": 0.4871,
+      "step": 33583
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.249243014101566e-06,
+      "loss": 0.3773,
+      "step": 33584
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2486662629408544e-06,
+      "loss": 0.3819,
+      "step": 33585
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2480895117801425e-06,
+      "loss": 0.2822,
+      "step": 33586
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2475127606194308e-06,
+      "loss": 0.2855,
+      "step": 33587
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.246936009458719e-06,
+      "loss": 0.3617,
+      "step": 33588
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2463592582980074e-06,
+      "loss": 0.4344,
+      "step": 33589
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2457825071372957e-06,
+      "loss": 0.354,
+      "step": 33590
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.245205755976584e-06,
+      "loss": 0.3585,
+      "step": 33591
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2446290048158722e-06,
+      "loss": 0.3806,
+      "step": 33592
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2440522536551605e-06,
+      "loss": 0.4493,
+      "step": 33593
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.243475502494449e-06,
+      "loss": 0.4162,
+      "step": 33594
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2428987513337371e-06,
+      "loss": 0.3927,
+      "step": 33595
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2423220001730255e-06,
+      "loss": 0.3547,
+      "step": 33596
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2417452490123138e-06,
+      "loss": 0.4661,
+      "step": 33597
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2411684978516019e-06,
+      "loss": 0.4053,
+      "step": 33598
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2405917466908904e-06,
+      "loss": 0.2855,
+      "step": 33599
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2400149955301787e-06,
+      "loss": 0.4401,
+      "step": 33600
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2394382443694668e-06,
+      "loss": 0.3567,
+      "step": 33601
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2388614932087552e-06,
+      "loss": 0.4574,
+      "step": 33602
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2382847420480435e-06,
+      "loss": 0.3252,
+      "step": 33603
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2377079908873318e-06,
+      "loss": 0.3382,
+      "step": 33604
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2371312397266201e-06,
+      "loss": 0.4638,
+      "step": 33605
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2365544885659084e-06,
+      "loss": 0.4226,
+      "step": 33606
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2359777374051966e-06,
+      "loss": 0.3791,
+      "step": 33607
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2354009862444849e-06,
+      "loss": 0.3477,
+      "step": 33608
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2348242350837732e-06,
+      "loss": 0.4505,
+      "step": 33609
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2342474839230615e-06,
+      "loss": 0.4299,
+      "step": 33610
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2336707327623498e-06,
+      "loss": 0.2226,
+      "step": 33611
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2330939816016382e-06,
+      "loss": 0.3585,
+      "step": 33612
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2325172304409263e-06,
+      "loss": 0.3819,
+      "step": 33613
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2319404792802146e-06,
+      "loss": 0.3802,
+      "step": 33614
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.231363728119503e-06,
+      "loss": 0.3391,
+      "step": 33615
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2307869769587912e-06,
+      "loss": 0.363,
+      "step": 33616
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2302102257980795e-06,
+      "loss": 0.3467,
+      "step": 33617
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2296334746373679e-06,
+      "loss": 0.391,
+      "step": 33618
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.229056723476656e-06,
+      "loss": 0.4951,
+      "step": 33619
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2284799723159443e-06,
+      "loss": 0.405,
+      "step": 33620
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2279032211552326e-06,
+      "loss": 0.3928,
+      "step": 33621
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.227326469994521e-06,
+      "loss": 0.3837,
+      "step": 33622
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2267497188338093e-06,
+      "loss": 0.455,
+      "step": 33623
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2261729676730976e-06,
+      "loss": 0.5207,
+      "step": 33624
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.225596216512386e-06,
+      "loss": 0.3983,
+      "step": 33625
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.225019465351674e-06,
+      "loss": 0.3264,
+      "step": 33626
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2244427141909625e-06,
+      "loss": 0.3508,
+      "step": 33627
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2238659630302507e-06,
+      "loss": 0.3216,
+      "step": 33628
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.223289211869539e-06,
+      "loss": 0.5199,
+      "step": 33629
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2227124607088273e-06,
+      "loss": 0.4471,
+      "step": 33630
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2221357095481156e-06,
+      "loss": 0.3869,
+      "step": 33631
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2215589583874037e-06,
+      "loss": 0.371,
+      "step": 33632
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2209822072266923e-06,
+      "loss": 0.3412,
+      "step": 33633
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2204054560659804e-06,
+      "loss": 0.327,
+      "step": 33634
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2198287049052687e-06,
+      "loss": 0.2884,
+      "step": 33635
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.219251953744557e-06,
+      "loss": 0.3871,
+      "step": 33636
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2186752025838453e-06,
+      "loss": 0.2798,
+      "step": 33637
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2180984514231336e-06,
+      "loss": 0.3606,
+      "step": 33638
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.217521700262422e-06,
+      "loss": 0.4013,
+      "step": 33639
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2169449491017103e-06,
+      "loss": 0.4143,
+      "step": 33640
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2163681979409984e-06,
+      "loss": 0.3736,
+      "step": 33641
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2157914467802867e-06,
+      "loss": 0.4469,
+      "step": 33642
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.215214695619575e-06,
+      "loss": 0.4134,
+      "step": 33643
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2146379444588634e-06,
+      "loss": 0.3976,
+      "step": 33644
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2140611932981517e-06,
+      "loss": 0.455,
+      "step": 33645
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.21348444213744e-06,
+      "loss": 0.426,
+      "step": 33646
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.212907690976728e-06,
+      "loss": 0.277,
+      "step": 33647
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2123309398160164e-06,
+      "loss": 0.3447,
+      "step": 33648
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2117541886553047e-06,
+      "loss": 0.3378,
+      "step": 33649
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.211177437494593e-06,
+      "loss": 0.3993,
+      "step": 33650
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2106006863338814e-06,
+      "loss": 0.4573,
+      "step": 33651
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2100239351731697e-06,
+      "loss": 0.4234,
+      "step": 33652
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2094471840124578e-06,
+      "loss": 0.377,
+      "step": 33653
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2088704328517461e-06,
+      "loss": 0.3839,
+      "step": 33654
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2082936816910345e-06,
+      "loss": 0.3024,
+      "step": 33655
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2077169305303228e-06,
+      "loss": 0.4148,
+      "step": 33656
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.207140179369611e-06,
+      "loss": 0.4543,
+      "step": 33657
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2065634282088994e-06,
+      "loss": 0.338,
+      "step": 33658
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2059866770481875e-06,
+      "loss": 0.4213,
+      "step": 33659
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2054099258874758e-06,
+      "loss": 0.3505,
+      "step": 33660
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2048331747267644e-06,
+      "loss": 0.3072,
+      "step": 33661
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2042564235660525e-06,
+      "loss": 0.3838,
+      "step": 33662
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2036796724053408e-06,
+      "loss": 0.4371,
+      "step": 33663
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2031029212446291e-06,
+      "loss": 0.402,
+      "step": 33664
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2025261700839172e-06,
+      "loss": 0.349,
+      "step": 33665
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2019494189232058e-06,
+      "loss": 0.4217,
+      "step": 33666
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.201372667762494e-06,
+      "loss": 0.4463,
+      "step": 33667
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2007959166017822e-06,
+      "loss": 0.3912,
+      "step": 33668
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.2002191654410705e-06,
+      "loss": 0.3281,
+      "step": 33669
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1996424142803588e-06,
+      "loss": 0.4298,
+      "step": 33670
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1990656631196472e-06,
+      "loss": 0.4455,
+      "step": 33671
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1984889119589355e-06,
+      "loss": 0.4077,
+      "step": 33672
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1979121607982238e-06,
+      "loss": 0.4189,
+      "step": 33673
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.197335409637512e-06,
+      "loss": 0.3659,
+      "step": 33674
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1967586584768002e-06,
+      "loss": 0.3454,
+      "step": 33675
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1961819073160885e-06,
+      "loss": 0.5029,
+      "step": 33676
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1956051561553769e-06,
+      "loss": 0.4054,
+      "step": 33677
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1950284049946652e-06,
+      "loss": 0.3966,
+      "step": 33678
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1944516538339535e-06,
+      "loss": 0.3109,
+      "step": 33679
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1938749026732416e-06,
+      "loss": 0.4939,
+      "step": 33680
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.19329815151253e-06,
+      "loss": 0.3358,
+      "step": 33681
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1927214003518183e-06,
+      "loss": 0.4523,
+      "step": 33682
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1921446491911066e-06,
+      "loss": 0.4184,
+      "step": 33683
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.191567898030395e-06,
+      "loss": 0.3428,
+      "step": 33684
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1909911468696832e-06,
+      "loss": 0.4539,
+      "step": 33685
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1904143957089715e-06,
+      "loss": 0.4836,
+      "step": 33686
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1898376445482596e-06,
+      "loss": 0.4382,
+      "step": 33687
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1892608933875482e-06,
+      "loss": 0.3403,
+      "step": 33688
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1886841422268363e-06,
+      "loss": 0.4464,
+      "step": 33689
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1881073910661246e-06,
+      "loss": 0.4109,
+      "step": 33690
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.187530639905413e-06,
+      "loss": 0.426,
+      "step": 33691
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1869538887447012e-06,
+      "loss": 0.4437,
+      "step": 33692
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1863771375839894e-06,
+      "loss": 0.3473,
+      "step": 33693
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.1858003864232779e-06,
+      "loss": 0.4399,
+      "step": 33694
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.185223635262566e-06,
+      "loss": 0.5469,
+      "step": 33695
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1846468841018543e-06,
+      "loss": 0.3419,
+      "step": 33696
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1840701329411426e-06,
+      "loss": 0.3062,
+      "step": 33697
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.183493381780431e-06,
+      "loss": 0.3054,
+      "step": 33698
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1829166306197193e-06,
+      "loss": 0.422,
+      "step": 33699
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1823398794590076e-06,
+      "loss": 0.3567,
+      "step": 33700
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1817631282982957e-06,
+      "loss": 0.4868,
+      "step": 33701
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.181186377137584e-06,
+      "loss": 0.4454,
+      "step": 33702
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1806096259768723e-06,
+      "loss": 0.4187,
+      "step": 33703
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1800328748161607e-06,
+      "loss": 0.305,
+      "step": 33704
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.179456123655449e-06,
+      "loss": 0.3538,
+      "step": 33705
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1788793724947373e-06,
+      "loss": 0.4112,
+      "step": 33706
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1783026213340256e-06,
+      "loss": 0.4567,
+      "step": 33707
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1777258701733137e-06,
+      "loss": 0.4552,
+      "step": 33708
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.177149119012602e-06,
+      "loss": 0.3994,
+      "step": 33709
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1765723678518904e-06,
+      "loss": 0.3196,
+      "step": 33710
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1759956166911787e-06,
+      "loss": 0.2563,
+      "step": 33711
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.175418865530467e-06,
+      "loss": 0.5041,
+      "step": 33712
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1748421143697553e-06,
+      "loss": 0.3347,
+      "step": 33713
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1742653632090434e-06,
+      "loss": 0.3262,
+      "step": 33714
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1736886120483318e-06,
+      "loss": 0.3832,
+      "step": 33715
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.17311186088762e-06,
+      "loss": 0.3896,
+      "step": 33716
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1725351097269084e-06,
+      "loss": 0.3858,
+      "step": 33717
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1719583585661967e-06,
+      "loss": 0.4571,
+      "step": 33718
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.171381607405485e-06,
+      "loss": 0.3839,
+      "step": 33719
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1708048562447732e-06,
+      "loss": 0.5531,
+      "step": 33720
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1702281050840615e-06,
+      "loss": 0.4432,
+      "step": 33721
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.16965135392335e-06,
+      "loss": 0.5038,
+      "step": 33722
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1690746027626381e-06,
+      "loss": 0.4084,
+      "step": 33723
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1684978516019264e-06,
+      "loss": 0.5188,
+      "step": 33724
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1679211004412148e-06,
+      "loss": 0.4558,
+      "step": 33725
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1673443492805029e-06,
+      "loss": 0.3717,
+      "step": 33726
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1667675981197914e-06,
+      "loss": 0.3913,
+      "step": 33727
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1661908469590797e-06,
+      "loss": 0.4492,
+      "step": 33728
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1656140957983678e-06,
+      "loss": 0.5248,
+      "step": 33729
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1650373446376562e-06,
+      "loss": 0.4227,
+      "step": 33730
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1644605934769445e-06,
+      "loss": 0.4436,
+      "step": 33731
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1638838423162328e-06,
+      "loss": 0.3597,
+      "step": 33732
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1633070911555211e-06,
+      "loss": 0.3376,
+      "step": 33733
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1627303399948094e-06,
+      "loss": 0.3246,
+      "step": 33734
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1621535888340975e-06,
+      "loss": 0.4312,
+      "step": 33735
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1615768376733859e-06,
+      "loss": 0.4175,
+      "step": 33736
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1610000865126742e-06,
+      "loss": 0.4128,
+      "step": 33737
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1604233353519625e-06,
+      "loss": 0.4003,
+      "step": 33738
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1598465841912508e-06,
+      "loss": 0.4575,
+      "step": 33739
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1592698330305391e-06,
+      "loss": 0.4426,
+      "step": 33740
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1586930818698273e-06,
+      "loss": 0.4068,
+      "step": 33741
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1581163307091156e-06,
+      "loss": 0.3871,
+      "step": 33742
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1575395795484039e-06,
+      "loss": 0.3921,
+      "step": 33743
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1569628283876922e-06,
+      "loss": 0.4501,
+      "step": 33744
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1563860772269805e-06,
+      "loss": 0.5931,
+      "step": 33745
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1558093260662689e-06,
+      "loss": 0.4314,
+      "step": 33746
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.155232574905557e-06,
+      "loss": 0.3866,
+      "step": 33747
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1546558237448453e-06,
+      "loss": 0.5231,
+      "step": 33748
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1540790725841338e-06,
+      "loss": 0.3731,
+      "step": 33749
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.153502321423422e-06,
+      "loss": 0.5094,
+      "step": 33750
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1529255702627102e-06,
+      "loss": 0.3296,
+      "step": 33751
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1523488191019986e-06,
+      "loss": 0.3758,
+      "step": 33752
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1517720679412869e-06,
+      "loss": 0.342,
+      "step": 33753
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.151195316780575e-06,
+      "loss": 0.4133,
+      "step": 33754
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1506185656198635e-06,
+      "loss": 0.3871,
+      "step": 33755
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1500418144591516e-06,
+      "loss": 0.3136,
+      "step": 33756
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.14946506329844e-06,
+      "loss": 0.4575,
+      "step": 33757
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1488883121377283e-06,
+      "loss": 0.4081,
+      "step": 33758
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1483115609770166e-06,
+      "loss": 0.4723,
+      "step": 33759
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.147734809816305e-06,
+      "loss": 0.3441,
+      "step": 33760
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1471580586555932e-06,
+      "loss": 0.3805,
+      "step": 33761
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1465813074948813e-06,
+      "loss": 0.3427,
+      "step": 33762
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1460045563341697e-06,
+      "loss": 0.3729,
+      "step": 33763
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.145427805173458e-06,
+      "loss": 0.4233,
+      "step": 33764
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1448510540127463e-06,
+      "loss": 0.3662,
+      "step": 33765
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1442743028520346e-06,
+      "loss": 0.3225,
+      "step": 33766
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.143697551691323e-06,
+      "loss": 0.4583,
+      "step": 33767
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1431208005306113e-06,
+      "loss": 0.3216,
+      "step": 33768
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1425440493698994e-06,
+      "loss": 0.3383,
+      "step": 33769
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1419672982091877e-06,
+      "loss": 0.3371,
+      "step": 33770
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.141390547048476e-06,
+      "loss": 0.3017,
+      "step": 33771
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1408137958877643e-06,
+      "loss": 0.3336,
+      "step": 33772
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1402370447270527e-06,
+      "loss": 0.381,
+      "step": 33773
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.139660293566341e-06,
+      "loss": 0.501,
+      "step": 33774
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.139083542405629e-06,
+      "loss": 0.4386,
+      "step": 33775
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1385067912449174e-06,
+      "loss": 0.4614,
+      "step": 33776
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1379300400842057e-06,
+      "loss": 0.3042,
+      "step": 33777
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.137353288923494e-06,
+      "loss": 0.4643,
+      "step": 33778
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1367765377627824e-06,
+      "loss": 0.3301,
+      "step": 33779
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1361997866020707e-06,
+      "loss": 0.3496,
+      "step": 33780
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1356230354413588e-06,
+      "loss": 0.3776,
+      "step": 33781
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1350462842806471e-06,
+      "loss": 0.504,
+      "step": 33782
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1344695331199357e-06,
+      "loss": 0.3001,
+      "step": 33783
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1338927819592238e-06,
+      "loss": 0.348,
+      "step": 33784
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.133316030798512e-06,
+      "loss": 0.4306,
+      "step": 33785
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1327392796378004e-06,
+      "loss": 0.3835,
+      "step": 33786
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1321625284770885e-06,
+      "loss": 0.3575,
+      "step": 33787
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.131585777316377e-06,
+      "loss": 0.3419,
+      "step": 33788
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1310090261556654e-06,
+      "loss": 0.3939,
+      "step": 33789
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1304322749949535e-06,
+      "loss": 0.3417,
+      "step": 33790
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1298555238342418e-06,
+      "loss": 0.4356,
+      "step": 33791
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1292787726735301e-06,
+      "loss": 0.3943,
+      "step": 33792
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1287020215128182e-06,
+      "loss": 0.475,
+      "step": 33793
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1281252703521068e-06,
+      "loss": 0.2929,
+      "step": 33794
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.127548519191395e-06,
+      "loss": 0.3947,
+      "step": 33795
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1269717680306832e-06,
+      "loss": 0.445,
+      "step": 33796
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1263950168699715e-06,
+      "loss": 0.3583,
+      "step": 33797
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1258182657092598e-06,
+      "loss": 0.3366,
+      "step": 33798
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1252415145485481e-06,
+      "loss": 0.4316,
+      "step": 33799
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1246647633878365e-06,
+      "loss": 0.3995,
+      "step": 33800
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1240880122271248e-06,
+      "loss": 0.4721,
+      "step": 33801
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1235112610664129e-06,
+      "loss": 0.4146,
+      "step": 33802
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1229345099057012e-06,
+      "loss": 0.3502,
+      "step": 33803
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1223577587449895e-06,
+      "loss": 0.3902,
+      "step": 33804
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1217810075842779e-06,
+      "loss": 0.3198,
+      "step": 33805
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1212042564235662e-06,
+      "loss": 0.3261,
+      "step": 33806
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1206275052628545e-06,
+      "loss": 0.3278,
+      "step": 33807
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1200507541021426e-06,
+      "loss": 0.3143,
+      "step": 33808
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.119474002941431e-06,
+      "loss": 0.4047,
+      "step": 33809
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1188972517807195e-06,
+      "loss": 0.3836,
+      "step": 33810
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1183205006200076e-06,
+      "loss": 0.3399,
+      "step": 33811
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1177437494592959e-06,
+      "loss": 0.3996,
+      "step": 33812
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1171669982985842e-06,
+      "loss": 0.3261,
+      "step": 33813
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1165902471378725e-06,
+      "loss": 0.436,
+      "step": 33814
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1160134959771606e-06,
+      "loss": 0.3176,
+      "step": 33815
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1154367448164492e-06,
+      "loss": 0.3491,
+      "step": 33816
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1148599936557373e-06,
+      "loss": 0.5169,
+      "step": 33817
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1142832424950256e-06,
+      "loss": 0.3934,
+      "step": 33818
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.113706491334314e-06,
+      "loss": 0.2663,
+      "step": 33819
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1131297401736022e-06,
+      "loss": 0.4125,
+      "step": 33820
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1125529890128903e-06,
+      "loss": 0.3696,
+      "step": 33821
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1119762378521789e-06,
+      "loss": 0.4449,
+      "step": 33822
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.111399486691467e-06,
+      "loss": 0.351,
+      "step": 33823
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1108227355307553e-06,
+      "loss": 0.4038,
+      "step": 33824
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1102459843700436e-06,
+      "loss": 0.3502,
+      "step": 33825
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.109669233209332e-06,
+      "loss": 0.448,
+      "step": 33826
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1090924820486203e-06,
+      "loss": 0.4144,
+      "step": 33827
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1085157308879086e-06,
+      "loss": 0.414,
+      "step": 33828
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.107938979727197e-06,
+      "loss": 0.3874,
+      "step": 33829
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.107362228566485e-06,
+      "loss": 0.4434,
+      "step": 33830
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1067854774057733e-06,
+      "loss": 0.3786,
+      "step": 33831
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1062087262450617e-06,
+      "loss": 0.2988,
+      "step": 33832
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.10563197508435e-06,
+      "loss": 0.3892,
+      "step": 33833
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1050552239236383e-06,
+      "loss": 0.3914,
+      "step": 33834
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1044784727629266e-06,
+      "loss": 0.3745,
+      "step": 33835
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1039017216022147e-06,
+      "loss": 0.5112,
+      "step": 33836
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.103324970441503e-06,
+      "loss": 0.3216,
+      "step": 33837
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1027482192807914e-06,
+      "loss": 0.3807,
+      "step": 33838
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1021714681200797e-06,
+      "loss": 0.2881,
+      "step": 33839
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.101594716959368e-06,
+      "loss": 0.4349,
+      "step": 33840
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1010179657986563e-06,
+      "loss": 0.3298,
+      "step": 33841
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.1004412146379444e-06,
+      "loss": 0.4344,
+      "step": 33842
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0998644634772328e-06,
+      "loss": 0.3305,
+      "step": 33843
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.099287712316521e-06,
+      "loss": 0.3004,
+      "step": 33844
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0987109611558094e-06,
+      "loss": 0.403,
+      "step": 33845
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0981342099950977e-06,
+      "loss": 0.5025,
+      "step": 33846
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.097557458834386e-06,
+      "loss": 0.2918,
+      "step": 33847
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0969807076736741e-06,
+      "loss": 0.2752,
+      "step": 33848
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0964039565129627e-06,
+      "loss": 0.378,
+      "step": 33849
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.095827205352251e-06,
+      "loss": 0.4028,
+      "step": 33850
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0952504541915391e-06,
+      "loss": 0.4043,
+      "step": 33851
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0946737030308274e-06,
+      "loss": 0.4289,
+      "step": 33852
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0940969518701157e-06,
+      "loss": 0.4614,
+      "step": 33853
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0935202007094039e-06,
+      "loss": 0.3635,
+      "step": 33854
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0929434495486924e-06,
+      "loss": 0.3821,
+      "step": 33855
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0923666983879807e-06,
+      "loss": 0.3735,
+      "step": 33856
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0917899472272688e-06,
+      "loss": 0.4415,
+      "step": 33857
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0912131960665571e-06,
+      "loss": 0.3659,
+      "step": 33858
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0906364449058455e-06,
+      "loss": 0.4614,
+      "step": 33859
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0900596937451338e-06,
+      "loss": 0.3715,
+      "step": 33860
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.089482942584422e-06,
+      "loss": 0.4689,
+      "step": 33861
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0889061914237104e-06,
+      "loss": 0.3522,
+      "step": 33862
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0883294402629985e-06,
+      "loss": 0.3983,
+      "step": 33863
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0877526891022868e-06,
+      "loss": 0.3592,
+      "step": 33864
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0871759379415752e-06,
+      "loss": 0.4347,
+      "step": 33865
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0865991867808635e-06,
+      "loss": 0.3421,
+      "step": 33866
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0860224356201518e-06,
+      "loss": 0.4055,
+      "step": 33867
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0854456844594401e-06,
+      "loss": 0.3833,
+      "step": 33868
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0848689332987282e-06,
+      "loss": 0.3695,
+      "step": 33869
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0842921821380166e-06,
+      "loss": 0.3673,
+      "step": 33870
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.083715430977305e-06,
+      "loss": 0.3426,
+      "step": 33871
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0831386798165932e-06,
+      "loss": 0.3733,
+      "step": 33872
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 1.0825619286558815e-06,
+      "loss": 0.2588,
+      "step": 33873
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0819851774951698e-06,
+      "loss": 0.4017,
+      "step": 33874
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0814084263344582e-06,
+      "loss": 0.3886,
+      "step": 33875
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0808316751737463e-06,
+      "loss": 0.3234,
+      "step": 33876
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0802549240130348e-06,
+      "loss": 0.415,
+      "step": 33877
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.079678172852323e-06,
+      "loss": 0.3946,
+      "step": 33878
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0791014216916112e-06,
+      "loss": 0.3677,
+      "step": 33879
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0785246705308996e-06,
+      "loss": 0.4131,
+      "step": 33880
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0779479193701879e-06,
+      "loss": 0.3735,
+      "step": 33881
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.077371168209476e-06,
+      "loss": 0.3445,
+      "step": 33882
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0767944170487645e-06,
+      "loss": 0.3864,
+      "step": 33883
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0762176658880526e-06,
+      "loss": 0.3284,
+      "step": 33884
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.075640914727341e-06,
+      "loss": 0.3189,
+      "step": 33885
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0750641635666293e-06,
+      "loss": 0.3972,
+      "step": 33886
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0744874124059176e-06,
+      "loss": 0.4109,
+      "step": 33887
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.073910661245206e-06,
+      "loss": 0.4088,
+      "step": 33888
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0733339100844942e-06,
+      "loss": 0.3503,
+      "step": 33889
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0727571589237823e-06,
+      "loss": 0.457,
+      "step": 33890
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0721804077630707e-06,
+      "loss": 0.3603,
+      "step": 33891
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.071603656602359e-06,
+      "loss": 0.4184,
+      "step": 33892
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0710269054416473e-06,
+      "loss": 0.3427,
+      "step": 33893
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0704501542809356e-06,
+      "loss": 0.4861,
+      "step": 33894
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.069873403120224e-06,
+      "loss": 0.2862,
+      "step": 33895
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0692966519595123e-06,
+      "loss": 0.4711,
+      "step": 33896
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0687199007988004e-06,
+      "loss": 0.4466,
+      "step": 33897
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0681431496380887e-06,
+      "loss": 0.3024,
+      "step": 33898
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.067566398477377e-06,
+      "loss": 0.4358,
+      "step": 33899
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0669896473166653e-06,
+      "loss": 0.3416,
+      "step": 33900
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0664128961559536e-06,
+      "loss": 0.4692,
+      "step": 33901
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.065836144995242e-06,
+      "loss": 0.4131,
+      "step": 33902
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.06525939383453e-06,
+      "loss": 0.4025,
+      "step": 33903
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0646826426738184e-06,
+      "loss": 0.3585,
+      "step": 33904
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0641058915131067e-06,
+      "loss": 0.4504,
+      "step": 33905
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.063529140352395e-06,
+      "loss": 0.4172,
+      "step": 33906
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0629523891916834e-06,
+      "loss": 0.4627,
+      "step": 33907
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0623756380309717e-06,
+      "loss": 0.4893,
+      "step": 33908
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0617988868702598e-06,
+      "loss": 0.3808,
+      "step": 33909
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0612221357095483e-06,
+      "loss": 0.3039,
+      "step": 33910
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0606453845488366e-06,
+      "loss": 0.3056,
+      "step": 33911
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0600686333881247e-06,
+      "loss": 0.3084,
+      "step": 33912
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.059491882227413e-06,
+      "loss": 0.3547,
+      "step": 33913
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0589151310667014e-06,
+      "loss": 0.4528,
+      "step": 33914
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0583383799059895e-06,
+      "loss": 0.3395,
+      "step": 33915
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.057761628745278e-06,
+      "loss": 0.3347,
+      "step": 33916
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0571848775845663e-06,
+      "loss": 0.456,
+      "step": 33917
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0566081264238545e-06,
+      "loss": 0.4449,
+      "step": 33918
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0560313752631428e-06,
+      "loss": 0.4723,
+      "step": 33919
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.055454624102431e-06,
+      "loss": 0.436,
+      "step": 33920
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0548778729417194e-06,
+      "loss": 0.2882,
+      "step": 33921
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0543011217810077e-06,
+      "loss": 0.4536,
+      "step": 33922
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.053724370620296e-06,
+      "loss": 0.5242,
+      "step": 33923
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0531476194595842e-06,
+      "loss": 0.3793,
+      "step": 33924
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0525708682988725e-06,
+      "loss": 0.3326,
+      "step": 33925
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0519941171381608e-06,
+      "loss": 0.4128,
+      "step": 33926
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0514173659774491e-06,
+      "loss": 0.33,
+      "step": 33927
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0508406148167374e-06,
+      "loss": 0.3871,
+      "step": 33928
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0502638636560258e-06,
+      "loss": 0.4165,
+      "step": 33929
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0496871124953139e-06,
+      "loss": 0.3824,
+      "step": 33930
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0491103613346022e-06,
+      "loss": 0.3521,
+      "step": 33931
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0485336101738905e-06,
+      "loss": 0.4015,
+      "step": 33932
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0479568590131788e-06,
+      "loss": 0.3915,
+      "step": 33933
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0473801078524672e-06,
+      "loss": 0.4477,
+      "step": 33934
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0468033566917555e-06,
+      "loss": 0.4221,
+      "step": 33935
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0462266055310436e-06,
+      "loss": 0.3353,
+      "step": 33936
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.045649854370332e-06,
+      "loss": 0.4759,
+      "step": 33937
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0450731032096204e-06,
+      "loss": 0.3096,
+      "step": 33938
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0444963520489085e-06,
+      "loss": 0.5191,
+      "step": 33939
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0439196008881969e-06,
+      "loss": 0.4187,
+      "step": 33940
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0433428497274852e-06,
+      "loss": 0.384,
+      "step": 33941
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0427660985667735e-06,
+      "loss": 0.2987,
+      "step": 33942
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0421893474060616e-06,
+      "loss": 0.4096,
+      "step": 33943
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0416125962453501e-06,
+      "loss": 0.3176,
+      "step": 33944
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0410358450846383e-06,
+      "loss": 0.4407,
+      "step": 33945
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0404590939239266e-06,
+      "loss": 0.3695,
+      "step": 33946
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.039882342763215e-06,
+      "loss": 0.4621,
+      "step": 33947
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0393055916025032e-06,
+      "loss": 0.5018,
+      "step": 33948
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0387288404417915e-06,
+      "loss": 0.3294,
+      "step": 33949
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0381520892810799e-06,
+      "loss": 0.4056,
+      "step": 33950
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.037575338120368e-06,
+      "loss": 0.3867,
+      "step": 33951
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0369985869596563e-06,
+      "loss": 0.3529,
+      "step": 33952
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0364218357989446e-06,
+      "loss": 0.4232,
+      "step": 33953
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.035845084638233e-06,
+      "loss": 0.4309,
+      "step": 33954
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0352683334775212e-06,
+      "loss": 0.3501,
+      "step": 33955
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0346915823168096e-06,
+      "loss": 0.4305,
+      "step": 33956
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0341148311560979e-06,
+      "loss": 0.4539,
+      "step": 33957
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.033538079995386e-06,
+      "loss": 0.3645,
+      "step": 33958
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0329613288346743e-06,
+      "loss": 0.3837,
+      "step": 33959
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0323845776739626e-06,
+      "loss": 0.4252,
+      "step": 33960
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.031807826513251e-06,
+      "loss": 0.369,
+      "step": 33961
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0312310753525393e-06,
+      "loss": 0.3857,
+      "step": 33962
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0306543241918276e-06,
+      "loss": 0.4013,
+      "step": 33963
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0300775730311157e-06,
+      "loss": 0.3475,
+      "step": 33964
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.029500821870404e-06,
+      "loss": 0.3509,
+      "step": 33965
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0289240707096924e-06,
+      "loss": 0.2651,
+      "step": 33966
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0283473195489807e-06,
+      "loss": 0.4125,
+      "step": 33967
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.027770568388269e-06,
+      "loss": 0.3442,
+      "step": 33968
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0271938172275573e-06,
+      "loss": 0.4257,
+      "step": 33969
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0266170660668454e-06,
+      "loss": 0.3312,
+      "step": 33970
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.026040314906134e-06,
+      "loss": 0.3455,
+      "step": 33971
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.025463563745422e-06,
+      "loss": 0.4637,
+      "step": 33972
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0248868125847104e-06,
+      "loss": 0.465,
+      "step": 33973
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0243100614239987e-06,
+      "loss": 0.3817,
+      "step": 33974
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.023733310263287e-06,
+      "loss": 0.5065,
+      "step": 33975
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0231565591025751e-06,
+      "loss": 0.458,
+      "step": 33976
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0225798079418637e-06,
+      "loss": 0.3808,
+      "step": 33977
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.022003056781152e-06,
+      "loss": 0.3766,
+      "step": 33978
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.02142630562044e-06,
+      "loss": 0.4455,
+      "step": 33979
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0208495544597284e-06,
+      "loss": 0.4252,
+      "step": 33980
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0202728032990167e-06,
+      "loss": 0.381,
+      "step": 33981
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0196960521383048e-06,
+      "loss": 0.4733,
+      "step": 33982
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0191193009775934e-06,
+      "loss": 0.4888,
+      "step": 33983
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0185425498168817e-06,
+      "loss": 0.3191,
+      "step": 33984
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0179657986561698e-06,
+      "loss": 0.495,
+      "step": 33985
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0173890474954581e-06,
+      "loss": 0.4384,
+      "step": 33986
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0168122963347464e-06,
+      "loss": 0.4096,
+      "step": 33987
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0162355451740348e-06,
+      "loss": 0.3929,
+      "step": 33988
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.015658794013323e-06,
+      "loss": 0.4077,
+      "step": 33989
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0150820428526114e-06,
+      "loss": 0.4577,
+      "step": 33990
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0145052916918995e-06,
+      "loss": 0.4024,
+      "step": 33991
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0139285405311878e-06,
+      "loss": 0.4431,
+      "step": 33992
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0133517893704762e-06,
+      "loss": 0.4317,
+      "step": 33993
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0127750382097645e-06,
+      "loss": 0.3071,
+      "step": 33994
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0121982870490528e-06,
+      "loss": 0.3244,
+      "step": 33995
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0116215358883411e-06,
+      "loss": 0.4172,
+      "step": 33996
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0110447847276292e-06,
+      "loss": 0.361,
+      "step": 33997
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0104680335669175e-06,
+      "loss": 0.3397,
+      "step": 33998
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.009891282406206e-06,
+      "loss": 0.4194,
+      "step": 33999
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0093145312454942e-06,
+      "loss": 0.4263,
+      "step": 34000
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0087377800847825e-06,
+      "loss": 0.4021,
+      "step": 34001
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0081610289240708e-06,
+      "loss": 0.2833,
+      "step": 34002
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0075842777633591e-06,
+      "loss": 0.495,
+      "step": 34003
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0070075266026473e-06,
+      "loss": 0.3716,
+      "step": 34004
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0064307754419358e-06,
+      "loss": 0.3277,
+      "step": 34005
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.005854024281224e-06,
+      "loss": 0.5154,
+      "step": 34006
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0052772731205122e-06,
+      "loss": 0.4219,
+      "step": 34007
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0047005219598005e-06,
+      "loss": 0.4829,
+      "step": 34008
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0041237707990889e-06,
+      "loss": 0.4884,
+      "step": 34009
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0035470196383772e-06,
+      "loss": 0.4118,
+      "step": 34010
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0029702684776655e-06,
+      "loss": 0.3192,
+      "step": 34011
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0023935173169536e-06,
+      "loss": 0.2845,
+      "step": 34012
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.001816766156242e-06,
+      "loss": 0.5616,
+      "step": 34013
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0012400149955302e-06,
+      "loss": 0.3595,
+      "step": 34014
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0006632638348186e-06,
+      "loss": 0.3301,
+      "step": 34015
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0000865126741069e-06,
+      "loss": 0.5359,
+      "step": 34016
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.995097615133952e-07,
+      "loss": 0.2897,
+      "step": 34017
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.989330103526833e-07,
+      "loss": 0.3548,
+      "step": 34018
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.983562591919716e-07,
+      "loss": 0.4492,
+      "step": 34019
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.9777950803126e-07,
+      "loss": 0.5055,
+      "step": 34020
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.972027568705483e-07,
+      "loss": 0.4067,
+      "step": 34021
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.966260057098366e-07,
+      "loss": 0.3906,
+      "step": 34022
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.96049254549125e-07,
+      "loss": 0.3212,
+      "step": 34023
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.954725033884132e-07,
+      "loss": 0.3188,
+      "step": 34024
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.948957522277013e-07,
+      "loss": 0.375,
+      "step": 34025
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.943190010669897e-07,
+      "loss": 0.3632,
+      "step": 34026
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.93742249906278e-07,
+      "loss": 0.3704,
+      "step": 34027
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.931654987455663e-07,
+      "loss": 0.3991,
+      "step": 34028
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.925887475848546e-07,
+      "loss": 0.5279,
+      "step": 34029
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.92011996424143e-07,
+      "loss": 0.4333,
+      "step": 34030
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.91435245263431e-07,
+      "loss": 0.4094,
+      "step": 34031
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.908584941027196e-07,
+      "loss": 0.4417,
+      "step": 34032
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.902817429420077e-07,
+      "loss": 0.5298,
+      "step": 34033
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.89704991781296e-07,
+      "loss": 0.4485,
+      "step": 34034
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.891282406205843e-07,
+      "loss": 0.3388,
+      "step": 34035
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.885514894598727e-07,
+      "loss": 0.3465,
+      "step": 34036
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.879747382991608e-07,
+      "loss": 0.3634,
+      "step": 34037
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.873979871384493e-07,
+      "loss": 0.389,
+      "step": 34038
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.868212359777376e-07,
+      "loss": 0.3124,
+      "step": 34039
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.862444848170257e-07,
+      "loss": 0.3772,
+      "step": 34040
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.85667733656314e-07,
+      "loss": 0.319,
+      "step": 34041
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.850909824956024e-07,
+      "loss": 0.316,
+      "step": 34042
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.845142313348905e-07,
+      "loss": 0.4348,
+      "step": 34043
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.83937480174179e-07,
+      "loss": 0.3486,
+      "step": 34044
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.833607290134673e-07,
+      "loss": 0.3885,
+      "step": 34045
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.827839778527554e-07,
+      "loss": 0.4896,
+      "step": 34046
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.822072266920438e-07,
+      "loss": 0.3503,
+      "step": 34047
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.81630475531332e-07,
+      "loss": 0.4259,
+      "step": 34048
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.810537243706204e-07,
+      "loss": 0.4333,
+      "step": 34049
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.804769732099087e-07,
+      "loss": 0.511,
+      "step": 34050
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 9.79900222049197e-07,
+      "loss": 0.304,
+      "step": 34051
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.793234708884851e-07,
+      "loss": 0.3505,
+      "step": 34052
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.787467197277735e-07,
+      "loss": 0.4406,
+      "step": 34053
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.781699685670618e-07,
+      "loss": 0.4428,
+      "step": 34054
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.775932174063501e-07,
+      "loss": 0.3601,
+      "step": 34055
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.770164662456384e-07,
+      "loss": 0.3696,
+      "step": 34056
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.764397150849268e-07,
+      "loss": 0.4135,
+      "step": 34057
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.758629639242149e-07,
+      "loss": 0.329,
+      "step": 34058
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.752862127635032e-07,
+      "loss": 0.3177,
+      "step": 34059
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.747094616027917e-07,
+      "loss": 0.3776,
+      "step": 34060
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.741327104420798e-07,
+      "loss": 0.2912,
+      "step": 34061
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.735559592813681e-07,
+      "loss": 0.4059,
+      "step": 34062
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.729792081206565e-07,
+      "loss": 0.2908,
+      "step": 34063
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.724024569599446e-07,
+      "loss": 0.3697,
+      "step": 34064
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.718257057992329e-07,
+      "loss": 0.4445,
+      "step": 34065
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.712489546385214e-07,
+      "loss": 0.3801,
+      "step": 34066
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.706722034778095e-07,
+      "loss": 0.4356,
+      "step": 34067
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.700954523170979e-07,
+      "loss": 0.3429,
+      "step": 34068
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.695187011563862e-07,
+      "loss": 0.3427,
+      "step": 34069
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.689419499956745e-07,
+      "loss": 0.4792,
+      "step": 34070
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.683651988349628e-07,
+      "loss": 0.3646,
+      "step": 34071
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.677884476742511e-07,
+      "loss": 0.4858,
+      "step": 34072
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.672116965135392e-07,
+      "loss": 0.3547,
+      "step": 34073
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.666349453528276e-07,
+      "loss": 0.2728,
+      "step": 34074
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.660581941921159e-07,
+      "loss": 0.4528,
+      "step": 34075
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.654814430314042e-07,
+      "loss": 0.3376,
+      "step": 34076
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.649046918706925e-07,
+      "loss": 0.4504,
+      "step": 34077
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.643279407099808e-07,
+      "loss": 0.423,
+      "step": 34078
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.63751189549269e-07,
+      "loss": 0.4533,
+      "step": 34079
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.631744383885573e-07,
+      "loss": 0.3685,
+      "step": 34080
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.625976872278456e-07,
+      "loss": 0.46,
+      "step": 34081
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.62020936067134e-07,
+      "loss": 0.4978,
+      "step": 34082
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.614441849064222e-07,
+      "loss": 0.4612,
+      "step": 34083
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.608674337457106e-07,
+      "loss": 0.3974,
+      "step": 34084
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.602906825849989e-07,
+      "loss": 0.478,
+      "step": 34085
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.59713931424287e-07,
+      "loss": 0.4109,
+      "step": 34086
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.591371802635753e-07,
+      "loss": 0.4186,
+      "step": 34087
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.585604291028636e-07,
+      "loss": 0.3925,
+      "step": 34088
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.57983677942152e-07,
+      "loss": 0.4513,
+      "step": 34089
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.574069267814403e-07,
+      "loss": 0.219,
+      "step": 34090
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.568301756207286e-07,
+      "loss": 0.3543,
+      "step": 34091
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.562534244600167e-07,
+      "loss": 0.3492,
+      "step": 34092
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.55676673299305e-07,
+      "loss": 0.4235,
+      "step": 34093
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.550999221385933e-07,
+      "loss": 0.3766,
+      "step": 34094
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.545231709778817e-07,
+      "loss": 0.3842,
+      "step": 34095
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.5394641981717e-07,
+      "loss": 0.4376,
+      "step": 34096
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.533696686564582e-07,
+      "loss": 0.4163,
+      "step": 34097
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.527929174957465e-07,
+      "loss": 0.3884,
+      "step": 34098
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.522161663350349e-07,
+      "loss": 0.3119,
+      "step": 34099
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.516394151743232e-07,
+      "loss": 0.4095,
+      "step": 34100
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.510626640136114e-07,
+      "loss": 0.4205,
+      "step": 34101
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.504859128528997e-07,
+      "loss": 0.3952,
+      "step": 34102
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.499091616921879e-07,
+      "loss": 0.4481,
+      "step": 34103
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.493324105314762e-07,
+      "loss": 0.4347,
+      "step": 34104
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.487556593707646e-07,
+      "loss": 0.3638,
+      "step": 34105
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.481789082100529e-07,
+      "loss": 0.3269,
+      "step": 34106
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.476021570493412e-07,
+      "loss": 0.4099,
+      "step": 34107
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.470254058886294e-07,
+      "loss": 0.4176,
+      "step": 34108
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.464486547279177e-07,
+      "loss": 0.3327,
+      "step": 34109
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.45871903567206e-07,
+      "loss": 0.454,
+      "step": 34110
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.452951524064944e-07,
+      "loss": 0.5284,
+      "step": 34111
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.447184012457826e-07,
+      "loss": 0.3099,
+      "step": 34112
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.441416500850709e-07,
+      "loss": 0.3212,
+      "step": 34113
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.435648989243591e-07,
+      "loss": 0.3666,
+      "step": 34114
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.429881477636474e-07,
+      "loss": 0.4688,
+      "step": 34115
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.424113966029357e-07,
+      "loss": 0.307,
+      "step": 34116
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.418346454422241e-07,
+      "loss": 0.3511,
+      "step": 34117
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.412578942815123e-07,
+      "loss": 0.3575,
+      "step": 34118
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.406811431208006e-07,
+      "loss": 0.3508,
+      "step": 34119
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.401043919600888e-07,
+      "loss": 0.3305,
+      "step": 34120
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.395276407993772e-07,
+      "loss": 0.4507,
+      "step": 34121
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.389508896386656e-07,
+      "loss": 0.3029,
+      "step": 34122
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.383741384779538e-07,
+      "loss": 0.3866,
+      "step": 34123
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.37797387317242e-07,
+      "loss": 0.4114,
+      "step": 34124
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.372206361565303e-07,
+      "loss": 0.3238,
+      "step": 34125
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.366438849958185e-07,
+      "loss": 0.3553,
+      "step": 34126
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.36067133835107e-07,
+      "loss": 0.4075,
+      "step": 34127
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.354903826743953e-07,
+      "loss": 0.2853,
+      "step": 34128
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.349136315136835e-07,
+      "loss": 0.3872,
+      "step": 34129
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.343368803529718e-07,
+      "loss": 0.4446,
+      "step": 34130
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.3376012919226e-07,
+      "loss": 0.3155,
+      "step": 34131
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.331833780315485e-07,
+      "loss": 0.476,
+      "step": 34132
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.326066268708367e-07,
+      "loss": 0.2663,
+      "step": 34133
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.32029875710125e-07,
+      "loss": 0.3891,
+      "step": 34134
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.314531245494132e-07,
+      "loss": 0.3198,
+      "step": 34135
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.308763733887015e-07,
+      "loss": 0.4272,
+      "step": 34136
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.302996222279897e-07,
+      "loss": 0.4847,
+      "step": 34137
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.297228710672782e-07,
+      "loss": 0.3275,
+      "step": 34138
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.291461199065664e-07,
+      "loss": 0.4049,
+      "step": 34139
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.285693687458547e-07,
+      "loss": 0.5027,
+      "step": 34140
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.279926175851429e-07,
+      "loss": 0.4074,
+      "step": 34141
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.274158664244312e-07,
+      "loss": 0.2778,
+      "step": 34142
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.268391152637197e-07,
+      "loss": 0.3742,
+      "step": 34143
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.262623641030079e-07,
+      "loss": 0.3083,
+      "step": 34144
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.256856129422962e-07,
+      "loss": 0.4995,
+      "step": 34145
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.251088617815844e-07,
+      "loss": 0.4628,
+      "step": 34146
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.245321106208726e-07,
+      "loss": 0.438,
+      "step": 34147
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.239553594601609e-07,
+      "loss": 0.4957,
+      "step": 34148
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.233786082994494e-07,
+      "loss": 0.4104,
+      "step": 34149
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.228018571387376e-07,
+      "loss": 0.3329,
+      "step": 34150
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.222251059780259e-07,
+      "loss": 0.4863,
+      "step": 34151
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.216483548173141e-07,
+      "loss": 0.4375,
+      "step": 34152
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.210716036566024e-07,
+      "loss": 0.4941,
+      "step": 34153
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.204948524958907e-07,
+      "loss": 0.4958,
+      "step": 34154
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.199181013351791e-07,
+      "loss": 0.3682,
+      "step": 34155
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.193413501744673e-07,
+      "loss": 0.4014,
+      "step": 34156
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.187645990137556e-07,
+      "loss": 0.3285,
+      "step": 34157
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.181878478530438e-07,
+      "loss": 0.3573,
+      "step": 34158
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.176110966923321e-07,
+      "loss": 0.4034,
+      "step": 34159
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.170343455316205e-07,
+      "loss": 0.4015,
+      "step": 34160
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.164575943709088e-07,
+      "loss": 0.3215,
+      "step": 34161
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.15880843210197e-07,
+      "loss": 0.2685,
+      "step": 34162
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.153040920494853e-07,
+      "loss": 0.3806,
+      "step": 34163
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.147273408887735e-07,
+      "loss": 0.3746,
+      "step": 34164
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.141505897280619e-07,
+      "loss": 0.4618,
+      "step": 34165
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.135738385673503e-07,
+      "loss": 0.3877,
+      "step": 34166
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.129970874066385e-07,
+      "loss": 0.4036,
+      "step": 34167
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.124203362459268e-07,
+      "loss": 0.2712,
+      "step": 34168
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.11843585085215e-07,
+      "loss": 0.4032,
+      "step": 34169
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.112668339245032e-07,
+      "loss": 0.4256,
+      "step": 34170
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.106900827637917e-07,
+      "loss": 0.4236,
+      "step": 34171
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.1011333160308e-07,
+      "loss": 0.4002,
+      "step": 34172
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.095365804423682e-07,
+      "loss": 0.4621,
+      "step": 34173
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.089598292816565e-07,
+      "loss": 0.4258,
+      "step": 34174
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.083830781209447e-07,
+      "loss": 0.301,
+      "step": 34175
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.078063269602331e-07,
+      "loss": 0.5125,
+      "step": 34176
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.072295757995214e-07,
+      "loss": 0.5408,
+      "step": 34177
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.066528246388097e-07,
+      "loss": 0.4077,
+      "step": 34178
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.060760734780979e-07,
+      "loss": 0.3523,
+      "step": 34179
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.054993223173862e-07,
+      "loss": 0.4387,
+      "step": 34180
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.049225711566745e-07,
+      "loss": 0.3391,
+      "step": 34181
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.043458199959629e-07,
+      "loss": 0.3336,
+      "step": 34182
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.037690688352511e-07,
+      "loss": 0.3589,
+      "step": 34183
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.031923176745394e-07,
+      "loss": 0.4923,
+      "step": 34184
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.026155665138276e-07,
+      "loss": 0.4636,
+      "step": 34185
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.02038815353116e-07,
+      "loss": 0.3568,
+      "step": 34186
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.014620641924042e-07,
+      "loss": 0.3721,
+      "step": 34187
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.008853130316926e-07,
+      "loss": 0.3712,
+      "step": 34188
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 9.003085618709809e-07,
+      "loss": 0.4026,
+      "step": 34189
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.997318107102691e-07,
+      "loss": 0.3975,
+      "step": 34190
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.991550595495574e-07,
+      "loss": 0.4148,
+      "step": 34191
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.985783083888457e-07,
+      "loss": 0.3273,
+      "step": 34192
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.980015572281341e-07,
+      "loss": 0.4319,
+      "step": 34193
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.974248060674223e-07,
+      "loss": 0.3333,
+      "step": 34194
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.968480549067106e-07,
+      "loss": 0.3413,
+      "step": 34195
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.962713037459988e-07,
+      "loss": 0.4615,
+      "step": 34196
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.956945525852872e-07,
+      "loss": 0.3444,
+      "step": 34197
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.951178014245754e-07,
+      "loss": 0.3731,
+      "step": 34198
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.945410502638638e-07,
+      "loss": 0.3873,
+      "step": 34199
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.93964299103152e-07,
+      "loss": 0.507,
+      "step": 34200
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.933875479424403e-07,
+      "loss": 0.45,
+      "step": 34201
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.928107967817285e-07,
+      "loss": 0.4411,
+      "step": 34202
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.922340456210169e-07,
+      "loss": 0.4077,
+      "step": 34203
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.916572944603051e-07,
+      "loss": 0.3566,
+      "step": 34204
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.910805432995935e-07,
+      "loss": 0.4246,
+      "step": 34205
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.905037921388817e-07,
+      "loss": 0.4444,
+      "step": 34206
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.8992704097817e-07,
+      "loss": 0.3993,
+      "step": 34207
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.893502898174583e-07,
+      "loss": 0.3202,
+      "step": 34208
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.887735386567466e-07,
+      "loss": 0.3152,
+      "step": 34209
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.88196787496035e-07,
+      "loss": 0.4413,
+      "step": 34210
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.876200363353232e-07,
+      "loss": 0.3303,
+      "step": 34211
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.870432851746115e-07,
+      "loss": 0.3755,
+      "step": 34212
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.864665340138998e-07,
+      "loss": 0.5192,
+      "step": 34213
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.858897828531881e-07,
+      "loss": 0.4227,
+      "step": 34214
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.853130316924763e-07,
+      "loss": 0.2645,
+      "step": 34215
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.847362805317647e-07,
+      "loss": 0.3898,
+      "step": 34216
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.841595293710529e-07,
+      "loss": 0.526,
+      "step": 34217
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.835827782103413e-07,
+      "loss": 0.2586,
+      "step": 34218
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.830060270496295e-07,
+      "loss": 0.3853,
+      "step": 34219
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.824292758889178e-07,
+      "loss": 0.5481,
+      "step": 34220
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.818525247282061e-07,
+      "loss": 0.3694,
+      "step": 34221
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.812757735674944e-07,
+      "loss": 0.4536,
+      "step": 34222
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.806990224067826e-07,
+      "loss": 0.4021,
+      "step": 34223
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.80122271246071e-07,
+      "loss": 0.4128,
+      "step": 34224
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.795455200853592e-07,
+      "loss": 0.5162,
+      "step": 34225
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.789687689246475e-07,
+      "loss": 0.4198,
+      "step": 34226
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.783920177639359e-07,
+      "loss": 0.3565,
+      "step": 34227
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.778152666032241e-07,
+      "loss": 0.4517,
+      "step": 34228
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.772385154425124e-07,
+      "loss": 0.4451,
+      "step": 34229
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 8.766617642818007e-07,
+      "loss": 0.3317,
+      "step": 34230
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.760850131210889e-07,
+      "loss": 0.4187,
+      "step": 34231
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.755082619603773e-07,
+      "loss": 0.4153,
+      "step": 34232
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.749315107996656e-07,
+      "loss": 0.3648,
+      "step": 34233
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.743547596389538e-07,
+      "loss": 0.3648,
+      "step": 34234
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.737780084782422e-07,
+      "loss": 0.374,
+      "step": 34235
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.732012573175304e-07,
+      "loss": 0.2886,
+      "step": 34236
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.726245061568187e-07,
+      "loss": 0.2574,
+      "step": 34237
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.72047754996107e-07,
+      "loss": 0.4477,
+      "step": 34238
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.714710038353953e-07,
+      "loss": 0.3882,
+      "step": 34239
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.708942526746836e-07,
+      "loss": 0.5108,
+      "step": 34240
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.703175015139719e-07,
+      "loss": 0.3785,
+      "step": 34241
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.697407503532601e-07,
+      "loss": 0.3258,
+      "step": 34242
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.691639991925485e-07,
+      "loss": 0.4022,
+      "step": 34243
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.685872480318367e-07,
+      "loss": 0.3477,
+      "step": 34244
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.680104968711251e-07,
+      "loss": 0.5554,
+      "step": 34245
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.674337457104133e-07,
+      "loss": 0.4038,
+      "step": 34246
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.668569945497016e-07,
+      "loss": 0.5556,
+      "step": 34247
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.662802433889898e-07,
+      "loss": 0.3442,
+      "step": 34248
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.657034922282782e-07,
+      "loss": 0.4316,
+      "step": 34249
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.651267410675665e-07,
+      "loss": 0.4425,
+      "step": 34250
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.645499899068548e-07,
+      "loss": 0.4091,
+      "step": 34251
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.63973238746143e-07,
+      "loss": 0.3542,
+      "step": 34252
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.633964875854313e-07,
+      "loss": 0.4232,
+      "step": 34253
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.628197364247195e-07,
+      "loss": 0.3552,
+      "step": 34254
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.622429852640079e-07,
+      "loss": 0.5335,
+      "step": 34255
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.616662341032963e-07,
+      "loss": 0.4371,
+      "step": 34256
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.610894829425845e-07,
+      "loss": 0.3879,
+      "step": 34257
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.605127317818728e-07,
+      "loss": 0.3522,
+      "step": 34258
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.59935980621161e-07,
+      "loss": 0.3421,
+      "step": 34259
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.593592294604494e-07,
+      "loss": 0.4784,
+      "step": 34260
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.587824782997377e-07,
+      "loss": 0.4286,
+      "step": 34261
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.58205727139026e-07,
+      "loss": 0.4111,
+      "step": 34262
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.576289759783142e-07,
+      "loss": 0.3484,
+      "step": 34263
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.570522248176025e-07,
+      "loss": 0.41,
+      "step": 34264
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.564754736568907e-07,
+      "loss": 0.3479,
+      "step": 34265
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.558987224961791e-07,
+      "loss": 0.2199,
+      "step": 34266
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.553219713354674e-07,
+      "loss": 0.3604,
+      "step": 34267
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.547452201747557e-07,
+      "loss": 0.4334,
+      "step": 34268
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.541684690140439e-07,
+      "loss": 0.379,
+      "step": 34269
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.535917178533322e-07,
+      "loss": 0.3721,
+      "step": 34270
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.530149666926206e-07,
+      "loss": 0.351,
+      "step": 34271
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.524382155319089e-07,
+      "loss": 0.3083,
+      "step": 34272
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.518614643711972e-07,
+      "loss": 0.3869,
+      "step": 34273
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.512847132104854e-07,
+      "loss": 0.459,
+      "step": 34274
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.507079620497736e-07,
+      "loss": 0.4469,
+      "step": 34275
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.501312108890619e-07,
+      "loss": 0.422,
+      "step": 34276
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.495544597283504e-07,
+      "loss": 0.3957,
+      "step": 34277
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.489777085676386e-07,
+      "loss": 0.3967,
+      "step": 34278
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.484009574069269e-07,
+      "loss": 0.3646,
+      "step": 34279
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.478242062462151e-07,
+      "loss": 0.3002,
+      "step": 34280
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.472474550855034e-07,
+      "loss": 0.3603,
+      "step": 34281
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.466707039247917e-07,
+      "loss": 0.325,
+      "step": 34282
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.460939527640801e-07,
+      "loss": 0.4299,
+      "step": 34283
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.455172016033683e-07,
+      "loss": 0.4941,
+      "step": 34284
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.449404504426566e-07,
+      "loss": 0.3268,
+      "step": 34285
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.443636992819448e-07,
+      "loss": 0.409,
+      "step": 34286
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.437869481212331e-07,
+      "loss": 0.5332,
+      "step": 34287
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.432101969605215e-07,
+      "loss": 0.4071,
+      "step": 34288
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.426334457998098e-07,
+      "loss": 0.3028,
+      "step": 34289
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.42056694639098e-07,
+      "loss": 0.3506,
+      "step": 34290
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.414799434783863e-07,
+      "loss": 0.4656,
+      "step": 34291
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.409031923176745e-07,
+      "loss": 0.4678,
+      "step": 34292
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.40326441156963e-07,
+      "loss": 0.4142,
+      "step": 34293
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.397496899962513e-07,
+      "loss": 0.5221,
+      "step": 34294
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.391729388355395e-07,
+      "loss": 0.3288,
+      "step": 34295
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.385961876748278e-07,
+      "loss": 0.4156,
+      "step": 34296
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.38019436514116e-07,
+      "loss": 0.3924,
+      "step": 34297
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.374426853534042e-07,
+      "loss": 0.414,
+      "step": 34298
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.368659341926927e-07,
+      "loss": 0.317,
+      "step": 34299
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.36289183031981e-07,
+      "loss": 0.3656,
+      "step": 34300
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.357124318712692e-07,
+      "loss": 0.3761,
+      "step": 34301
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.351356807105575e-07,
+      "loss": 0.3985,
+      "step": 34302
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.345589295498457e-07,
+      "loss": 0.3935,
+      "step": 34303
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.339821783891342e-07,
+      "loss": 0.3884,
+      "step": 34304
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.334054272284224e-07,
+      "loss": 0.3752,
+      "step": 34305
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.328286760677107e-07,
+      "loss": 0.4453,
+      "step": 34306
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.322519249069989e-07,
+      "loss": 0.3703,
+      "step": 34307
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.316751737462872e-07,
+      "loss": 0.4304,
+      "step": 34308
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.310984225855754e-07,
+      "loss": 0.5893,
+      "step": 34309
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.305216714248639e-07,
+      "loss": 0.3687,
+      "step": 34310
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.299449202641521e-07,
+      "loss": 0.465,
+      "step": 34311
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.293681691034404e-07,
+      "loss": 0.336,
+      "step": 34312
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.287914179427286e-07,
+      "loss": 0.3161,
+      "step": 34313
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.282146667820169e-07,
+      "loss": 0.2955,
+      "step": 34314
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.276379156213052e-07,
+      "loss": 0.4297,
+      "step": 34315
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.270611644605936e-07,
+      "loss": 0.3775,
+      "step": 34316
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.264844132998819e-07,
+      "loss": 0.4874,
+      "step": 34317
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.259076621391701e-07,
+      "loss": 0.3323,
+      "step": 34318
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.253309109784584e-07,
+      "loss": 0.3422,
+      "step": 34319
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.247541598177466e-07,
+      "loss": 0.4599,
+      "step": 34320
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.241774086570351e-07,
+      "loss": 0.3151,
+      "step": 34321
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.236006574963233e-07,
+      "loss": 0.3698,
+      "step": 34322
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.230239063356116e-07,
+      "loss": 0.4155,
+      "step": 34323
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.224471551748998e-07,
+      "loss": 0.338,
+      "step": 34324
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.218704040141881e-07,
+      "loss": 0.3774,
+      "step": 34325
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.212936528534764e-07,
+      "loss": 0.4878,
+      "step": 34326
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.207169016927648e-07,
+      "loss": 0.5044,
+      "step": 34327
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.20140150532053e-07,
+      "loss": 0.4791,
+      "step": 34328
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.195633993713413e-07,
+      "loss": 0.3664,
+      "step": 34329
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.189866482106295e-07,
+      "loss": 0.4407,
+      "step": 34330
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.184098970499179e-07,
+      "loss": 0.3542,
+      "step": 34331
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.178331458892063e-07,
+      "loss": 0.4131,
+      "step": 34332
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.172563947284945e-07,
+      "loss": 0.3327,
+      "step": 34333
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.166796435677827e-07,
+      "loss": 0.3795,
+      "step": 34334
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.16102892407071e-07,
+      "loss": 0.3326,
+      "step": 34335
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.155261412463592e-07,
+      "loss": 0.4136,
+      "step": 34336
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.149493900856476e-07,
+      "loss": 0.3964,
+      "step": 34337
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.14372638924936e-07,
+      "loss": 0.2997,
+      "step": 34338
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.137958877642242e-07,
+      "loss": 0.4788,
+      "step": 34339
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.132191366035125e-07,
+      "loss": 0.4025,
+      "step": 34340
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.126423854428007e-07,
+      "loss": 0.4303,
+      "step": 34341
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.120656342820891e-07,
+      "loss": 0.4097,
+      "step": 34342
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.114888831213774e-07,
+      "loss": 0.4444,
+      "step": 34343
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.109121319606657e-07,
+      "loss": 0.4042,
+      "step": 34344
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.103353807999539e-07,
+      "loss": 0.44,
+      "step": 34345
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.097586296392422e-07,
+      "loss": 0.3762,
+      "step": 34346
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.091818784785304e-07,
+      "loss": 0.3788,
+      "step": 34347
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.086051273178188e-07,
+      "loss": 0.4715,
+      "step": 34348
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.080283761571071e-07,
+      "loss": 0.3311,
+      "step": 34349
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.074516249963954e-07,
+      "loss": 0.3764,
+      "step": 34350
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.068748738356836e-07,
+      "loss": 0.4945,
+      "step": 34351
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.062981226749719e-07,
+      "loss": 0.3469,
+      "step": 34352
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.057213715142602e-07,
+      "loss": 0.3134,
+      "step": 34353
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.051446203535486e-07,
+      "loss": 0.3673,
+      "step": 34354
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.045678691928369e-07,
+      "loss": 0.1489,
+      "step": 34355
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.039911180321251e-07,
+      "loss": 0.366,
+      "step": 34356
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.034143668714133e-07,
+      "loss": 0.3495,
+      "step": 34357
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.028376157107017e-07,
+      "loss": 0.4575,
+      "step": 34358
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.022608645499899e-07,
+      "loss": 0.3637,
+      "step": 34359
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.016841133892783e-07,
+      "loss": 0.3727,
+      "step": 34360
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.011073622285666e-07,
+      "loss": 0.355,
+      "step": 34361
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 8.005306110678548e-07,
+      "loss": 0.3909,
+      "step": 34362
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.999538599071432e-07,
+      "loss": 0.4428,
+      "step": 34363
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.993771087464314e-07,
+      "loss": 0.4419,
+      "step": 34364
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.988003575857197e-07,
+      "loss": 0.4153,
+      "step": 34365
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.98223606425008e-07,
+      "loss": 0.3843,
+      "step": 34366
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.976468552642963e-07,
+      "loss": 0.3374,
+      "step": 34367
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.970701041035845e-07,
+      "loss": 0.3088,
+      "step": 34368
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.964933529428729e-07,
+      "loss": 0.4622,
+      "step": 34369
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.959166017821611e-07,
+      "loss": 0.4467,
+      "step": 34370
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.953398506214495e-07,
+      "loss": 0.4622,
+      "step": 34371
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.947630994607377e-07,
+      "loss": 0.4898,
+      "step": 34372
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.94186348300026e-07,
+      "loss": 0.4617,
+      "step": 34373
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.936095971393143e-07,
+      "loss": 0.3254,
+      "step": 34374
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.930328459786026e-07,
+      "loss": 0.4063,
+      "step": 34375
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.924560948178908e-07,
+      "loss": 0.3845,
+      "step": 34376
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.918793436571792e-07,
+      "loss": 0.4653,
+      "step": 34377
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.913025924964675e-07,
+      "loss": 0.3813,
+      "step": 34378
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.907258413357557e-07,
+      "loss": 0.4364,
+      "step": 34379
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.90149090175044e-07,
+      "loss": 0.3118,
+      "step": 34380
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.895723390143323e-07,
+      "loss": 0.3773,
+      "step": 34381
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.889955878536207e-07,
+      "loss": 0.4533,
+      "step": 34382
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.884188366929089e-07,
+      "loss": 0.404,
+      "step": 34383
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.878420855321972e-07,
+      "loss": 0.399,
+      "step": 34384
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.872653343714855e-07,
+      "loss": 0.3648,
+      "step": 34385
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.866885832107738e-07,
+      "loss": 0.3753,
+      "step": 34386
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.86111832050062e-07,
+      "loss": 0.2848,
+      "step": 34387
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.855350808893504e-07,
+      "loss": 0.4388,
+      "step": 34388
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.849583297286386e-07,
+      "loss": 0.4524,
+      "step": 34389
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.84381578567927e-07,
+      "loss": 0.2584,
+      "step": 34390
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.838048274072152e-07,
+      "loss": 0.4582,
+      "step": 34391
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.832280762465035e-07,
+      "loss": 0.5173,
+      "step": 34392
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.826513250857918e-07,
+      "loss": 0.3941,
+      "step": 34393
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.820745739250801e-07,
+      "loss": 0.4015,
+      "step": 34394
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.814978227643683e-07,
+      "loss": 0.3461,
+      "step": 34395
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.809210716036567e-07,
+      "loss": 0.364,
+      "step": 34396
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.803443204429449e-07,
+      "loss": 0.397,
+      "step": 34397
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.797675692822332e-07,
+      "loss": 0.3584,
+      "step": 34398
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.791908181215216e-07,
+      "loss": 0.2902,
+      "step": 34399
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 7.786140669608098e-07,
+      "loss": 0.2814,
+      "step": 34400
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 35750,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 2,
+  "save_steps": 192608170000.0,
+  "total_flos": 1.9795437252074268e+24,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}