diff --git "a/checkpoint-20000/trainer_state.json" "b/checkpoint-20000/trainer_state.json"
new file mode 100644--- /dev/null
+++ "b/checkpoint-20000/trainer_state.json"
@@ -0,0 +1,120021 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.6957368723148905,
+  "eval_steps": 500,
+  "global_step": 20000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3174971031286212e-08,
+      "loss": 4.331,
+      "step": 1
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.6349942062572424e-08,
+      "loss": 4.1443,
+      "step": 2
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.952491309385863e-08,
+      "loss": 4.3343,
+      "step": 3
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.269988412514485e-08,
+      "loss": 4.2615,
+      "step": 4
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1587485515643108e-07,
+      "loss": 4.0824,
+      "step": 5
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3904982618771727e-07,
+      "loss": 4.1118,
+      "step": 6
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6222479721900348e-07,
+      "loss": 4.1886,
+      "step": 7
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.853997682502897e-07,
+      "loss": 4.2092,
+      "step": 8
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.085747392815759e-07,
+      "loss": 4.1067,
+      "step": 9
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3174971031286215e-07,
+      "loss": 4.1774,
+      "step": 10
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.549246813441483e-07,
+      "loss": 4.4161,
+      "step": 11
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7809965237543453e-07,
+      "loss": 4.375,
+      "step": 12
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0127462340672075e-07,
+      "loss": 4.3113,
+      "step": 13
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2444959443800696e-07,
+      "loss": 4.148,
+      "step": 14
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.4762456546929323e-07,
+      "loss": 4.1949,
+      "step": 15
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.707995365005794e-07,
+      "loss": 4.177,
+      "step": 16
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.939745075318656e-07,
+      "loss": 4.1539,
+      "step": 17
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.171494785631518e-07,
+      "loss": 4.1751,
+      "step": 18
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.4032444959443804e-07,
+      "loss": 4.375,
+      "step": 19
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.634994206257243e-07,
+      "loss": 3.711,
+      "step": 20
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 4.866743916570104e-07,
+      "loss": 3.9811,
+      "step": 21
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.098493626882966e-07,
+      "loss": 3.7419,
+      "step": 22
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.33024333719583e-07,
+      "loss": 3.7219,
+      "step": 23
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.561993047508691e-07,
+      "loss": 3.7399,
+      "step": 24
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 5.793742757821554e-07,
+      "loss": 3.7148,
+      "step": 25
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.025492468134415e-07,
+      "loss": 3.7045,
+      "step": 26
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.257242178447277e-07,
+      "loss": 3.4791,
+      "step": 27
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.488991888760139e-07,
+      "loss": 2.9588,
+      "step": 28
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.720741599073002e-07,
+      "loss": 2.81,
+      "step": 29
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 6.952491309385865e-07,
+      "loss": 2.9663,
+      "step": 30
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.184241019698726e-07,
+      "loss": 2.8941,
+      "step": 31
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.415990730011588e-07,
+      "loss": 3.0079,
+      "step": 32
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.64774044032445e-07,
+      "loss": 2.7529,
+      "step": 33
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 7.879490150637312e-07,
+      "loss": 2.7569,
+      "step": 34
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.111239860950175e-07,
+      "loss": 2.8331,
+      "step": 35
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.342989571263037e-07,
+      "loss": 2.7629,
+      "step": 36
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.574739281575899e-07,
+      "loss": 2.6754,
+      "step": 37
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 8.806488991888761e-07,
+      "loss": 2.4825,
+      "step": 38
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.038238702201622e-07,
+      "loss": 2.3577,
+      "step": 39
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.269988412514486e-07,
+      "loss": 2.0555,
+      "step": 40
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.501738122827347e-07,
+      "loss": 2.1276,
+      "step": 41
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.733487833140208e-07,
+      "loss": 2.1449,
+      "step": 42
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 9.96523754345307e-07,
+      "loss": 1.8829,
+      "step": 43
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0196987253765933e-06,
+      "loss": 2.0511,
+      "step": 44
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.0428736964078797e-06,
+      "loss": 1.9704,
+      "step": 45
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.066048667439166e-06,
+      "loss": 2.0253,
+      "step": 46
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.089223638470452e-06,
+      "loss": 1.9847,
+      "step": 47
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1123986095017381e-06,
+      "loss": 1.9195,
+      "step": 48
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1355735805330243e-06,
+      "loss": 1.9268,
+      "step": 49
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.1587485515643108e-06,
+      "loss": 1.9207,
+      "step": 50
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.181923522595597e-06,
+      "loss": 1.879,
+      "step": 51
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.205098493626883e-06,
+      "loss": 1.7418,
+      "step": 52
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2282734646581692e-06,
+      "loss": 1.7899,
+      "step": 53
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2514484356894554e-06,
+      "loss": 1.781,
+      "step": 54
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2746234067207416e-06,
+      "loss": 1.7557,
+      "step": 55
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.2977983777520279e-06,
+      "loss": 1.8487,
+      "step": 56
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.320973348783314e-06,
+      "loss": 1.7376,
+      "step": 57
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3441483198146005e-06,
+      "loss": 1.6492,
+      "step": 58
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.3673232908458867e-06,
+      "loss": 1.6299,
+      "step": 59
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.390498261877173e-06,
+      "loss": 1.6794,
+      "step": 60
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.413673232908459e-06,
+      "loss": 1.6868,
+      "step": 61
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4368482039397451e-06,
+      "loss": 1.611,
+      "step": 62
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4600231749710314e-06,
+      "loss": 1.7118,
+      "step": 63
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.4831981460023176e-06,
+      "loss": 1.6163,
+      "step": 64
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5063731170336038e-06,
+      "loss": 1.5804,
+      "step": 65
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.52954808806489e-06,
+      "loss": 1.7108,
+      "step": 66
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5527230590961762e-06,
+      "loss": 1.5946,
+      "step": 67
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5758980301274624e-06,
+      "loss": 1.6452,
+      "step": 68
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.5990730011587489e-06,
+      "loss": 1.6334,
+      "step": 69
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.622247972190035e-06,
+      "loss": 1.5181,
+      "step": 70
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.645422943221321e-06,
+      "loss": 1.6854,
+      "step": 71
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6685979142526073e-06,
+      "loss": 1.6866,
+      "step": 72
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.6917728852838935e-06,
+      "loss": 1.5256,
+      "step": 73
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7149478563151797e-06,
+      "loss": 1.4504,
+      "step": 74
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.738122827346466e-06,
+      "loss": 1.5092,
+      "step": 75
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7612977983777522e-06,
+      "loss": 1.4502,
+      "step": 76
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.7844727694090384e-06,
+      "loss": 1.5657,
+      "step": 77
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8076477404403244e-06,
+      "loss": 1.5777,
+      "step": 78
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.830822711471611e-06,
+      "loss": 1.4891,
+      "step": 79
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8539976825028972e-06,
+      "loss": 1.4015,
+      "step": 80
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.8771726535341832e-06,
+      "loss": 1.5116,
+      "step": 81
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9003476245654695e-06,
+      "loss": 1.4748,
+      "step": 82
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9235225955967557e-06,
+      "loss": 1.5411,
+      "step": 83
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.9466975666280417e-06,
+      "loss": 1.4901,
+      "step": 84
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.969872537659328e-06,
+      "loss": 1.473,
+      "step": 85
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 1.993047508690614e-06,
+      "loss": 1.4419,
+      "step": 86
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0162224797219005e-06,
+      "loss": 1.4667,
+      "step": 87
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0393974507531865e-06,
+      "loss": 1.4351,
+      "step": 88
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.062572421784473e-06,
+      "loss": 1.4754,
+      "step": 89
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.0857473928157594e-06,
+      "loss": 1.6083,
+      "step": 90
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.1089223638470454e-06,
+      "loss": 1.5417,
+      "step": 91
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.132097334878332e-06,
+      "loss": 1.471,
+      "step": 92
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.155272305909618e-06,
+      "loss": 1.5131,
+      "step": 93
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.178447276940904e-06,
+      "loss": 1.4507,
+      "step": 94
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2016222479721903e-06,
+      "loss": 1.5716,
+      "step": 95
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2247972190034763e-06,
+      "loss": 1.4328,
+      "step": 96
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2479721900347627e-06,
+      "loss": 1.5551,
+      "step": 97
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.2711471610660487e-06,
+      "loss": 1.4151,
+      "step": 98
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.294322132097335e-06,
+      "loss": 1.5183,
+      "step": 99
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3174971031286215e-06,
+      "loss": 1.4731,
+      "step": 100
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.3406720741599076e-06,
+      "loss": 1.4112,
+      "step": 101
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.363847045191194e-06,
+      "loss": 1.3972,
+      "step": 102
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.38702201622248e-06,
+      "loss": 1.423,
+      "step": 103
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.410196987253766e-06,
+      "loss": 1.6216,
+      "step": 104
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4333719582850524e-06,
+      "loss": 1.4468,
+      "step": 105
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.4565469293163384e-06,
+      "loss": 1.4796,
+      "step": 106
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.479721900347625e-06,
+      "loss": 1.4696,
+      "step": 107
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.502896871378911e-06,
+      "loss": 1.3445,
+      "step": 108
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5260718424101973e-06,
+      "loss": 1.4379,
+      "step": 109
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5492468134414833e-06,
+      "loss": 1.4166,
+      "step": 110
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5724217844727693e-06,
+      "loss": 1.5335,
+      "step": 111
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.5955967555040557e-06,
+      "loss": 1.4822,
+      "step": 112
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.6187717265353417e-06,
+      "loss": 1.3208,
+      "step": 113
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.641946697566628e-06,
+      "loss": 1.4495,
+      "step": 114
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.665121668597914e-06,
+      "loss": 1.2604,
+      "step": 115
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.688296639629201e-06,
+      "loss": 1.4519,
+      "step": 116
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.711471610660487e-06,
+      "loss": 1.4941,
+      "step": 117
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7346465816917734e-06,
+      "loss": 1.3101,
+      "step": 118
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.7578215527230594e-06,
+      "loss": 1.3204,
+      "step": 119
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.780996523754346e-06,
+      "loss": 1.3517,
+      "step": 120
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.804171494785632e-06,
+      "loss": 1.4944,
+      "step": 121
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.827346465816918e-06,
+      "loss": 1.3268,
+      "step": 122
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8505214368482043e-06,
+      "loss": 1.3229,
+      "step": 123
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8736964078794903e-06,
+      "loss": 1.3578,
+      "step": 124
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.8968713789107767e-06,
+      "loss": 1.4631,
+      "step": 125
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9200463499420627e-06,
+      "loss": 1.4886,
+      "step": 126
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.943221320973349e-06,
+      "loss": 1.3679,
+      "step": 127
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.966396292004635e-06,
+      "loss": 1.2978,
+      "step": 128
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 2.9895712630359216e-06,
+      "loss": 1.4547,
+      "step": 129
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0127462340672076e-06,
+      "loss": 1.3934,
+      "step": 130
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.0359212050984936e-06,
+      "loss": 1.3027,
+      "step": 131
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.05909617612978e-06,
+      "loss": 1.3163,
+      "step": 132
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.082271147161066e-06,
+      "loss": 1.4777,
+      "step": 133
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1054461181923525e-06,
+      "loss": 1.4758,
+      "step": 134
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1286210892236385e-06,
+      "loss": 1.5456,
+      "step": 135
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.151796060254925e-06,
+      "loss": 1.383,
+      "step": 136
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1749710312862113e-06,
+      "loss": 1.5003,
+      "step": 137
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.1981460023174977e-06,
+      "loss": 1.4451,
+      "step": 138
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.2213209733487837e-06,
+      "loss": 1.2943,
+      "step": 139
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.24449594438007e-06,
+      "loss": 1.457,
+      "step": 140
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.267670915411356e-06,
+      "loss": 1.4172,
+      "step": 141
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.290845886442642e-06,
+      "loss": 1.3365,
+      "step": 142
+    },
+    {
+      "epoch": 0.0,
+      "learning_rate": 3.3140208574739286e-06,
+      "loss": 1.3542,
+      "step": 143
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.3371958285052146e-06,
+      "loss": 1.3233,
+      "step": 144
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.360370799536501e-06,
+      "loss": 1.3646,
+      "step": 145
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.383545770567787e-06,
+      "loss": 1.4117,
+      "step": 146
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4067207415990735e-06,
+      "loss": 1.2635,
+      "step": 147
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4298957126303595e-06,
+      "loss": 1.4685,
+      "step": 148
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.4530706836616455e-06,
+      "loss": 1.2896,
+      "step": 149
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.476245654692932e-06,
+      "loss": 1.366,
+      "step": 150
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.499420625724218e-06,
+      "loss": 1.2317,
+      "step": 151
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5225955967555043e-06,
+      "loss": 1.3525,
+      "step": 152
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5457705677867903e-06,
+      "loss": 1.3392,
+      "step": 153
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5689455388180768e-06,
+      "loss": 1.3983,
+      "step": 154
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.5921205098493628e-06,
+      "loss": 1.3681,
+      "step": 155
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.6152954808806488e-06,
+      "loss": 1.2763,
+      "step": 156
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.638470451911935e-06,
+      "loss": 1.3988,
+      "step": 157
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.661645422943222e-06,
+      "loss": 1.3144,
+      "step": 158
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.684820393974508e-06,
+      "loss": 1.3876,
+      "step": 159
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7079953650057945e-06,
+      "loss": 1.2529,
+      "step": 160
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7311703360370805e-06,
+      "loss": 1.3259,
+      "step": 161
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.7543453070683665e-06,
+      "loss": 1.3387,
+      "step": 162
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.777520278099653e-06,
+      "loss": 1.3119,
+      "step": 163
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.800695249130939e-06,
+      "loss": 1.3617,
+      "step": 164
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.823870220162225e-06,
+      "loss": 1.3279,
+      "step": 165
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.847045191193511e-06,
+      "loss": 1.3199,
+      "step": 166
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.870220162224797e-06,
+      "loss": 1.323,
+      "step": 167
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.893395133256083e-06,
+      "loss": 1.4096,
+      "step": 168
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.91657010428737e-06,
+      "loss": 1.1955,
+      "step": 169
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.939745075318656e-06,
+      "loss": 1.211,
+      "step": 170
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.962920046349942e-06,
+      "loss": 1.3134,
+      "step": 171
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 3.986095017381228e-06,
+      "loss": 1.2566,
+      "step": 172
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.009269988412515e-06,
+      "loss": 1.4465,
+      "step": 173
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.032444959443801e-06,
+      "loss": 1.4253,
+      "step": 174
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.055619930475087e-06,
+      "loss": 1.2931,
+      "step": 175
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.078794901506373e-06,
+      "loss": 1.407,
+      "step": 176
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.101969872537659e-06,
+      "loss": 1.2553,
+      "step": 177
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.125144843568946e-06,
+      "loss": 1.3596,
+      "step": 178
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.148319814600233e-06,
+      "loss": 1.3583,
+      "step": 179
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.171494785631519e-06,
+      "loss": 1.1883,
+      "step": 180
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.194669756662805e-06,
+      "loss": 1.2345,
+      "step": 181
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.217844727694091e-06,
+      "loss": 1.2118,
+      "step": 182
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.241019698725377e-06,
+      "loss": 1.2084,
+      "step": 183
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.264194669756664e-06,
+      "loss": 1.3394,
+      "step": 184
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.28736964078795e-06,
+      "loss": 1.3872,
+      "step": 185
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.310544611819236e-06,
+      "loss": 1.2985,
+      "step": 186
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.333719582850522e-06,
+      "loss": 1.4032,
+      "step": 187
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.356894553881808e-06,
+      "loss": 1.3743,
+      "step": 188
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.3800695249130945e-06,
+      "loss": 1.4178,
+      "step": 189
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4032444959443805e-06,
+      "loss": 1.3476,
+      "step": 190
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4264194669756665e-06,
+      "loss": 1.324,
+      "step": 191
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.4495944380069525e-06,
+      "loss": 1.3169,
+      "step": 192
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.472769409038239e-06,
+      "loss": 1.2274,
+      "step": 193
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.495944380069525e-06,
+      "loss": 1.2428,
+      "step": 194
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.519119351100811e-06,
+      "loss": 1.2079,
+      "step": 195
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.542294322132097e-06,
+      "loss": 1.2342,
+      "step": 196
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.565469293163383e-06,
+      "loss": 1.3472,
+      "step": 197
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.58864426419467e-06,
+      "loss": 1.2774,
+      "step": 198
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.611819235225956e-06,
+      "loss": 1.2213,
+      "step": 199
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.634994206257243e-06,
+      "loss": 1.4101,
+      "step": 200
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.658169177288529e-06,
+      "loss": 1.364,
+      "step": 201
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.681344148319815e-06,
+      "loss": 1.2526,
+      "step": 202
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.704519119351101e-06,
+      "loss": 1.1154,
+      "step": 203
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.727694090382388e-06,
+      "loss": 1.3616,
+      "step": 204
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.750869061413674e-06,
+      "loss": 1.1447,
+      "step": 205
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.77404403244496e-06,
+      "loss": 1.2978,
+      "step": 206
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.797219003476246e-06,
+      "loss": 1.296,
+      "step": 207
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.820393974507532e-06,
+      "loss": 1.3098,
+      "step": 208
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.843568945538819e-06,
+      "loss": 1.4279,
+      "step": 209
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.866743916570105e-06,
+      "loss": 1.224,
+      "step": 210
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.889918887601391e-06,
+      "loss": 1.2611,
+      "step": 211
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.913093858632677e-06,
+      "loss": 1.2993,
+      "step": 212
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.936268829663963e-06,
+      "loss": 1.3809,
+      "step": 213
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.95944380069525e-06,
+      "loss": 1.2763,
+      "step": 214
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 4.982618771726536e-06,
+      "loss": 1.2674,
+      "step": 215
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.005793742757822e-06,
+      "loss": 1.256,
+      "step": 216
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0289687137891085e-06,
+      "loss": 1.1977,
+      "step": 217
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0521436848203946e-06,
+      "loss": 1.3078,
+      "step": 218
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0753186558516806e-06,
+      "loss": 1.1963,
+      "step": 219
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.0984936268829666e-06,
+      "loss": 1.1974,
+      "step": 220
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.121668597914253e-06,
+      "loss": 1.2663,
+      "step": 221
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.1448435689455386e-06,
+      "loss": 1.2195,
+      "step": 222
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.168018539976825e-06,
+      "loss": 1.2194,
+      "step": 223
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.191193511008111e-06,
+      "loss": 1.3938,
+      "step": 224
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.214368482039398e-06,
+      "loss": 1.2542,
+      "step": 225
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.2375434530706834e-06,
+      "loss": 1.309,
+      "step": 226
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.26071842410197e-06,
+      "loss": 1.4199,
+      "step": 227
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.283893395133256e-06,
+      "loss": 1.2687,
+      "step": 228
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.307068366164543e-06,
+      "loss": 1.1901,
+      "step": 229
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.330243337195828e-06,
+      "loss": 1.2215,
+      "step": 230
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.353418308227115e-06,
+      "loss": 1.3511,
+      "step": 231
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.376593279258402e-06,
+      "loss": 1.222,
+      "step": 232
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.399768250289687e-06,
+      "loss": 1.2128,
+      "step": 233
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.422943221320974e-06,
+      "loss": 1.3293,
+      "step": 234
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.44611819235226e-06,
+      "loss": 1.2664,
+      "step": 235
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.469293163383547e-06,
+      "loss": 1.2495,
+      "step": 236
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.492468134414832e-06,
+      "loss": 1.1753,
+      "step": 237
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.515643105446119e-06,
+      "loss": 1.2558,
+      "step": 238
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.538818076477405e-06,
+      "loss": 1.2174,
+      "step": 239
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.561993047508692e-06,
+      "loss": 1.3259,
+      "step": 240
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.585168018539977e-06,
+      "loss": 1.2661,
+      "step": 241
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.608342989571264e-06,
+      "loss": 1.2546,
+      "step": 242
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.63151796060255e-06,
+      "loss": 1.2338,
+      "step": 243
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.654692931633836e-06,
+      "loss": 1.2415,
+      "step": 244
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.677867902665122e-06,
+      "loss": 1.3597,
+      "step": 245
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.701042873696409e-06,
+      "loss": 1.2764,
+      "step": 246
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.724217844727694e-06,
+      "loss": 1.3041,
+      "step": 247
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.747392815758981e-06,
+      "loss": 1.2378,
+      "step": 248
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.770567786790267e-06,
+      "loss": 1.3366,
+      "step": 249
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.7937427578215534e-06,
+      "loss": 1.314,
+      "step": 250
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.816917728852839e-06,
+      "loss": 1.3427,
+      "step": 251
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.8400926998841255e-06,
+      "loss": 1.4306,
+      "step": 252
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.863267670915412e-06,
+      "loss": 1.2206,
+      "step": 253
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.886442641946698e-06,
+      "loss": 1.197,
+      "step": 254
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.909617612977985e-06,
+      "loss": 1.1663,
+      "step": 255
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.93279258400927e-06,
+      "loss": 1.0977,
+      "step": 256
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.955967555040557e-06,
+      "loss": 1.3843,
+      "step": 257
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 5.979142526071843e-06,
+      "loss": 1.2093,
+      "step": 258
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.002317497103129e-06,
+      "loss": 1.3903,
+      "step": 259
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.025492468134415e-06,
+      "loss": 1.3856,
+      "step": 260
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.048667439165702e-06,
+      "loss": 1.2102,
+      "step": 261
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.071842410196987e-06,
+      "loss": 1.2195,
+      "step": 262
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.095017381228274e-06,
+      "loss": 1.2127,
+      "step": 263
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.11819235225956e-06,
+      "loss": 1.2707,
+      "step": 264
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.141367323290847e-06,
+      "loss": 1.3381,
+      "step": 265
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.164542294322132e-06,
+      "loss": 1.144,
+      "step": 266
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.187717265353419e-06,
+      "loss": 1.2813,
+      "step": 267
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.210892236384705e-06,
+      "loss": 1.275,
+      "step": 268
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.234067207415992e-06,
+      "loss": 1.3214,
+      "step": 269
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.257242178447277e-06,
+      "loss": 1.3547,
+      "step": 270
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.280417149478564e-06,
+      "loss": 1.2512,
+      "step": 271
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.30359212050985e-06,
+      "loss": 1.2169,
+      "step": 272
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.326767091541136e-06,
+      "loss": 1.3818,
+      "step": 273
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.349942062572423e-06,
+      "loss": 1.1657,
+      "step": 274
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.373117033603709e-06,
+      "loss": 1.2702,
+      "step": 275
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.3962920046349955e-06,
+      "loss": 1.2579,
+      "step": 276
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.419466975666281e-06,
+      "loss": 1.2152,
+      "step": 277
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4426419466975675e-06,
+      "loss": 1.295,
+      "step": 278
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.4658169177288535e-06,
+      "loss": 1.2144,
+      "step": 279
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.48899188876014e-06,
+      "loss": 1.195,
+      "step": 280
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.5121668597914255e-06,
+      "loss": 1.3117,
+      "step": 281
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.535341830822712e-06,
+      "loss": 1.2747,
+      "step": 282
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.558516801853998e-06,
+      "loss": 1.3421,
+      "step": 283
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.581691772885284e-06,
+      "loss": 1.2658,
+      "step": 284
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.60486674391657e-06,
+      "loss": 1.2378,
+      "step": 285
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.628041714947857e-06,
+      "loss": 1.2319,
+      "step": 286
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.651216685979142e-06,
+      "loss": 1.1425,
+      "step": 287
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.674391657010429e-06,
+      "loss": 1.1886,
+      "step": 288
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.697566628041715e-06,
+      "loss": 1.3463,
+      "step": 289
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.720741599073002e-06,
+      "loss": 1.1561,
+      "step": 290
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.743916570104287e-06,
+      "loss": 1.3277,
+      "step": 291
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.767091541135574e-06,
+      "loss": 1.0838,
+      "step": 292
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.79026651216686e-06,
+      "loss": 1.2278,
+      "step": 293
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.813441483198147e-06,
+      "loss": 1.2119,
+      "step": 294
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.836616454229433e-06,
+      "loss": 1.204,
+      "step": 295
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.859791425260719e-06,
+      "loss": 1.2598,
+      "step": 296
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.882966396292006e-06,
+      "loss": 1.2218,
+      "step": 297
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.906141367323291e-06,
+      "loss": 1.2813,
+      "step": 298
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.929316338354578e-06,
+      "loss": 1.3206,
+      "step": 299
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.952491309385864e-06,
+      "loss": 1.2501,
+      "step": 300
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.975666280417151e-06,
+      "loss": 1.2238,
+      "step": 301
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.998841251448436e-06,
+      "loss": 1.2749,
+      "step": 302
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.022016222479723e-06,
+      "loss": 1.0769,
+      "step": 303
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.045191193511009e-06,
+      "loss": 1.2558,
+      "step": 304
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.0683661645422955e-06,
+      "loss": 1.3472,
+      "step": 305
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.091541135573581e-06,
+      "loss": 1.3524,
+      "step": 306
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1147161066048675e-06,
+      "loss": 1.2932,
+      "step": 307
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1378910776361535e-06,
+      "loss": 1.19,
+      "step": 308
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1610660486674395e-06,
+      "loss": 1.2316,
+      "step": 309
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.1842410196987255e-06,
+      "loss": 1.1326,
+      "step": 310
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.207415990730012e-06,
+      "loss": 1.2482,
+      "step": 311
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.2305909617612975e-06,
+      "loss": 1.2347,
+      "step": 312
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.253765932792584e-06,
+      "loss": 1.3884,
+      "step": 313
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.27694090382387e-06,
+      "loss": 1.2716,
+      "step": 314
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.300115874855157e-06,
+      "loss": 1.2873,
+      "step": 315
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.323290845886444e-06,
+      "loss": 1.2994,
+      "step": 316
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.346465816917729e-06,
+      "loss": 1.1863,
+      "step": 317
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.369640787949016e-06,
+      "loss": 1.1461,
+      "step": 318
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.392815758980302e-06,
+      "loss": 1.1764,
+      "step": 319
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.415990730011589e-06,
+      "loss": 1.374,
+      "step": 320
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.439165701042874e-06,
+      "loss": 1.2571,
+      "step": 321
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.462340672074161e-06,
+      "loss": 1.3222,
+      "step": 322
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.485515643105446e-06,
+      "loss": 1.1194,
+      "step": 323
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.508690614136733e-06,
+      "loss": 1.2224,
+      "step": 324
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.531865585168019e-06,
+      "loss": 1.3594,
+      "step": 325
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.555040556199306e-06,
+      "loss": 1.2318,
+      "step": 326
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.578215527230591e-06,
+      "loss": 1.2939,
+      "step": 327
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.601390498261878e-06,
+      "loss": 1.1865,
+      "step": 328
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.624565469293164e-06,
+      "loss": 1.2992,
+      "step": 329
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.64774044032445e-06,
+      "loss": 1.1826,
+      "step": 330
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.670915411355736e-06,
+      "loss": 1.2005,
+      "step": 331
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.694090382387023e-06,
+      "loss": 1.4001,
+      "step": 332
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.717265353418308e-06,
+      "loss": 1.2739,
+      "step": 333
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.740440324449595e-06,
+      "loss": 1.2228,
+      "step": 334
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.763615295480882e-06,
+      "loss": 1.2789,
+      "step": 335
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.786790266512167e-06,
+      "loss": 1.2346,
+      "step": 336
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.809965237543454e-06,
+      "loss": 1.2563,
+      "step": 337
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.83314020857474e-06,
+      "loss": 1.1487,
+      "step": 338
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.856315179606027e-06,
+      "loss": 1.1666,
+      "step": 339
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.879490150637312e-06,
+      "loss": 1.3232,
+      "step": 340
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.9026651216686e-06,
+      "loss": 1.2362,
+      "step": 341
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.925840092699884e-06,
+      "loss": 1.2199,
+      "step": 342
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.949015063731171e-06,
+      "loss": 1.2432,
+      "step": 343
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.972190034762456e-06,
+      "loss": 1.2211,
+      "step": 344
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.995365005793743e-06,
+      "loss": 1.2846,
+      "step": 345
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.01853997682503e-06,
+      "loss": 1.3266,
+      "step": 346
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.041714947856315e-06,
+      "loss": 1.0763,
+      "step": 347
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.064889918887602e-06,
+      "loss": 1.1457,
+      "step": 348
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.088064889918889e-06,
+      "loss": 1.1522,
+      "step": 349
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.111239860950174e-06,
+      "loss": 1.2994,
+      "step": 350
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.134414831981461e-06,
+      "loss": 1.2386,
+      "step": 351
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.157589803012746e-06,
+      "loss": 1.3012,
+      "step": 352
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.180764774044033e-06,
+      "loss": 1.2759,
+      "step": 353
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.203939745075318e-06,
+      "loss": 1.2821,
+      "step": 354
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.227114716106605e-06,
+      "loss": 1.3713,
+      "step": 355
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.250289687137892e-06,
+      "loss": 1.4134,
+      "step": 356
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.273464658169179e-06,
+      "loss": 1.1621,
+      "step": 357
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.296639629200466e-06,
+      "loss": 1.1288,
+      "step": 358
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.31981460023175e-06,
+      "loss": 1.3591,
+      "step": 359
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.342989571263038e-06,
+      "loss": 1.1823,
+      "step": 360
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.366164542294323e-06,
+      "loss": 1.2759,
+      "step": 361
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.38933951332561e-06,
+      "loss": 1.2009,
+      "step": 362
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.412514484356895e-06,
+      "loss": 1.2153,
+      "step": 363
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.435689455388182e-06,
+      "loss": 1.3043,
+      "step": 364
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.458864426419467e-06,
+      "loss": 1.2518,
+      "step": 365
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.482039397450754e-06,
+      "loss": 1.2459,
+      "step": 366
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.50521436848204e-06,
+      "loss": 1.4058,
+      "step": 367
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.528389339513327e-06,
+      "loss": 1.271,
+      "step": 368
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.551564310544612e-06,
+      "loss": 1.2738,
+      "step": 369
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.5747392815759e-06,
+      "loss": 1.3313,
+      "step": 370
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.597914252607184e-06,
+      "loss": 1.2573,
+      "step": 371
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.621089223638471e-06,
+      "loss": 1.3021,
+      "step": 372
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.644264194669756e-06,
+      "loss": 1.3416,
+      "step": 373
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.667439165701043e-06,
+      "loss": 1.1423,
+      "step": 374
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.69061413673233e-06,
+      "loss": 1.2571,
+      "step": 375
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.713789107763615e-06,
+      "loss": 1.3715,
+      "step": 376
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.736964078794902e-06,
+      "loss": 1.2848,
+      "step": 377
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.760139049826189e-06,
+      "loss": 1.286,
+      "step": 378
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.783314020857476e-06,
+      "loss": 1.17,
+      "step": 379
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.806488991888761e-06,
+      "loss": 1.2189,
+      "step": 380
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.829663962920048e-06,
+      "loss": 1.156,
+      "step": 381
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.852838933951333e-06,
+      "loss": 1.1054,
+      "step": 382
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.87601390498262e-06,
+      "loss": 1.1919,
+      "step": 383
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.899188876013905e-06,
+      "loss": 1.2503,
+      "step": 384
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.922363847045192e-06,
+      "loss": 1.2174,
+      "step": 385
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.945538818076479e-06,
+      "loss": 1.2408,
+      "step": 386
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.968713789107764e-06,
+      "loss": 1.2825,
+      "step": 387
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 8.99188876013905e-06,
+      "loss": 1.2712,
+      "step": 388
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.015063731170338e-06,
+      "loss": 1.2388,
+      "step": 389
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.038238702201623e-06,
+      "loss": 1.1851,
+      "step": 390
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.06141367323291e-06,
+      "loss": 1.1212,
+      "step": 391
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.084588644264195e-06,
+      "loss": 1.328,
+      "step": 392
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.107763615295482e-06,
+      "loss": 1.2091,
+      "step": 393
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.130938586326767e-06,
+      "loss": 1.1803,
+      "step": 394
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.154113557358054e-06,
+      "loss": 1.2003,
+      "step": 395
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.17728852838934e-06,
+      "loss": 1.3538,
+      "step": 396
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.200463499420627e-06,
+      "loss": 1.2539,
+      "step": 397
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.223638470451912e-06,
+      "loss": 1.3982,
+      "step": 398
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.2468134414832e-06,
+      "loss": 1.1922,
+      "step": 399
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.269988412514486e-06,
+      "loss": 1.0989,
+      "step": 400
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.293163383545771e-06,
+      "loss": 1.1027,
+      "step": 401
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.316338354577058e-06,
+      "loss": 1.2419,
+      "step": 402
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.339513325608343e-06,
+      "loss": 1.3572,
+      "step": 403
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.36268829663963e-06,
+      "loss": 1.1874,
+      "step": 404
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.385863267670915e-06,
+      "loss": 1.1213,
+      "step": 405
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.409038238702202e-06,
+      "loss": 1.1555,
+      "step": 406
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.432213209733489e-06,
+      "loss": 1.2364,
+      "step": 407
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.455388180764776e-06,
+      "loss": 1.2551,
+      "step": 408
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.478563151796061e-06,
+      "loss": 1.25,
+      "step": 409
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.501738122827348e-06,
+      "loss": 1.2159,
+      "step": 410
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.524913093858633e-06,
+      "loss": 1.4126,
+      "step": 411
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.54808806488992e-06,
+      "loss": 1.1719,
+      "step": 412
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.571263035921205e-06,
+      "loss": 1.1493,
+      "step": 413
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.594438006952492e-06,
+      "loss": 1.2618,
+      "step": 414
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.617612977983777e-06,
+      "loss": 1.2438,
+      "step": 415
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.640787949015064e-06,
+      "loss": 1.1576,
+      "step": 416
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.66396292004635e-06,
+      "loss": 1.1366,
+      "step": 417
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.687137891077638e-06,
+      "loss": 1.1961,
+      "step": 418
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.710312862108923e-06,
+      "loss": 1.2343,
+      "step": 419
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.73348783314021e-06,
+      "loss": 1.149,
+      "step": 420
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.756662804171497e-06,
+      "loss": 1.3491,
+      "step": 421
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.779837775202782e-06,
+      "loss": 1.1381,
+      "step": 422
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.803012746234069e-06,
+      "loss": 1.2721,
+      "step": 423
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.826187717265354e-06,
+      "loss": 1.2457,
+      "step": 424
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.84936268829664e-06,
+      "loss": 1.2742,
+      "step": 425
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.872537659327926e-06,
+      "loss": 1.127,
+      "step": 426
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.895712630359213e-06,
+      "loss": 1.1352,
+      "step": 427
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.9188876013905e-06,
+      "loss": 1.2484,
+      "step": 428
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.942062572421786e-06,
+      "loss": 1.2262,
+      "step": 429
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.965237543453071e-06,
+      "loss": 1.186,
+      "step": 430
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 9.988412514484358e-06,
+      "loss": 1.2427,
+      "step": 431
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0011587485515643e-05,
+      "loss": 1.265,
+      "step": 432
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0034762456546929e-05,
+      "loss": 1.1791,
+      "step": 433
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0057937427578217e-05,
+      "loss": 1.2196,
+      "step": 434
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0081112398609502e-05,
+      "loss": 1.1266,
+      "step": 435
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0104287369640789e-05,
+      "loss": 1.2165,
+      "step": 436
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0127462340672074e-05,
+      "loss": 1.2016,
+      "step": 437
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0150637311703361e-05,
+      "loss": 1.2934,
+      "step": 438
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0173812282734648e-05,
+      "loss": 1.197,
+      "step": 439
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0196987253765933e-05,
+      "loss": 1.374,
+      "step": 440
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0220162224797218e-05,
+      "loss": 1.3707,
+      "step": 441
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0243337195828507e-05,
+      "loss": 1.1192,
+      "step": 442
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0266512166859792e-05,
+      "loss": 1.2127,
+      "step": 443
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0289687137891077e-05,
+      "loss": 1.2768,
+      "step": 444
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0312862108922366e-05,
+      "loss": 1.3152,
+      "step": 445
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.033603707995365e-05,
+      "loss": 1.3584,
+      "step": 446
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0359212050984938e-05,
+      "loss": 1.1822,
+      "step": 447
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0382387022016223e-05,
+      "loss": 1.2084,
+      "step": 448
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.040556199304751e-05,
+      "loss": 1.1878,
+      "step": 449
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0428736964078797e-05,
+      "loss": 1.3507,
+      "step": 450
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0451911935110082e-05,
+      "loss": 1.1811,
+      "step": 451
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0475086906141367e-05,
+      "loss": 1.3342,
+      "step": 452
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0498261877172655e-05,
+      "loss": 1.1947,
+      "step": 453
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.052143684820394e-05,
+      "loss": 1.2386,
+      "step": 454
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0544611819235226e-05,
+      "loss": 1.2322,
+      "step": 455
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0567786790266513e-05,
+      "loss": 1.2011,
+      "step": 456
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.05909617612978e-05,
+      "loss": 1.329,
+      "step": 457
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0614136732329086e-05,
+      "loss": 1.3236,
+      "step": 458
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0637311703360371e-05,
+      "loss": 1.1773,
+      "step": 459
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0660486674391657e-05,
+      "loss": 1.2949,
+      "step": 460
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0683661645422945e-05,
+      "loss": 1.4357,
+      "step": 461
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.070683661645423e-05,
+      "loss": 1.11,
+      "step": 462
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0730011587485515e-05,
+      "loss": 1.3087,
+      "step": 463
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0753186558516804e-05,
+      "loss": 1.1575,
+      "step": 464
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0776361529548089e-05,
+      "loss": 1.0626,
+      "step": 465
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0799536500579374e-05,
+      "loss": 1.0906,
+      "step": 466
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0822711471610661e-05,
+      "loss": 1.2195,
+      "step": 467
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0845886442641948e-05,
+      "loss": 1.3958,
+      "step": 468
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0869061413673235e-05,
+      "loss": 1.2755,
+      "step": 469
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.089223638470452e-05,
+      "loss": 1.2384,
+      "step": 470
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0915411355735805e-05,
+      "loss": 1.4126,
+      "step": 471
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0938586326767094e-05,
+      "loss": 1.1886,
+      "step": 472
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0961761297798379e-05,
+      "loss": 1.0915,
+      "step": 473
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.0984936268829664e-05,
+      "loss": 1.2959,
+      "step": 474
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1008111239860951e-05,
+      "loss": 1.1083,
+      "step": 475
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1031286210892238e-05,
+      "loss": 1.1772,
+      "step": 476
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1054461181923523e-05,
+      "loss": 1.1577,
+      "step": 477
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.107763615295481e-05,
+      "loss": 1.2426,
+      "step": 478
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1100811123986095e-05,
+      "loss": 1.2709,
+      "step": 479
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1123986095017383e-05,
+      "loss": 1.0812,
+      "step": 480
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1147161066048669e-05,
+      "loss": 1.2771,
+      "step": 481
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1170336037079954e-05,
+      "loss": 1.2645,
+      "step": 482
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1193511008111239e-05,
+      "loss": 1.1445,
+      "step": 483
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1216685979142527e-05,
+      "loss": 1.1601,
+      "step": 484
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1239860950173813e-05,
+      "loss": 1.2854,
+      "step": 485
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.12630359212051e-05,
+      "loss": 1.1835,
+      "step": 486
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1286210892236386e-05,
+      "loss": 1.3512,
+      "step": 487
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1309385863267671e-05,
+      "loss": 1.2079,
+      "step": 488
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1332560834298958e-05,
+      "loss": 1.255,
+      "step": 489
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1355735805330243e-05,
+      "loss": 1.3812,
+      "step": 490
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1378910776361532e-05,
+      "loss": 1.2226,
+      "step": 491
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1402085747392817e-05,
+      "loss": 1.2318,
+      "step": 492
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1425260718424102e-05,
+      "loss": 1.3457,
+      "step": 493
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1448435689455387e-05,
+      "loss": 1.2329,
+      "step": 494
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1471610660486676e-05,
+      "loss": 1.192,
+      "step": 495
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1494785631517961e-05,
+      "loss": 1.3504,
+      "step": 496
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1517960602549248e-05,
+      "loss": 1.2561,
+      "step": 497
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1541135573580533e-05,
+      "loss": 1.1249,
+      "step": 498
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1564310544611822e-05,
+      "loss": 1.1469,
+      "step": 499
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1587485515643107e-05,
+      "loss": 1.3383,
+      "step": 500
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1610660486674392e-05,
+      "loss": 1.1702,
+      "step": 501
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1633835457705677e-05,
+      "loss": 1.2509,
+      "step": 502
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1657010428736966e-05,
+      "loss": 1.2859,
+      "step": 503
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1680185399768251e-05,
+      "loss": 1.2048,
+      "step": 504
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1703360370799536e-05,
+      "loss": 1.1733,
+      "step": 505
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1726535341830825e-05,
+      "loss": 1.2526,
+      "step": 506
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.174971031286211e-05,
+      "loss": 1.1868,
+      "step": 507
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1772885283893397e-05,
+      "loss": 1.2694,
+      "step": 508
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1796060254924682e-05,
+      "loss": 1.2105,
+      "step": 509
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.181923522595597e-05,
+      "loss": 1.24,
+      "step": 510
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1842410196987255e-05,
+      "loss": 1.233,
+      "step": 511
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.186558516801854e-05,
+      "loss": 1.1752,
+      "step": 512
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1888760139049826e-05,
+      "loss": 1.1991,
+      "step": 513
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1911935110081114e-05,
+      "loss": 1.3157,
+      "step": 514
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.19351100811124e-05,
+      "loss": 1.1731,
+      "step": 515
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1958285052143686e-05,
+      "loss": 1.2418,
+      "step": 516
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.1981460023174972e-05,
+      "loss": 1.409,
+      "step": 517
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2004634994206258e-05,
+      "loss": 1.2516,
+      "step": 518
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2027809965237545e-05,
+      "loss": 1.1483,
+      "step": 519
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.205098493626883e-05,
+      "loss": 1.27,
+      "step": 520
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2074159907300116e-05,
+      "loss": 1.0864,
+      "step": 521
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2097334878331404e-05,
+      "loss": 1.2635,
+      "step": 522
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.212050984936269e-05,
+      "loss": 1.1913,
+      "step": 523
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2143684820393974e-05,
+      "loss": 1.104,
+      "step": 524
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2166859791425261e-05,
+      "loss": 1.2542,
+      "step": 525
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2190034762456548e-05,
+      "loss": 1.1459,
+      "step": 526
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2213209733487835e-05,
+      "loss": 1.1064,
+      "step": 527
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.223638470451912e-05,
+      "loss": 1.2256,
+      "step": 528
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2259559675550407e-05,
+      "loss": 1.2285,
+      "step": 529
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2282734646581694e-05,
+      "loss": 1.1405,
+      "step": 530
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2305909617612979e-05,
+      "loss": 1.2059,
+      "step": 531
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2329084588644264e-05,
+      "loss": 1.3453,
+      "step": 532
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2352259559675553e-05,
+      "loss": 1.227,
+      "step": 533
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2375434530706838e-05,
+      "loss": 1.2261,
+      "step": 534
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2398609501738123e-05,
+      "loss": 1.22,
+      "step": 535
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.242178447276941e-05,
+      "loss": 1.3093,
+      "step": 536
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2444959443800697e-05,
+      "loss": 1.1704,
+      "step": 537
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2468134414831984e-05,
+      "loss": 1.2756,
+      "step": 538
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2491309385863269e-05,
+      "loss": 1.1522,
+      "step": 539
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2514484356894554e-05,
+      "loss": 1.2829,
+      "step": 540
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2537659327925842e-05,
+      "loss": 1.1881,
+      "step": 541
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2560834298957128e-05,
+      "loss": 1.4517,
+      "step": 542
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2584009269988413e-05,
+      "loss": 1.1895,
+      "step": 543
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.26071842410197e-05,
+      "loss": 1.2109,
+      "step": 544
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2630359212050986e-05,
+      "loss": 1.2928,
+      "step": 545
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2653534183082272e-05,
+      "loss": 1.291,
+      "step": 546
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2676709154113558e-05,
+      "loss": 1.3514,
+      "step": 547
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2699884125144845e-05,
+      "loss": 1.1274,
+      "step": 548
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2723059096176132e-05,
+      "loss": 1.3392,
+      "step": 549
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2746234067207417e-05,
+      "loss": 1.3049,
+      "step": 550
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2769409038238702e-05,
+      "loss": 1.2824,
+      "step": 551
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2792584009269991e-05,
+      "loss": 1.156,
+      "step": 552
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2815758980301276e-05,
+      "loss": 1.1467,
+      "step": 553
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2838933951332561e-05,
+      "loss": 1.2113,
+      "step": 554
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2862108922363848e-05,
+      "loss": 1.3152,
+      "step": 555
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2885283893395135e-05,
+      "loss": 1.3146,
+      "step": 556
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.290845886442642e-05,
+      "loss": 1.3111,
+      "step": 557
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2931633835457707e-05,
+      "loss": 1.1657,
+      "step": 558
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.2954808806488992e-05,
+      "loss": 1.0955,
+      "step": 559
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.297798377752028e-05,
+      "loss": 1.1531,
+      "step": 560
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3001158748551566e-05,
+      "loss": 1.3286,
+      "step": 561
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3024333719582851e-05,
+      "loss": 1.3318,
+      "step": 562
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3047508690614136e-05,
+      "loss": 1.1955,
+      "step": 563
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3070683661645425e-05,
+      "loss": 1.3295,
+      "step": 564
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.309385863267671e-05,
+      "loss": 1.0913,
+      "step": 565
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3117033603707997e-05,
+      "loss": 1.2519,
+      "step": 566
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3140208574739282e-05,
+      "loss": 1.2497,
+      "step": 567
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3163383545770569e-05,
+      "loss": 1.215,
+      "step": 568
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3186558516801856e-05,
+      "loss": 1.2756,
+      "step": 569
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.320973348783314e-05,
+      "loss": 1.1845,
+      "step": 570
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.323290845886443e-05,
+      "loss": 1.2836,
+      "step": 571
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3256083429895714e-05,
+      "loss": 1.3184,
+      "step": 572
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3279258400927e-05,
+      "loss": 1.191,
+      "step": 573
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3302433371958285e-05,
+      "loss": 1.1178,
+      "step": 574
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3325608342989573e-05,
+      "loss": 1.3155,
+      "step": 575
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3348783314020858e-05,
+      "loss": 1.2157,
+      "step": 576
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3371958285052145e-05,
+      "loss": 1.3466,
+      "step": 577
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.339513325608343e-05,
+      "loss": 1.1959,
+      "step": 578
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3418308227114717e-05,
+      "loss": 1.1288,
+      "step": 579
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3441483198146004e-05,
+      "loss": 1.3847,
+      "step": 580
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.346465816917729e-05,
+      "loss": 1.3049,
+      "step": 581
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3487833140208574e-05,
+      "loss": 1.139,
+      "step": 582
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3511008111239863e-05,
+      "loss": 1.3216,
+      "step": 583
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3534183082271148e-05,
+      "loss": 1.3151,
+      "step": 584
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3557358053302433e-05,
+      "loss": 1.3872,
+      "step": 585
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.358053302433372e-05,
+      "loss": 1.2205,
+      "step": 586
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3603707995365007e-05,
+      "loss": 1.172,
+      "step": 587
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3626882966396294e-05,
+      "loss": 1.1892,
+      "step": 588
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3650057937427579e-05,
+      "loss": 1.304,
+      "step": 589
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3673232908458866e-05,
+      "loss": 1.1012,
+      "step": 590
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3696407879490153e-05,
+      "loss": 1.3961,
+      "step": 591
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3719582850521438e-05,
+      "loss": 1.1697,
+      "step": 592
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3742757821552723e-05,
+      "loss": 1.1765,
+      "step": 593
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3765932792584012e-05,
+      "loss": 1.2128,
+      "step": 594
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3789107763615297e-05,
+      "loss": 1.1697,
+      "step": 595
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3812282734646582e-05,
+      "loss": 1.1295,
+      "step": 596
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3835457705677869e-05,
+      "loss": 1.2765,
+      "step": 597
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3858632676709156e-05,
+      "loss": 1.2348,
+      "step": 598
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3881807647740442e-05,
+      "loss": 1.2199,
+      "step": 599
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3904982618771728e-05,
+      "loss": 1.1766,
+      "step": 600
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3928157589803013e-05,
+      "loss": 1.4241,
+      "step": 601
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3951332560834301e-05,
+      "loss": 1.2056,
+      "step": 602
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3974507531865586e-05,
+      "loss": 1.2225,
+      "step": 603
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.3997682502896872e-05,
+      "loss": 1.0769,
+      "step": 604
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4020857473928158e-05,
+      "loss": 1.1364,
+      "step": 605
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4044032444959445e-05,
+      "loss": 1.2734,
+      "step": 606
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.406720741599073e-05,
+      "loss": 1.2298,
+      "step": 607
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4090382387022017e-05,
+      "loss": 1.2697,
+      "step": 608
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4113557358053302e-05,
+      "loss": 1.255,
+      "step": 609
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4136732329084591e-05,
+      "loss": 1.1631,
+      "step": 610
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4159907300115876e-05,
+      "loss": 1.1943,
+      "step": 611
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4183082271147161e-05,
+      "loss": 1.2556,
+      "step": 612
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.420625724217845e-05,
+      "loss": 1.1281,
+      "step": 613
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4229432213209735e-05,
+      "loss": 1.2272,
+      "step": 614
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.425260718424102e-05,
+      "loss": 1.2477,
+      "step": 615
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4275782155272307e-05,
+      "loss": 1.2278,
+      "step": 616
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4298957126303594e-05,
+      "loss": 1.0939,
+      "step": 617
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4322132097334879e-05,
+      "loss": 1.2704,
+      "step": 618
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4345307068366166e-05,
+      "loss": 1.1159,
+      "step": 619
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4368482039397451e-05,
+      "loss": 1.3127,
+      "step": 620
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.439165701042874e-05,
+      "loss": 1.2659,
+      "step": 621
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4414831981460025e-05,
+      "loss": 1.2991,
+      "step": 622
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.443800695249131e-05,
+      "loss": 1.3061,
+      "step": 623
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4461181923522595e-05,
+      "loss": 1.3739,
+      "step": 624
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4484356894553884e-05,
+      "loss": 1.2564,
+      "step": 625
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4507531865585169e-05,
+      "loss": 1.2131,
+      "step": 626
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4530706836616456e-05,
+      "loss": 1.343,
+      "step": 627
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.455388180764774e-05,
+      "loss": 1.2859,
+      "step": 628
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4577056778679028e-05,
+      "loss": 1.3386,
+      "step": 629
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4600231749710314e-05,
+      "loss": 1.3168,
+      "step": 630
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.46234067207416e-05,
+      "loss": 1.2363,
+      "step": 631
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4646581691772888e-05,
+      "loss": 1.1974,
+      "step": 632
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4669756662804173e-05,
+      "loss": 1.0715,
+      "step": 633
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4692931633835459e-05,
+      "loss": 1.0945,
+      "step": 634
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4716106604866744e-05,
+      "loss": 1.2618,
+      "step": 635
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4739281575898032e-05,
+      "loss": 1.2296,
+      "step": 636
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4762456546929317e-05,
+      "loss": 1.1105,
+      "step": 637
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4785631517960604e-05,
+      "loss": 1.3371,
+      "step": 638
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.480880648899189e-05,
+      "loss": 1.1972,
+      "step": 639
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4831981460023178e-05,
+      "loss": 1.2095,
+      "step": 640
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4855156431054463e-05,
+      "loss": 1.1882,
+      "step": 641
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4878331402085748e-05,
+      "loss": 1.2154,
+      "step": 642
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4901506373117033e-05,
+      "loss": 1.2554,
+      "step": 643
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4924681344148322e-05,
+      "loss": 1.2393,
+      "step": 644
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4947856315179607e-05,
+      "loss": 1.2518,
+      "step": 645
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4971031286210892e-05,
+      "loss": 1.1733,
+      "step": 646
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.4994206257242179e-05,
+      "loss": 1.1914,
+      "step": 647
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5017381228273466e-05,
+      "loss": 1.1216,
+      "step": 648
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5040556199304753e-05,
+      "loss": 1.2365,
+      "step": 649
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5063731170336038e-05,
+      "loss": 1.3571,
+      "step": 650
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5086906141367323e-05,
+      "loss": 1.2524,
+      "step": 651
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5110081112398612e-05,
+      "loss": 1.3171,
+      "step": 652
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5133256083429897e-05,
+      "loss": 1.1702,
+      "step": 653
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5156431054461182e-05,
+      "loss": 1.1161,
+      "step": 654
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.517960602549247e-05,
+      "loss": 1.1978,
+      "step": 655
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5202780996523756e-05,
+      "loss": 1.3299,
+      "step": 656
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5225955967555043e-05,
+      "loss": 1.2644,
+      "step": 657
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5249130938586328e-05,
+      "loss": 1.2075,
+      "step": 658
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5272305909617616e-05,
+      "loss": 1.2274,
+      "step": 659
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.52954808806489e-05,
+      "loss": 1.2937,
+      "step": 660
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5318655851680187e-05,
+      "loss": 1.2088,
+      "step": 661
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.534183082271147e-05,
+      "loss": 1.1354,
+      "step": 662
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.536500579374276e-05,
+      "loss": 1.1721,
+      "step": 663
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5388180764774045e-05,
+      "loss": 1.2358,
+      "step": 664
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.541135573580533e-05,
+      "loss": 1.176,
+      "step": 665
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5434530706836616e-05,
+      "loss": 1.1279,
+      "step": 666
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5457705677867904e-05,
+      "loss": 1.1296,
+      "step": 667
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.548088064889919e-05,
+      "loss": 1.1284,
+      "step": 668
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5504055619930475e-05,
+      "loss": 1.3452,
+      "step": 669
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5527230590961763e-05,
+      "loss": 1.127,
+      "step": 670
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5550405561993048e-05,
+      "loss": 1.2249,
+      "step": 671
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5573580533024333e-05,
+      "loss": 1.1445,
+      "step": 672
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5596755504055622e-05,
+      "loss": 1.2892,
+      "step": 673
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5619930475086907e-05,
+      "loss": 1.2462,
+      "step": 674
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5643105446118196e-05,
+      "loss": 1.3188,
+      "step": 675
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.566628041714948e-05,
+      "loss": 1.2418,
+      "step": 676
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5689455388180766e-05,
+      "loss": 1.2275,
+      "step": 677
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5712630359212055e-05,
+      "loss": 1.2184,
+      "step": 678
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.573580533024334e-05,
+      "loss": 1.3112,
+      "step": 679
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5758980301274625e-05,
+      "loss": 1.2364,
+      "step": 680
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.578215527230591e-05,
+      "loss": 1.2646,
+      "step": 681
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.58053302433372e-05,
+      "loss": 1.2406,
+      "step": 682
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5828505214368484e-05,
+      "loss": 1.225,
+      "step": 683
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.585168018539977e-05,
+      "loss": 1.1831,
+      "step": 684
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5874855156431054e-05,
+      "loss": 1.2387,
+      "step": 685
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5898030127462343e-05,
+      "loss": 1.3882,
+      "step": 686
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5921205098493628e-05,
+      "loss": 1.289,
+      "step": 687
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5944380069524913e-05,
+      "loss": 1.3917,
+      "step": 688
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5967555040556198e-05,
+      "loss": 1.2,
+      "step": 689
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.5990730011587487e-05,
+      "loss": 1.0123,
+      "step": 690
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6013904982618772e-05,
+      "loss": 1.1799,
+      "step": 691
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.603707995365006e-05,
+      "loss": 1.2307,
+      "step": 692
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6060254924681345e-05,
+      "loss": 1.2674,
+      "step": 693
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.608342989571263e-05,
+      "loss": 1.1098,
+      "step": 694
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.610660486674392e-05,
+      "loss": 1.1441,
+      "step": 695
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6129779837775204e-05,
+      "loss": 1.2729,
+      "step": 696
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6152954808806493e-05,
+      "loss": 1.3778,
+      "step": 697
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6176129779837778e-05,
+      "loss": 1.2933,
+      "step": 698
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6199304750869063e-05,
+      "loss": 1.2875,
+      "step": 699
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6222479721900348e-05,
+      "loss": 1.2849,
+      "step": 700
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6245654692931637e-05,
+      "loss": 1.2119,
+      "step": 701
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6268829663962922e-05,
+      "loss": 1.237,
+      "step": 702
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6292004634994207e-05,
+      "loss": 1.1899,
+      "step": 703
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6315179606025492e-05,
+      "loss": 1.2581,
+      "step": 704
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.633835457705678e-05,
+      "loss": 1.191,
+      "step": 705
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6361529548088066e-05,
+      "loss": 1.1519,
+      "step": 706
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.638470451911935e-05,
+      "loss": 1.1671,
+      "step": 707
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6407879490150636e-05,
+      "loss": 1.227,
+      "step": 708
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6431054461181925e-05,
+      "loss": 1.148,
+      "step": 709
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.645422943221321e-05,
+      "loss": 1.2253,
+      "step": 710
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6477404403244495e-05,
+      "loss": 1.225,
+      "step": 711
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6500579374275784e-05,
+      "loss": 1.3192,
+      "step": 712
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.652375434530707e-05,
+      "loss": 1.2532,
+      "step": 713
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6546929316338357e-05,
+      "loss": 1.304,
+      "step": 714
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6570104287369643e-05,
+      "loss": 1.3162,
+      "step": 715
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.659327925840093e-05,
+      "loss": 1.2187,
+      "step": 716
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.6616454229432216e-05,
+      "loss": 1.1327,
+      "step": 717
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 1.66396292004635e-05,
+      "loss": 1.3472,
+      "step": 718
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6662804171494787e-05,
+      "loss": 1.2279,
+      "step": 719
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6685979142526075e-05,
+      "loss": 1.2734,
+      "step": 720
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.670915411355736e-05,
+      "loss": 1.2452,
+      "step": 721
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6732329084588645e-05,
+      "loss": 1.1819,
+      "step": 722
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.675550405561993e-05,
+      "loss": 1.2525,
+      "step": 723
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.677867902665122e-05,
+      "loss": 1.1121,
+      "step": 724
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6801853997682504e-05,
+      "loss": 1.3078,
+      "step": 725
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.682502896871379e-05,
+      "loss": 1.1586,
+      "step": 726
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6848203939745075e-05,
+      "loss": 1.2535,
+      "step": 727
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6871378910776363e-05,
+      "loss": 1.255,
+      "step": 728
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.689455388180765e-05,
+      "loss": 1.3167,
+      "step": 729
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6917728852838933e-05,
+      "loss": 1.2487,
+      "step": 730
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6940903823870222e-05,
+      "loss": 1.1791,
+      "step": 731
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6964078794901507e-05,
+      "loss": 1.1872,
+      "step": 732
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.6987253765932796e-05,
+      "loss": 1.0884,
+      "step": 733
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.701042873696408e-05,
+      "loss": 1.1615,
+      "step": 734
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7033603707995366e-05,
+      "loss": 1.1527,
+      "step": 735
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7056778679026655e-05,
+      "loss": 1.2627,
+      "step": 736
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.707995365005794e-05,
+      "loss": 1.2236,
+      "step": 737
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7103128621089225e-05,
+      "loss": 1.3329,
+      "step": 738
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7126303592120513e-05,
+      "loss": 1.1912,
+      "step": 739
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.71494785631518e-05,
+      "loss": 1.2182,
+      "step": 740
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7172653534183084e-05,
+      "loss": 1.279,
+      "step": 741
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.719582850521437e-05,
+      "loss": 1.3116,
+      "step": 742
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7219003476245657e-05,
+      "loss": 1.2473,
+      "step": 743
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7242178447276943e-05,
+      "loss": 1.0933,
+      "step": 744
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7265353418308228e-05,
+      "loss": 1.3257,
+      "step": 745
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7288528389339513e-05,
+      "loss": 1.1668,
+      "step": 746
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.73117033603708e-05,
+      "loss": 1.1363,
+      "step": 747
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7334878331402087e-05,
+      "loss": 1.3069,
+      "step": 748
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7358053302433372e-05,
+      "loss": 1.3082,
+      "step": 749
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.738122827346466e-05,
+      "loss": 1.1473,
+      "step": 750
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7404403244495946e-05,
+      "loss": 1.1757,
+      "step": 751
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.742757821552723e-05,
+      "loss": 1.3049,
+      "step": 752
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.745075318655852e-05,
+      "loss": 1.2678,
+      "step": 753
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7473928157589804e-05,
+      "loss": 1.272,
+      "step": 754
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7497103128621093e-05,
+      "loss": 1.2668,
+      "step": 755
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7520278099652378e-05,
+      "loss": 1.1606,
+      "step": 756
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7543453070683663e-05,
+      "loss": 1.3804,
+      "step": 757
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7566628041714952e-05,
+      "loss": 1.2803,
+      "step": 758
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7589803012746237e-05,
+      "loss": 1.4519,
+      "step": 759
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7612977983777522e-05,
+      "loss": 1.2835,
+      "step": 760
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7636152954808807e-05,
+      "loss": 1.1756,
+      "step": 761
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7659327925840096e-05,
+      "loss": 1.2756,
+      "step": 762
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.768250289687138e-05,
+      "loss": 1.0478,
+      "step": 763
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7705677867902666e-05,
+      "loss": 1.2258,
+      "step": 764
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.772885283893395e-05,
+      "loss": 1.1886,
+      "step": 765
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.775202780996524e-05,
+      "loss": 1.1108,
+      "step": 766
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7775202780996525e-05,
+      "loss": 1.3664,
+      "step": 767
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.779837775202781e-05,
+      "loss": 1.2906,
+      "step": 768
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7821552723059095e-05,
+      "loss": 1.3345,
+      "step": 769
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7844727694090384e-05,
+      "loss": 1.1992,
+      "step": 770
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.786790266512167e-05,
+      "loss": 1.1707,
+      "step": 771
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7891077636152958e-05,
+      "loss": 1.2474,
+      "step": 772
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7914252607184243e-05,
+      "loss": 1.2699,
+      "step": 773
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7937427578215528e-05,
+      "loss": 1.3194,
+      "step": 774
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.7960602549246816e-05,
+      "loss": 1.1484,
+      "step": 775
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.79837775202781e-05,
+      "loss": 1.3034,
+      "step": 776
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8006952491309387e-05,
+      "loss": 1.2677,
+      "step": 777
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8030127462340675e-05,
+      "loss": 1.2094,
+      "step": 778
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.805330243337196e-05,
+      "loss": 1.1195,
+      "step": 779
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8076477404403246e-05,
+      "loss": 1.2398,
+      "step": 780
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8099652375434534e-05,
+      "loss": 1.399,
+      "step": 781
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.812282734646582e-05,
+      "loss": 1.084,
+      "step": 782
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8146002317497104e-05,
+      "loss": 1.2813,
+      "step": 783
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.816917728852839e-05,
+      "loss": 1.2718,
+      "step": 784
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8192352259559678e-05,
+      "loss": 1.2848,
+      "step": 785
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8215527230590963e-05,
+      "loss": 1.2942,
+      "step": 786
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.823870220162225e-05,
+      "loss": 1.1597,
+      "step": 787
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8261877172653534e-05,
+      "loss": 1.3942,
+      "step": 788
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8285052143684822e-05,
+      "loss": 1.3332,
+      "step": 789
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8308227114716107e-05,
+      "loss": 1.1556,
+      "step": 790
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8331402085747392e-05,
+      "loss": 1.1674,
+      "step": 791
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.835457705677868e-05,
+      "loss": 1.2879,
+      "step": 792
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8377752027809966e-05,
+      "loss": 1.3483,
+      "step": 793
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8400926998841255e-05,
+      "loss": 1.2127,
+      "step": 794
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.842410196987254e-05,
+      "loss": 1.1822,
+      "step": 795
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8447276940903825e-05,
+      "loss": 1.2595,
+      "step": 796
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8470451911935114e-05,
+      "loss": 1.1626,
+      "step": 797
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.84936268829664e-05,
+      "loss": 1.2544,
+      "step": 798
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8516801853997684e-05,
+      "loss": 1.1996,
+      "step": 799
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8539976825028972e-05,
+      "loss": 1.2049,
+      "step": 800
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8563151796060258e-05,
+      "loss": 1.3082,
+      "step": 801
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8586326767091543e-05,
+      "loss": 1.1913,
+      "step": 802
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8609501738122828e-05,
+      "loss": 1.1941,
+      "step": 803
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8632676709154116e-05,
+      "loss": 1.1271,
+      "step": 804
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.86558516801854e-05,
+      "loss": 1.2331,
+      "step": 805
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8679026651216687e-05,
+      "loss": 1.0685,
+      "step": 806
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8702201622247972e-05,
+      "loss": 1.3959,
+      "step": 807
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.872537659327926e-05,
+      "loss": 1.2842,
+      "step": 808
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8748551564310546e-05,
+      "loss": 1.1866,
+      "step": 809
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.877172653534183e-05,
+      "loss": 1.2341,
+      "step": 810
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.879490150637312e-05,
+      "loss": 1.1601,
+      "step": 811
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8818076477404404e-05,
+      "loss": 1.1941,
+      "step": 812
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.884125144843569e-05,
+      "loss": 1.1267,
+      "step": 813
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8864426419466978e-05,
+      "loss": 1.2761,
+      "step": 814
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8887601390498263e-05,
+      "loss": 1.1118,
+      "step": 815
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8910776361529552e-05,
+      "loss": 1.3469,
+      "step": 816
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8933951332560837e-05,
+      "loss": 1.3707,
+      "step": 817
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8957126303592122e-05,
+      "loss": 1.1746,
+      "step": 818
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.8980301274623407e-05,
+      "loss": 1.2224,
+      "step": 819
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9003476245654696e-05,
+      "loss": 1.2455,
+      "step": 820
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.902665121668598e-05,
+      "loss": 1.2265,
+      "step": 821
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9049826187717266e-05,
+      "loss": 1.215,
+      "step": 822
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9073001158748555e-05,
+      "loss": 1.2527,
+      "step": 823
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.909617612977984e-05,
+      "loss": 1.1776,
+      "step": 824
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9119351100811125e-05,
+      "loss": 1.2486,
+      "step": 825
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.914252607184241e-05,
+      "loss": 1.2041,
+      "step": 826
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.91657010428737e-05,
+      "loss": 1.1765,
+      "step": 827
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9188876013904984e-05,
+      "loss": 1.3034,
+      "step": 828
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.921205098493627e-05,
+      "loss": 1.3496,
+      "step": 829
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9235225955967554e-05,
+      "loss": 1.2521,
+      "step": 830
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9258400926998843e-05,
+      "loss": 1.2155,
+      "step": 831
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9281575898030128e-05,
+      "loss": 1.1472,
+      "step": 832
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9304750869061416e-05,
+      "loss": 1.2835,
+      "step": 833
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.93279258400927e-05,
+      "loss": 1.1578,
+      "step": 834
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9351100811123987e-05,
+      "loss": 1.187,
+      "step": 835
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9374275782155275e-05,
+      "loss": 1.1735,
+      "step": 836
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.939745075318656e-05,
+      "loss": 1.1863,
+      "step": 837
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9420625724217846e-05,
+      "loss": 1.161,
+      "step": 838
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9443800695249134e-05,
+      "loss": 1.0821,
+      "step": 839
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.946697566628042e-05,
+      "loss": 1.269,
+      "step": 840
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9490150637311704e-05,
+      "loss": 1.2607,
+      "step": 841
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9513325608342993e-05,
+      "loss": 1.3457,
+      "step": 842
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9536500579374278e-05,
+      "loss": 1.1421,
+      "step": 843
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9559675550405563e-05,
+      "loss": 1.2807,
+      "step": 844
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.958285052143685e-05,
+      "loss": 1.3072,
+      "step": 845
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9606025492468137e-05,
+      "loss": 1.191,
+      "step": 846
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9629200463499422e-05,
+      "loss": 1.2983,
+      "step": 847
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9652375434530707e-05,
+      "loss": 1.3099,
+      "step": 848
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9675550405561993e-05,
+      "loss": 1.1707,
+      "step": 849
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.969872537659328e-05,
+      "loss": 1.2565,
+      "step": 850
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9721900347624566e-05,
+      "loss": 1.3269,
+      "step": 851
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.974507531865585e-05,
+      "loss": 1.2272,
+      "step": 852
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.976825028968714e-05,
+      "loss": 1.1616,
+      "step": 853
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9791425260718425e-05,
+      "loss": 1.2061,
+      "step": 854
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9814600231749714e-05,
+      "loss": 1.1621,
+      "step": 855
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9837775202781e-05,
+      "loss": 1.2071,
+      "step": 856
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9860950173812284e-05,
+      "loss": 1.2492,
+      "step": 857
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9884125144843572e-05,
+      "loss": 1.0957,
+      "step": 858
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9907300115874858e-05,
+      "loss": 1.3049,
+      "step": 859
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9930475086906143e-05,
+      "loss": 1.2199,
+      "step": 860
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9953650057937428e-05,
+      "loss": 1.1197,
+      "step": 861
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9976825028968716e-05,
+      "loss": 1.2833,
+      "step": 862
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 2e-05,
+      "loss": 1.1996,
+      "step": 863
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999936526747e-05,
+      "loss": 1.2613,
+      "step": 864
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999746106992e-05,
+      "loss": 1.2907,
+      "step": 865
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999999428740735e-05,
+      "loss": 1.2987,
+      "step": 866
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999999898442798e-05,
+      "loss": 1.2085,
+      "step": 867
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999998413168736e-05,
+      "loss": 1.1778,
+      "step": 868
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999997714963e-05,
+      "loss": 1.2145,
+      "step": 869
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999996889810795e-05,
+      "loss": 1.1439,
+      "step": 870
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999995937712125e-05,
+      "loss": 1.0203,
+      "step": 871
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999994858666998e-05,
+      "loss": 1.1971,
+      "step": 872
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999993652675433e-05,
+      "loss": 1.3784,
+      "step": 873
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999992319737448e-05,
+      "loss": 1.295,
+      "step": 874
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999990859853053e-05,
+      "loss": 1.2208,
+      "step": 875
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999989273022268e-05,
+      "loss": 1.188,
+      "step": 876
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999987559245114e-05,
+      "loss": 1.1517,
+      "step": 877
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999985718521614e-05,
+      "loss": 1.317,
+      "step": 878
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999983750851792e-05,
+      "loss": 1.1638,
+      "step": 879
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999981656235672e-05,
+      "loss": 1.2722,
+      "step": 880
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999979434673277e-05,
+      "loss": 1.2465,
+      "step": 881
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999977086164642e-05,
+      "loss": 1.1243,
+      "step": 882
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999997461070979e-05,
+      "loss": 1.2222,
+      "step": 883
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999972008308758e-05,
+      "loss": 1.207,
+      "step": 884
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999969278961577e-05,
+      "loss": 1.1027,
+      "step": 885
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996642266828e-05,
+      "loss": 1.2635,
+      "step": 886
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999963439428906e-05,
+      "loss": 1.2579,
+      "step": 887
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999996032924349e-05,
+      "loss": 1.2093,
+      "step": 888
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999957092112076e-05,
+      "loss": 1.2073,
+      "step": 889
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999953728034698e-05,
+      "loss": 1.3724,
+      "step": 890
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999950237011404e-05,
+      "loss": 1.1551,
+      "step": 891
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999994661904224e-05,
+      "loss": 1.1161,
+      "step": 892
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999942874127244e-05,
+      "loss": 1.1263,
+      "step": 893
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999939002266472e-05,
+      "loss": 1.0824,
+      "step": 894
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999993500345997e-05,
+      "loss": 1.1538,
+      "step": 895
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999930877707788e-05,
+      "loss": 1.2478,
+      "step": 896
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999926625009975e-05,
+      "loss": 1.1143,
+      "step": 897
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999922245366595e-05,
+      "loss": 1.1809,
+      "step": 898
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999917738777693e-05,
+      "loss": 1.1312,
+      "step": 899
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999913105243332e-05,
+      "loss": 1.1349,
+      "step": 900
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999990834476357e-05,
+      "loss": 1.1927,
+      "step": 901
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999903457338465e-05,
+      "loss": 1.1677,
+      "step": 902
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999898442968083e-05,
+      "loss": 1.0813,
+      "step": 903
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999893301652484e-05,
+      "loss": 1.1917,
+      "step": 904
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999888033391738e-05,
+      "loss": 1.1709,
+      "step": 905
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999882638185907e-05,
+      "loss": 1.0928,
+      "step": 906
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999987711603506e-05,
+      "loss": 1.107,
+      "step": 907
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999871466939267e-05,
+      "loss": 1.285,
+      "step": 908
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999865690898604e-05,
+      "loss": 1.2167,
+      "step": 909
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999985978791314e-05,
+      "loss": 1.1431,
+      "step": 910
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999853757982953e-05,
+      "loss": 1.2832,
+      "step": 911
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999984760110812e-05,
+      "loss": 1.1284,
+      "step": 912
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999984131728871e-05,
+      "loss": 1.1549,
+      "step": 913
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999834906524815e-05,
+      "loss": 1.1584,
+      "step": 914
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982836881651e-05,
+      "loss": 1.3204,
+      "step": 915
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999982170416388e-05,
+      "loss": 1.3,
+      "step": 916
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999814912567005e-05,
+      "loss": 1.1667,
+      "step": 917
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999807994025978e-05,
+      "loss": 1.2544,
+      "step": 918
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999800948540886e-05,
+      "loss": 1.0847,
+      "step": 919
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999793776111812e-05,
+      "loss": 1.1222,
+      "step": 920
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999786476738853e-05,
+      "loss": 1.0658,
+      "step": 921
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99997790504221e-05,
+      "loss": 1.1931,
+      "step": 922
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999977149716165e-05,
+      "loss": 1.3188,
+      "step": 923
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999976381695759e-05,
+      "loss": 1.0531,
+      "step": 924
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999756009810028e-05,
+      "loss": 1.2484,
+      "step": 925
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999974807571906e-05,
+      "loss": 1.17,
+      "step": 926
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999974001468478e-05,
+      "loss": 1.1535,
+      "step": 927
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.99997318267073e-05,
+      "loss": 1.3533,
+      "step": 928
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999723511786716e-05,
+      "loss": 1.1768,
+      "step": 929
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999971506992314e-05,
+      "loss": 1.2114,
+      "step": 930
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999970650111668e-05,
+      "loss": 1.2389,
+      "step": 931
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999697805367437e-05,
+      "loss": 1.2499,
+      "step": 932
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999688982675524e-05,
+      "loss": 1.147,
+      "step": 933
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999680033041057e-05,
+      "loss": 1.3151,
+      "step": 934
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999670956464148e-05,
+      "loss": 1.1745,
+      "step": 935
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999966175294491e-05,
+      "loss": 1.1204,
+      "step": 936
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999652422483466e-05,
+      "loss": 1.169,
+      "step": 937
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999642965079923e-05,
+      "loss": 1.1785,
+      "step": 938
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999633380734413e-05,
+      "loss": 1.1812,
+      "step": 939
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999623669447045e-05,
+      "loss": 1.2342,
+      "step": 940
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999613831217957e-05,
+      "loss": 1.1225,
+      "step": 941
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999603866047264e-05,
+      "loss": 1.348,
+      "step": 942
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999593773935097e-05,
+      "loss": 1.2059,
+      "step": 943
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999583554881578e-05,
+      "loss": 1.2105,
+      "step": 944
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999573208886843e-05,
+      "loss": 1.2324,
+      "step": 945
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999562735951024e-05,
+      "loss": 1.1582,
+      "step": 946
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999955213607425e-05,
+      "loss": 1.2089,
+      "step": 947
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999541409256656e-05,
+      "loss": 1.3254,
+      "step": 948
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999953055549838e-05,
+      "loss": 1.1955,
+      "step": 949
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999951957479956e-05,
+      "loss": 1.0923,
+      "step": 950
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999508467160326e-05,
+      "loss": 1.1778,
+      "step": 951
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999497232580837e-05,
+      "loss": 1.2797,
+      "step": 952
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999485871061218e-05,
+      "loss": 1.1906,
+      "step": 953
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999474382601627e-05,
+      "loss": 1.0612,
+      "step": 954
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999462767202203e-05,
+      "loss": 1.2593,
+      "step": 955
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999451024863093e-05,
+      "loss": 1.145,
+      "step": 956
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999943915558445e-05,
+      "loss": 1.1383,
+      "step": 957
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999427159366417e-05,
+      "loss": 1.1602,
+      "step": 958
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999415036209156e-05,
+      "loss": 1.3563,
+      "step": 959
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999402786112812e-05,
+      "loss": 1.2406,
+      "step": 960
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999939040907755e-05,
+      "loss": 1.2211,
+      "step": 961
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999377905103518e-05,
+      "loss": 1.2018,
+      "step": 962
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999365274190882e-05,
+      "loss": 1.1388,
+      "step": 963
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999352516339794e-05,
+      "loss": 1.1577,
+      "step": 964
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999339631550425e-05,
+      "loss": 1.2316,
+      "step": 965
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999326619822933e-05,
+      "loss": 1.1167,
+      "step": 966
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999313481157485e-05,
+      "loss": 1.1554,
+      "step": 967
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999930021555425e-05,
+      "loss": 1.2092,
+      "step": 968
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999286823013392e-05,
+      "loss": 1.2115,
+      "step": 969
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999927330353508e-05,
+      "loss": 1.2089,
+      "step": 970
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999925965711949e-05,
+      "loss": 1.2011,
+      "step": 971
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999245883766797e-05,
+      "loss": 1.2057,
+      "step": 972
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999923198347717e-05,
+      "loss": 1.2072,
+      "step": 973
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999217956250788e-05,
+      "loss": 1.2593,
+      "step": 974
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999920380208783e-05,
+      "loss": 1.2857,
+      "step": 975
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999189520988475e-05,
+      "loss": 1.1419,
+      "step": 976
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999175112952902e-05,
+      "loss": 1.2363,
+      "step": 977
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999160577981297e-05,
+      "loss": 1.2848,
+      "step": 978
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999145916073847e-05,
+      "loss": 1.1961,
+      "step": 979
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999913112723073e-05,
+      "loss": 1.3292,
+      "step": 980
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999911621145214e-05,
+      "loss": 1.0994,
+      "step": 981
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999101168738263e-05,
+      "loss": 1.1508,
+      "step": 982
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999085999089295e-05,
+      "loss": 1.1744,
+      "step": 983
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999070702505424e-05,
+      "loss": 1.2824,
+      "step": 984
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999055278986842e-05,
+      "loss": 1.1478,
+      "step": 985
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999039728533752e-05,
+      "loss": 1.0227,
+      "step": 986
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999024051146348e-05,
+      "loss": 1.1399,
+      "step": 987
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9999008246824826e-05,
+      "loss": 1.3923,
+      "step": 988
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998992315569393e-05,
+      "loss": 1.286,
+      "step": 989
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998976257380245e-05,
+      "loss": 1.1558,
+      "step": 990
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999896007225759e-05,
+      "loss": 1.1074,
+      "step": 991
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999894376020163e-05,
+      "loss": 1.1887,
+      "step": 992
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998927321212576e-05,
+      "loss": 1.2681,
+      "step": 993
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998910755290636e-05,
+      "loss": 1.2628,
+      "step": 994
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998894062436014e-05,
+      "loss": 1.2225,
+      "step": 995
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999887724264893e-05,
+      "loss": 1.1469,
+      "step": 996
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998860295929595e-05,
+      "loss": 1.2345,
+      "step": 997
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999884322227822e-05,
+      "loss": 1.1236,
+      "step": 998
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998826021695028e-05,
+      "loss": 1.2921,
+      "step": 999
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998808694180235e-05,
+      "loss": 1.2393,
+      "step": 1000
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999879123973406e-05,
+      "loss": 1.0472,
+      "step": 1001
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998773658356723e-05,
+      "loss": 1.124,
+      "step": 1002
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998755950048453e-05,
+      "loss": 1.2475,
+      "step": 1003
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998738114809467e-05,
+      "loss": 1.0872,
+      "step": 1004
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.999872015264e-05,
+      "loss": 1.2218,
+      "step": 1005
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 1.9998702063540272e-05,
+      "loss": 1.2646,
+      "step": 1006
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998683847510517e-05,
+      "loss": 1.3117,
+      "step": 1007
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998665504550967e-05,
+      "loss": 1.2338,
+      "step": 1008
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998647034661852e-05,
+      "loss": 1.2372,
+      "step": 1009
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999862843784341e-05,
+      "loss": 1.1217,
+      "step": 1010
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998609714095873e-05,
+      "loss": 1.2827,
+      "step": 1011
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998590863419482e-05,
+      "loss": 1.16,
+      "step": 1012
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998571885814472e-05,
+      "loss": 1.1289,
+      "step": 1013
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998552781281087e-05,
+      "loss": 1.3314,
+      "step": 1014
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999853354981957e-05,
+      "loss": 1.2046,
+      "step": 1015
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998514191430168e-05,
+      "loss": 1.1749,
+      "step": 1016
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999849470611312e-05,
+      "loss": 1.0816,
+      "step": 1017
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998475093868676e-05,
+      "loss": 1.1387,
+      "step": 1018
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998455354697088e-05,
+      "loss": 1.2461,
+      "step": 1019
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998435488598603e-05,
+      "loss": 1.2732,
+      "step": 1020
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998415495573474e-05,
+      "loss": 1.1552,
+      "step": 1021
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998395375621952e-05,
+      "loss": 1.2689,
+      "step": 1022
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99983751287443e-05,
+      "loss": 1.3062,
+      "step": 1023
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998354754940766e-05,
+      "loss": 1.1473,
+      "step": 1024
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998334254211617e-05,
+      "loss": 1.3128,
+      "step": 1025
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999831362655711e-05,
+      "loss": 1.0528,
+      "step": 1026
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998292871977505e-05,
+      "loss": 1.2571,
+      "step": 1027
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998271990473066e-05,
+      "loss": 1.2264,
+      "step": 1028
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999825098204406e-05,
+      "loss": 1.2256,
+      "step": 1029
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998229846690753e-05,
+      "loss": 1.2178,
+      "step": 1030
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999820858441341e-05,
+      "loss": 1.1772,
+      "step": 1031
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998187195212308e-05,
+      "loss": 1.2403,
+      "step": 1032
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999816567908771e-05,
+      "loss": 1.1925,
+      "step": 1033
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998144036039894e-05,
+      "loss": 1.1168,
+      "step": 1034
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998122266069137e-05,
+      "loss": 1.1947,
+      "step": 1035
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999810036917571e-05,
+      "loss": 1.2906,
+      "step": 1036
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998078345359894e-05,
+      "loss": 1.2803,
+      "step": 1037
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998056194621968e-05,
+      "loss": 1.0647,
+      "step": 1038
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998033916962213e-05,
+      "loss": 1.1949,
+      "step": 1039
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9998011512380913e-05,
+      "loss": 1.1708,
+      "step": 1040
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997988980878354e-05,
+      "loss": 1.3477,
+      "step": 1041
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997966322454815e-05,
+      "loss": 1.1935,
+      "step": 1042
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999794353711059e-05,
+      "loss": 1.3623,
+      "step": 1043
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997920624845968e-05,
+      "loss": 1.1475,
+      "step": 1044
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999789758566124e-05,
+      "loss": 1.2621,
+      "step": 1045
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997874419556692e-05,
+      "loss": 1.159,
+      "step": 1046
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997851126532624e-05,
+      "loss": 1.1653,
+      "step": 1047
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999782770658933e-05,
+      "loss": 1.1406,
+      "step": 1048
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999780415972711e-05,
+      "loss": 1.2809,
+      "step": 1049
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999778048594626e-05,
+      "loss": 1.1535,
+      "step": 1050
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999775668524708e-05,
+      "loss": 1.1535,
+      "step": 1051
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997732757629874e-05,
+      "loss": 1.3082,
+      "step": 1052
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997708703094946e-05,
+      "loss": 1.1808,
+      "step": 1053
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99976845216426e-05,
+      "loss": 1.2011,
+      "step": 1054
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997660213273146e-05,
+      "loss": 1.236,
+      "step": 1055
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999763577798689e-05,
+      "loss": 1.1756,
+      "step": 1056
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997611215784137e-05,
+      "loss": 1.1212,
+      "step": 1057
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999758652666521e-05,
+      "loss": 1.3145,
+      "step": 1058
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997561710630414e-05,
+      "loss": 1.1306,
+      "step": 1059
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997536767680066e-05,
+      "loss": 1.3102,
+      "step": 1060
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997511697814483e-05,
+      "loss": 1.2678,
+      "step": 1061
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997486501033986e-05,
+      "loss": 1.1824,
+      "step": 1062
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999746117733889e-05,
+      "loss": 1.1877,
+      "step": 1063
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999743572672952e-05,
+      "loss": 1.2082,
+      "step": 1064
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99974101492062e-05,
+      "loss": 1.1579,
+      "step": 1065
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997384444769244e-05,
+      "loss": 1.1337,
+      "step": 1066
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997358613418995e-05,
+      "loss": 1.1946,
+      "step": 1067
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997332655155773e-05,
+      "loss": 1.1458,
+      "step": 1068
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997306569979904e-05,
+      "loss": 1.1581,
+      "step": 1069
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997280357891723e-05,
+      "loss": 1.1608,
+      "step": 1070
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997254018891566e-05,
+      "loss": 1.1833,
+      "step": 1071
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999722755297976e-05,
+      "loss": 1.247,
+      "step": 1072
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997200960156642e-05,
+      "loss": 1.2559,
+      "step": 1073
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997174240422558e-05,
+      "loss": 1.058,
+      "step": 1074
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997147393777842e-05,
+      "loss": 1.259,
+      "step": 1075
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999712042022283e-05,
+      "loss": 1.2815,
+      "step": 1076
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997093319757868e-05,
+      "loss": 1.2793,
+      "step": 1077
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997066092383304e-05,
+      "loss": 1.3176,
+      "step": 1078
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9997038738099478e-05,
+      "loss": 1.151,
+      "step": 1079
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999701125690674e-05,
+      "loss": 1.1143,
+      "step": 1080
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996983648805437e-05,
+      "loss": 1.2139,
+      "step": 1081
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996955913795924e-05,
+      "loss": 1.2709,
+      "step": 1082
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996928051878547e-05,
+      "loss": 1.2414,
+      "step": 1083
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996900063053666e-05,
+      "loss": 1.2344,
+      "step": 1084
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996871947321634e-05,
+      "loss": 1.1385,
+      "step": 1085
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99968437046828e-05,
+      "loss": 1.2995,
+      "step": 1086
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996815335137533e-05,
+      "loss": 1.1293,
+      "step": 1087
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999678683868619e-05,
+      "loss": 1.1352,
+      "step": 1088
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996758215329134e-05,
+      "loss": 1.18,
+      "step": 1089
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996729465066727e-05,
+      "loss": 1.1754,
+      "step": 1090
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999670058789933e-05,
+      "loss": 1.1452,
+      "step": 1091
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996671583827317e-05,
+      "loss": 1.267,
+      "step": 1092
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999664245285105e-05,
+      "loss": 1.2141,
+      "step": 1093
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99966131949709e-05,
+      "loss": 1.1572,
+      "step": 1094
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999658381018724e-05,
+      "loss": 1.1221,
+      "step": 1095
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996554298500447e-05,
+      "loss": 1.2119,
+      "step": 1096
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996524659910887e-05,
+      "loss": 1.3477,
+      "step": 1097
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999649489441894e-05,
+      "loss": 1.0857,
+      "step": 1098
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999646500202499e-05,
+      "loss": 1.227,
+      "step": 1099
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996434982729408e-05,
+      "loss": 1.1433,
+      "step": 1100
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996404836532576e-05,
+      "loss": 1.1501,
+      "step": 1101
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999637456343488e-05,
+      "loss": 1.1159,
+      "step": 1102
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996344163436702e-05,
+      "loss": 1.2807,
+      "step": 1103
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996313636538432e-05,
+      "loss": 1.0132,
+      "step": 1104
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999628298274045e-05,
+      "loss": 1.2425,
+      "step": 1105
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996252202043154e-05,
+      "loss": 1.1766,
+      "step": 1106
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999622129444693e-05,
+      "loss": 1.3087,
+      "step": 1107
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996190259952168e-05,
+      "loss": 1.1559,
+      "step": 1108
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996159098559264e-05,
+      "loss": 1.1688,
+      "step": 1109
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996127810268618e-05,
+      "loss": 1.1254,
+      "step": 1110
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999609639508062e-05,
+      "loss": 1.1217,
+      "step": 1111
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996064852995674e-05,
+      "loss": 1.181,
+      "step": 1112
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996033184014177e-05,
+      "loss": 1.2875,
+      "step": 1113
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9996001388136537e-05,
+      "loss": 1.2111,
+      "step": 1114
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999596946536315e-05,
+      "loss": 1.3029,
+      "step": 1115
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995937415694423e-05,
+      "loss": 1.1988,
+      "step": 1116
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995905239130767e-05,
+      "loss": 1.3137,
+      "step": 1117
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999587293567259e-05,
+      "loss": 1.1718,
+      "step": 1118
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995840505320295e-05,
+      "loss": 1.1673,
+      "step": 1119
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995807948074303e-05,
+      "loss": 1.1833,
+      "step": 1120
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995775263935023e-05,
+      "loss": 1.2292,
+      "step": 1121
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995742452902866e-05,
+      "loss": 1.2171,
+      "step": 1122
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995709514978255e-05,
+      "loss": 1.1499,
+      "step": 1123
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995676450161606e-05,
+      "loss": 1.3256,
+      "step": 1124
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995643258453337e-05,
+      "loss": 1.195,
+      "step": 1125
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999560993985387e-05,
+      "loss": 1.2676,
+      "step": 1126
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999557649436363e-05,
+      "loss": 1.0346,
+      "step": 1127
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999554292198304e-05,
+      "loss": 1.1799,
+      "step": 1128
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995509222712524e-05,
+      "loss": 1.0164,
+      "step": 1129
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995475396552518e-05,
+      "loss": 1.125,
+      "step": 1130
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995441443503443e-05,
+      "loss": 1.1528,
+      "step": 1131
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995407363565732e-05,
+      "loss": 1.2417,
+      "step": 1132
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999537315673982e-05,
+      "loss": 1.3036,
+      "step": 1133
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995338823026135e-05,
+      "loss": 1.0966,
+      "step": 1134
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999530436242512e-05,
+      "loss": 1.2114,
+      "step": 1135
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995269774937213e-05,
+      "loss": 1.0664,
+      "step": 1136
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995235060562846e-05,
+      "loss": 1.2378,
+      "step": 1137
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995200219302462e-05,
+      "loss": 1.0846,
+      "step": 1138
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999516525115651e-05,
+      "loss": 1.2319,
+      "step": 1139
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995130156125424e-05,
+      "loss": 1.1113,
+      "step": 1140
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995094934209657e-05,
+      "loss": 1.1714,
+      "step": 1141
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9995059585409653e-05,
+      "loss": 1.3146,
+      "step": 1142
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999502410972586e-05,
+      "loss": 1.311,
+      "step": 1143
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994988507158732e-05,
+      "loss": 1.0528,
+      "step": 1144
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994952777708718e-05,
+      "loss": 1.2106,
+      "step": 1145
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999491692137627e-05,
+      "loss": 1.1788,
+      "step": 1146
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994880938161847e-05,
+      "loss": 1.243,
+      "step": 1147
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99948448280659e-05,
+      "loss": 1.1449,
+      "step": 1148
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99948085910889e-05,
+      "loss": 1.2144,
+      "step": 1149
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994772227231293e-05,
+      "loss": 1.0759,
+      "step": 1150
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994735736493548e-05,
+      "loss": 1.1816,
+      "step": 1151
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999469911887612e-05,
+      "loss": 1.1826,
+      "step": 1152
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999466237437949e-05,
+      "loss": 1.2378,
+      "step": 1153
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999462550300411e-05,
+      "loss": 1.1967,
+      "step": 1154
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999458850475045e-05,
+      "loss": 1.2111,
+      "step": 1155
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994551379618987e-05,
+      "loss": 1.2488,
+      "step": 1156
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994514127610184e-05,
+      "loss": 1.1516,
+      "step": 1157
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994476748724522e-05,
+      "loss": 1.1194,
+      "step": 1158
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994439242962466e-05,
+      "loss": 1.2208,
+      "step": 1159
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99944016103245e-05,
+      "loss": 1.111,
+      "step": 1160
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994363850811102e-05,
+      "loss": 1.1395,
+      "step": 1161
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994325964422743e-05,
+      "loss": 1.0007,
+      "step": 1162
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999428795115991e-05,
+      "loss": 1.2404,
+      "step": 1163
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994249811023086e-05,
+      "loss": 1.2006,
+      "step": 1164
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994211544012754e-05,
+      "loss": 1.053,
+      "step": 1165
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994173150129402e-05,
+      "loss": 1.2119,
+      "step": 1166
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994134629373512e-05,
+      "loss": 1.0814,
+      "step": 1167
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994095981745577e-05,
+      "loss": 1.1045,
+      "step": 1168
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994057207246087e-05,
+      "loss": 1.2752,
+      "step": 1169
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9994018305875535e-05,
+      "loss": 1.1788,
+      "step": 1170
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993979277634413e-05,
+      "loss": 1.1446,
+      "step": 1171
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993940122523215e-05,
+      "loss": 1.2249,
+      "step": 1172
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993900840542444e-05,
+      "loss": 1.2273,
+      "step": 1173
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993861431692593e-05,
+      "loss": 1.2924,
+      "step": 1174
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993821895974164e-05,
+      "loss": 1.3323,
+      "step": 1175
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993782233387663e-05,
+      "loss": 1.04,
+      "step": 1176
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993742443933586e-05,
+      "loss": 1.1913,
+      "step": 1177
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993702527612446e-05,
+      "loss": 1.1786,
+      "step": 1178
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999366248442474e-05,
+      "loss": 1.2673,
+      "step": 1179
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993622314370987e-05,
+      "loss": 1.1743,
+      "step": 1180
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993582017451686e-05,
+      "loss": 1.061,
+      "step": 1181
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993541593667357e-05,
+      "loss": 1.0816,
+      "step": 1182
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993501043018513e-05,
+      "loss": 1.2808,
+      "step": 1183
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993460365505662e-05,
+      "loss": 1.2125,
+      "step": 1184
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993419561129332e-05,
+      "loss": 1.1254,
+      "step": 1185
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993378629890028e-05,
+      "loss": 1.2558,
+      "step": 1186
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993337571788275e-05,
+      "loss": 1.2076,
+      "step": 1187
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993296386824595e-05,
+      "loss": 1.1917,
+      "step": 1188
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993255074999513e-05,
+      "loss": 1.1736,
+      "step": 1189
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993213636313548e-05,
+      "loss": 1.108,
+      "step": 1190
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999317207076723e-05,
+      "loss": 1.2926,
+      "step": 1191
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993130378361085e-05,
+      "loss": 1.2285,
+      "step": 1192
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993088559095644e-05,
+      "loss": 1.2121,
+      "step": 1193
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993046612971437e-05,
+      "loss": 1.1244,
+      "step": 1194
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9993004539988997e-05,
+      "loss": 1.2734,
+      "step": 1195
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992962340148858e-05,
+      "loss": 1.2926,
+      "step": 1196
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992920013451548e-05,
+      "loss": 1.2598,
+      "step": 1197
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992877559897618e-05,
+      "loss": 1.2515,
+      "step": 1198
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992834979487598e-05,
+      "loss": 1.4369,
+      "step": 1199
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999279227222203e-05,
+      "loss": 1.2991,
+      "step": 1200
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999274943810146e-05,
+      "loss": 1.128,
+      "step": 1201
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992706477126424e-05,
+      "loss": 1.0873,
+      "step": 1202
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992663389297476e-05,
+      "loss": 1.1819,
+      "step": 1203
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992620174615158e-05,
+      "loss": 1.1919,
+      "step": 1204
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999257683308002e-05,
+      "loss": 1.1417,
+      "step": 1205
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999253336469261e-05,
+      "loss": 1.1278,
+      "step": 1206
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992489769453486e-05,
+      "loss": 1.1557,
+      "step": 1207
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999244604736319e-05,
+      "loss": 1.2115,
+      "step": 1208
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999240219842229e-05,
+      "loss": 1.1544,
+      "step": 1209
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992358222631335e-05,
+      "loss": 1.3334,
+      "step": 1210
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992314119990884e-05,
+      "loss": 1.2194,
+      "step": 1211
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992269890501503e-05,
+      "loss": 1.1395,
+      "step": 1212
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992225534163744e-05,
+      "loss": 1.1315,
+      "step": 1213
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992181050978173e-05,
+      "loss": 1.1386,
+      "step": 1214
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999213644094536e-05,
+      "loss": 1.1187,
+      "step": 1215
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992091704065864e-05,
+      "loss": 1.1703,
+      "step": 1216
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999204684034026e-05,
+      "loss": 1.1628,
+      "step": 1217
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9992001849769108e-05,
+      "loss": 1.0561,
+      "step": 1218
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999195673235299e-05,
+      "loss": 1.1647,
+      "step": 1219
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991911488092474e-05,
+      "loss": 1.2026,
+      "step": 1220
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991866116988133e-05,
+      "loss": 1.2018,
+      "step": 1221
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991820619040543e-05,
+      "loss": 1.2739,
+      "step": 1222
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991774994250285e-05,
+      "loss": 1.1892,
+      "step": 1223
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991729242617937e-05,
+      "loss": 1.1539,
+      "step": 1224
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991683364144076e-05,
+      "loss": 1.1729,
+      "step": 1225
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991637358829287e-05,
+      "loss": 1.1757,
+      "step": 1226
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991591226674157e-05,
+      "loss": 1.24,
+      "step": 1227
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991544967679265e-05,
+      "loss": 1.132,
+      "step": 1228
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991498581845205e-05,
+      "loss": 1.1406,
+      "step": 1229
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991452069172558e-05,
+      "loss": 1.1861,
+      "step": 1230
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991405429661925e-05,
+      "loss": 1.1997,
+      "step": 1231
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999135866331389e-05,
+      "loss": 1.0702,
+      "step": 1232
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999131177012905e-05,
+      "loss": 1.1336,
+      "step": 1233
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991264750107998e-05,
+      "loss": 1.0316,
+      "step": 1234
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991217603251335e-05,
+      "loss": 1.0699,
+      "step": 1235
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991170329559656e-05,
+      "loss": 1.0757,
+      "step": 1236
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991122929033562e-05,
+      "loss": 1.2963,
+      "step": 1237
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991075401673653e-05,
+      "loss": 1.0534,
+      "step": 1238
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9991027747480538e-05,
+      "loss": 1.2426,
+      "step": 1239
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999097996645481e-05,
+      "loss": 1.1646,
+      "step": 1240
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999093205859709e-05,
+      "loss": 1.2555,
+      "step": 1241
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990884023907977e-05,
+      "loss": 1.2903,
+      "step": 1242
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990835862388083e-05,
+      "loss": 1.2479,
+      "step": 1243
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990787574038023e-05,
+      "loss": 1.1449,
+      "step": 1244
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999073915885841e-05,
+      "loss": 1.0646,
+      "step": 1245
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990690616849847e-05,
+      "loss": 1.1662,
+      "step": 1246
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990641948012963e-05,
+      "loss": 1.2031,
+      "step": 1247
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999059315234837e-05,
+      "loss": 1.2199,
+      "step": 1248
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999054422985669e-05,
+      "loss": 1.2539,
+      "step": 1249
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990495180538545e-05,
+      "loss": 1.117,
+      "step": 1250
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990446004394553e-05,
+      "loss": 1.1111,
+      "step": 1251
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990396701425344e-05,
+      "loss": 1.2025,
+      "step": 1252
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999034727163154e-05,
+      "loss": 1.2467,
+      "step": 1253
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990297715013764e-05,
+      "loss": 1.2523,
+      "step": 1254
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990248031572657e-05,
+      "loss": 1.196,
+      "step": 1255
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999019822130884e-05,
+      "loss": 1.1868,
+      "step": 1256
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.999014828422295e-05,
+      "loss": 1.1805,
+      "step": 1257
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990098220315616e-05,
+      "loss": 1.1022,
+      "step": 1258
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9990048029587482e-05,
+      "loss": 1.2622,
+      "step": 1259
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989997712039177e-05,
+      "loss": 1.2226,
+      "step": 1260
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989947267671344e-05,
+      "loss": 1.1849,
+      "step": 1261
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989896696484624e-05,
+      "loss": 1.213,
+      "step": 1262
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989845998479656e-05,
+      "loss": 1.1453,
+      "step": 1263
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989795173657085e-05,
+      "loss": 1.1658,
+      "step": 1264
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989744222017554e-05,
+      "loss": 1.1442,
+      "step": 1265
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989693143561714e-05,
+      "loss": 1.1462,
+      "step": 1266
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989641938290208e-05,
+      "loss": 1.1298,
+      "step": 1267
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989590606203695e-05,
+      "loss": 1.2678,
+      "step": 1268
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989539147302816e-05,
+      "loss": 1.1678,
+      "step": 1269
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998948756158823e-05,
+      "loss": 1.1467,
+      "step": 1270
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989435849060593e-05,
+      "loss": 1.2688,
+      "step": 1271
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989384009720562e-05,
+      "loss": 1.002,
+      "step": 1272
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998933204356879e-05,
+      "loss": 1.0777,
+      "step": 1273
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989279950605938e-05,
+      "loss": 1.2487,
+      "step": 1274
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989227730832667e-05,
+      "loss": 1.2776,
+      "step": 1275
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989175384249642e-05,
+      "loss": 1.202,
+      "step": 1276
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989122910857532e-05,
+      "loss": 1.1057,
+      "step": 1277
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9989070310656993e-05,
+      "loss": 1.1303,
+      "step": 1278
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.99890175836487e-05,
+      "loss": 1.3503,
+      "step": 1279
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988964729833324e-05,
+      "loss": 1.266,
+      "step": 1280
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998891174921153e-05,
+      "loss": 1.1284,
+      "step": 1281
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998885864178399e-05,
+      "loss": 1.2359,
+      "step": 1282
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988805407551384e-05,
+      "loss": 1.0437,
+      "step": 1283
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998875204651438e-05,
+      "loss": 1.0977,
+      "step": 1284
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988698558673664e-05,
+      "loss": 1.3001,
+      "step": 1285
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988644944029913e-05,
+      "loss": 1.1794,
+      "step": 1286
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988591202583802e-05,
+      "loss": 1.1963,
+      "step": 1287
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998853733433602e-05,
+      "loss": 1.147,
+      "step": 1288
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988483339287247e-05,
+      "loss": 1.2528,
+      "step": 1289
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988429217438168e-05,
+      "loss": 1.2167,
+      "step": 1290
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988374968789473e-05,
+      "loss": 1.1661,
+      "step": 1291
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.998832059334185e-05,
+      "loss": 1.168,
+      "step": 1292
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 1.9988266091095988e-05,
+      "loss": 1.1889,
+      "step": 1293
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988211462052576e-05,
+      "loss": 1.0659,
+      "step": 1294
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988156706212313e-05,
+      "loss": 1.2841,
+      "step": 1295
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998810182357589e-05,
+      "loss": 1.1428,
+      "step": 1296
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9988046814144008e-05,
+      "loss": 1.2688,
+      "step": 1297
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987991677917362e-05,
+      "loss": 1.2573,
+      "step": 1298
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987936414896654e-05,
+      "loss": 1.2428,
+      "step": 1299
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987881025082584e-05,
+      "loss": 1.1744,
+      "step": 1300
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987825508475854e-05,
+      "loss": 1.0511,
+      "step": 1301
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998776986507717e-05,
+      "loss": 1.185,
+      "step": 1302
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998771409488724e-05,
+      "loss": 1.2843,
+      "step": 1303
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987658197906768e-05,
+      "loss": 1.3053,
+      "step": 1304
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987602174136473e-05,
+      "loss": 0.9467,
+      "step": 1305
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987546023577054e-05,
+      "loss": 1.1008,
+      "step": 1306
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987489746229228e-05,
+      "loss": 1.1426,
+      "step": 1307
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987433342093715e-05,
+      "loss": 1.2206,
+      "step": 1308
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987376811171224e-05,
+      "loss": 1.1479,
+      "step": 1309
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987320153462474e-05,
+      "loss": 1.2025,
+      "step": 1310
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987263368968192e-05,
+      "loss": 1.1081,
+      "step": 1311
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987206457689087e-05,
+      "loss": 1.2337,
+      "step": 1312
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987149419625886e-05,
+      "loss": 1.0648,
+      "step": 1313
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987092254779317e-05,
+      "loss": 1.2082,
+      "step": 1314
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9987034963150102e-05,
+      "loss": 1.1537,
+      "step": 1315
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986977544738968e-05,
+      "loss": 1.1165,
+      "step": 1316
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986919999546646e-05,
+      "loss": 1.2243,
+      "step": 1317
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986862327573864e-05,
+      "loss": 1.2077,
+      "step": 1318
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998680452882136e-05,
+      "loss": 1.244,
+      "step": 1319
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986746603289857e-05,
+      "loss": 1.1018,
+      "step": 1320
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99866885509801e-05,
+      "loss": 1.2168,
+      "step": 1321
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986630371892822e-05,
+      "loss": 1.1413,
+      "step": 1322
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998657206602876e-05,
+      "loss": 1.0806,
+      "step": 1323
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986513633388657e-05,
+      "loss": 1.1709,
+      "step": 1324
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986455073973254e-05,
+      "loss": 1.1229,
+      "step": 1325
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99863963877833e-05,
+      "loss": 1.3134,
+      "step": 1326
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986337574819526e-05,
+      "loss": 1.1394,
+      "step": 1327
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986278635082693e-05,
+      "loss": 1.294,
+      "step": 1328
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986219568573543e-05,
+      "loss": 1.1786,
+      "step": 1329
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986160375292824e-05,
+      "loss": 1.2638,
+      "step": 1330
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998610105524129e-05,
+      "loss": 1.2315,
+      "step": 1331
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9986041608419694e-05,
+      "loss": 1.2469,
+      "step": 1332
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998598203482879e-05,
+      "loss": 1.0443,
+      "step": 1333
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985922334469334e-05,
+      "loss": 1.1953,
+      "step": 1334
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985862507342087e-05,
+      "loss": 1.328,
+      "step": 1335
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99858025534478e-05,
+      "loss": 1.1989,
+      "step": 1336
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985742472787248e-05,
+      "loss": 1.3497,
+      "step": 1337
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998568226536118e-05,
+      "loss": 1.1729,
+      "step": 1338
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985621931170368e-05,
+      "loss": 1.301,
+      "step": 1339
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985561470215576e-05,
+      "loss": 1.1031,
+      "step": 1340
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998550088249757e-05,
+      "loss": 1.1985,
+      "step": 1341
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998544016801712e-05,
+      "loss": 1.1216,
+      "step": 1342
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985379326775e-05,
+      "loss": 1.0813,
+      "step": 1343
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985318358771978e-05,
+      "loss": 1.1556,
+      "step": 1344
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985257264008832e-05,
+      "loss": 1.1867,
+      "step": 1345
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985196042486334e-05,
+      "loss": 1.1391,
+      "step": 1346
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998513469420526e-05,
+      "loss": 1.179,
+      "step": 1347
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985073219166393e-05,
+      "loss": 1.1339,
+      "step": 1348
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9985011617370514e-05,
+      "loss": 1.2721,
+      "step": 1349
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99849498888184e-05,
+      "loss": 1.184,
+      "step": 1350
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984888033510835e-05,
+      "loss": 1.2213,
+      "step": 1351
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984826051448612e-05,
+      "loss": 1.0352,
+      "step": 1352
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998476394263251e-05,
+      "loss": 1.2318,
+      "step": 1353
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998470170706332e-05,
+      "loss": 1.1563,
+      "step": 1354
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998463934474183e-05,
+      "loss": 1.1756,
+      "step": 1355
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984576855668837e-05,
+      "loss": 1.2541,
+      "step": 1356
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984514239845127e-05,
+      "loss": 1.2041,
+      "step": 1357
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99844514972715e-05,
+      "loss": 1.0895,
+      "step": 1358
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998438862794875e-05,
+      "loss": 1.1589,
+      "step": 1359
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984325631877677e-05,
+      "loss": 1.1537,
+      "step": 1360
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984262509059084e-05,
+      "loss": 1.3605,
+      "step": 1361
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984199259493763e-05,
+      "loss": 1.1069,
+      "step": 1362
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984135883182524e-05,
+      "loss": 1.2561,
+      "step": 1363
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998407238012617e-05,
+      "loss": 1.224,
+      "step": 1364
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9984008750325507e-05,
+      "loss": 1.2271,
+      "step": 1365
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983944993781345e-05,
+      "loss": 1.1948,
+      "step": 1366
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983881110494487e-05,
+      "loss": 1.1395,
+      "step": 1367
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983817100465748e-05,
+      "loss": 1.3096,
+      "step": 1368
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983752963695942e-05,
+      "loss": 1.1981,
+      "step": 1369
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983688700185885e-05,
+      "loss": 1.174,
+      "step": 1370
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983624309936384e-05,
+      "loss": 1.2844,
+      "step": 1371
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983559792948265e-05,
+      "loss": 1.3186,
+      "step": 1372
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983495149222346e-05,
+      "loss": 1.1845,
+      "step": 1373
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998343037875944e-05,
+      "loss": 1.0467,
+      "step": 1374
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983365481560382e-05,
+      "loss": 1.1575,
+      "step": 1375
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983300457625984e-05,
+      "loss": 1.1215,
+      "step": 1376
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983235306957077e-05,
+      "loss": 1.0947,
+      "step": 1377
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983170029554487e-05,
+      "loss": 1.0064,
+      "step": 1378
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983104625419044e-05,
+      "loss": 1.2465,
+      "step": 1379
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9983039094551576e-05,
+      "loss": 1.2066,
+      "step": 1380
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982973436952918e-05,
+      "loss": 1.1992,
+      "step": 1381
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99829076526239e-05,
+      "loss": 1.2655,
+      "step": 1382
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982841741565358e-05,
+      "loss": 1.1667,
+      "step": 1383
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982775703778134e-05,
+      "loss": 1.1594,
+      "step": 1384
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982709539263057e-05,
+      "loss": 1.2877,
+      "step": 1385
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982643248020973e-05,
+      "loss": 1.1018,
+      "step": 1386
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982576830052723e-05,
+      "loss": 1.1189,
+      "step": 1387
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998251028535915e-05,
+      "loss": 1.1005,
+      "step": 1388
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982443613941096e-05,
+      "loss": 1.133,
+      "step": 1389
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998237681579941e-05,
+      "loss": 1.2471,
+      "step": 1390
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982309890934944e-05,
+      "loss": 1.2071,
+      "step": 1391
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982242839348537e-05,
+      "loss": 1.1404,
+      "step": 1392
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998217566104105e-05,
+      "loss": 1.1351,
+      "step": 1393
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998210835601333e-05,
+      "loss": 1.1411,
+      "step": 1394
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9982040924266234e-05,
+      "loss": 1.0325,
+      "step": 1395
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981973365800617e-05,
+      "loss": 1.2748,
+      "step": 1396
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998190568061734e-05,
+      "loss": 1.1937,
+      "step": 1397
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998183786871726e-05,
+      "loss": 1.2336,
+      "step": 1398
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981769930101232e-05,
+      "loss": 1.093,
+      "step": 1399
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981701864770127e-05,
+      "loss": 1.1011,
+      "step": 1400
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981633672724804e-05,
+      "loss": 1.2633,
+      "step": 1401
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981565353966134e-05,
+      "loss": 1.3208,
+      "step": 1402
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981496908494977e-05,
+      "loss": 1.1902,
+      "step": 1403
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981428336312207e-05,
+      "loss": 1.1907,
+      "step": 1404
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981359637418692e-05,
+      "loss": 1.1865,
+      "step": 1405
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981290811815307e-05,
+      "loss": 1.1111,
+      "step": 1406
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981221859502924e-05,
+      "loss": 1.0542,
+      "step": 1407
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998115278048242e-05,
+      "loss": 1.2875,
+      "step": 1408
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981083574754665e-05,
+      "loss": 1.2631,
+      "step": 1409
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9981014242320546e-05,
+      "loss": 1.1491,
+      "step": 1410
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998094478318094e-05,
+      "loss": 1.1475,
+      "step": 1411
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998087519733673e-05,
+      "loss": 1.0373,
+      "step": 1412
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980805484788798e-05,
+      "loss": 1.1829,
+      "step": 1413
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998073564553802e-05,
+      "loss": 1.0962,
+      "step": 1414
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980665679585302e-05,
+      "loss": 1.2927,
+      "step": 1415
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998059558693152e-05,
+      "loss": 1.1671,
+      "step": 1416
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980525367577565e-05,
+      "loss": 1.2635,
+      "step": 1417
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998045502152433e-05,
+      "loss": 1.1726,
+      "step": 1418
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980384548772703e-05,
+      "loss": 1.0883,
+      "step": 1419
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980313949323585e-05,
+      "loss": 1.1567,
+      "step": 1420
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980243223177874e-05,
+      "loss": 1.3529,
+      "step": 1421
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980172370336458e-05,
+      "loss": 1.1752,
+      "step": 1422
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9980101390800244e-05,
+      "loss": 1.1522,
+      "step": 1423
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.998003028457013e-05,
+      "loss": 1.228,
+      "step": 1424
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997995905164702e-05,
+      "loss": 1.2575,
+      "step": 1425
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997988769203182e-05,
+      "loss": 1.1489,
+      "step": 1426
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979816205725436e-05,
+      "loss": 1.1336,
+      "step": 1427
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997974459272877e-05,
+      "loss": 1.1772,
+      "step": 1428
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997967285304274e-05,
+      "loss": 1.2567,
+      "step": 1429
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997960098666824e-05,
+      "loss": 1.1547,
+      "step": 1430
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979528993606206e-05,
+      "loss": 1.2204,
+      "step": 1431
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979456873857532e-05,
+      "loss": 1.1592,
+      "step": 1432
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979384627423145e-05,
+      "loss": 1.2392,
+      "step": 1433
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979312254303956e-05,
+      "loss": 1.048,
+      "step": 1434
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979239754500886e-05,
+      "loss": 1.1598,
+      "step": 1435
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979167128014857e-05,
+      "loss": 1.1835,
+      "step": 1436
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997909437484679e-05,
+      "loss": 1.1501,
+      "step": 1437
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9979021494997605e-05,
+      "loss": 1.1487,
+      "step": 1438
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997894848846823e-05,
+      "loss": 1.2584,
+      "step": 1439
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99788753552596e-05,
+      "loss": 1.1826,
+      "step": 1440
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978802095372628e-05,
+      "loss": 1.2813,
+      "step": 1441
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978728708808252e-05,
+      "loss": 1.1277,
+      "step": 1442
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978655195567405e-05,
+      "loss": 1.1716,
+      "step": 1443
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978581555651015e-05,
+      "loss": 1.1584,
+      "step": 1444
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997850778906002e-05,
+      "loss": 1.211,
+      "step": 1445
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997843389579536e-05,
+      "loss": 1.1778,
+      "step": 1446
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978359875857966e-05,
+      "loss": 1.1757,
+      "step": 1447
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978285729248785e-05,
+      "loss": 1.1748,
+      "step": 1448
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978211455968753e-05,
+      "loss": 1.0987,
+      "step": 1449
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978137056018814e-05,
+      "loss": 1.1879,
+      "step": 1450
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9978062529399912e-05,
+      "loss": 1.0707,
+      "step": 1451
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977987876112993e-05,
+      "loss": 1.1401,
+      "step": 1452
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977913096159008e-05,
+      "loss": 1.3079,
+      "step": 1453
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977838189538903e-05,
+      "loss": 1.3126,
+      "step": 1454
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997776315625363e-05,
+      "loss": 1.1541,
+      "step": 1455
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977687996304142e-05,
+      "loss": 1.197,
+      "step": 1456
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997761270969139e-05,
+      "loss": 1.0669,
+      "step": 1457
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977537296416335e-05,
+      "loss": 1.0454,
+      "step": 1458
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997746175647993e-05,
+      "loss": 1.1896,
+      "step": 1459
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977386089883136e-05,
+      "loss": 1.2174,
+      "step": 1460
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997731029662691e-05,
+      "loss": 1.1455,
+      "step": 1461
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977234376712224e-05,
+      "loss": 1.1972,
+      "step": 1462
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977158330140027e-05,
+      "loss": 0.99,
+      "step": 1463
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977082156911298e-05,
+      "loss": 1.2564,
+      "step": 1464
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9977005857026997e-05,
+      "loss": 1.0694,
+      "step": 1465
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997692943048809e-05,
+      "loss": 1.2031,
+      "step": 1466
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976852877295558e-05,
+      "loss": 1.0288,
+      "step": 1467
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976776197450362e-05,
+      "loss": 1.1409,
+      "step": 1468
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997669939095348e-05,
+      "loss": 1.2519,
+      "step": 1469
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997662245780589e-05,
+      "loss": 1.1341,
+      "step": 1470
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997654539800856e-05,
+      "loss": 1.1007,
+      "step": 1471
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976468211562477e-05,
+      "loss": 1.2415,
+      "step": 1472
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976390898468613e-05,
+      "loss": 1.16,
+      "step": 1473
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976313458727954e-05,
+      "loss": 1.1169,
+      "step": 1474
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976235892341488e-05,
+      "loss": 1.2281,
+      "step": 1475
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997615819931019e-05,
+      "loss": 1.2179,
+      "step": 1476
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9976080379635053e-05,
+      "loss": 1.039,
+      "step": 1477
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997600243331706e-05,
+      "loss": 1.1969,
+      "step": 1478
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975924360357206e-05,
+      "loss": 1.122,
+      "step": 1479
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975846160756477e-05,
+      "loss": 1.1866,
+      "step": 1480
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997576783451587e-05,
+      "loss": 1.2392,
+      "step": 1481
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975689381636375e-05,
+      "loss": 1.3008,
+      "step": 1482
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997561080211899e-05,
+      "loss": 0.9898,
+      "step": 1483
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975532095964713e-05,
+      "loss": 1.185,
+      "step": 1484
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975453263174543e-05,
+      "loss": 1.1663,
+      "step": 1485
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975374303749482e-05,
+      "loss": 1.0843,
+      "step": 1486
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997529521769053e-05,
+      "loss": 1.1218,
+      "step": 1487
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975216004998696e-05,
+      "loss": 1.166,
+      "step": 1488
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975136665674973e-05,
+      "loss": 1.1552,
+      "step": 1489
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9975057199720384e-05,
+      "loss": 1.0517,
+      "step": 1490
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997497760713593e-05,
+      "loss": 1.0805,
+      "step": 1491
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997489788792262e-05,
+      "loss": 1.027,
+      "step": 1492
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974818042081464e-05,
+      "loss": 1.0696,
+      "step": 1493
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974738069613485e-05,
+      "loss": 1.2385,
+      "step": 1494
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997465797051969e-05,
+      "loss": 1.1097,
+      "step": 1495
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99745777448011e-05,
+      "loss": 1.1572,
+      "step": 1496
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997449739245873e-05,
+      "loss": 1.1936,
+      "step": 1497
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974416913493605e-05,
+      "loss": 1.1821,
+      "step": 1498
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997433630790674e-05,
+      "loss": 1.1516,
+      "step": 1499
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974255575699163e-05,
+      "loss": 1.1741,
+      "step": 1500
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974174716871903e-05,
+      "loss": 1.1677,
+      "step": 1501
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9974093731425977e-05,
+      "loss": 1.0739,
+      "step": 1502
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997401261936242e-05,
+      "loss": 1.0774,
+      "step": 1503
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997393138068225e-05,
+      "loss": 1.2262,
+      "step": 1504
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973850015386518e-05,
+      "loss": 1.1843,
+      "step": 1505
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973768523476242e-05,
+      "loss": 1.2179,
+      "step": 1506
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973686904952464e-05,
+      "loss": 1.2138,
+      "step": 1507
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973605159816213e-05,
+      "loss": 1.1521,
+      "step": 1508
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997352328806853e-05,
+      "loss": 1.442,
+      "step": 1509
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973441289710455e-05,
+      "loss": 1.2133,
+      "step": 1510
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973359164743035e-05,
+      "loss": 1.144,
+      "step": 1511
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99732769131673e-05,
+      "loss": 1.1361,
+      "step": 1512
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973194534984303e-05,
+      "loss": 1.3351,
+      "step": 1513
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9973112030195088e-05,
+      "loss": 1.2056,
+      "step": 1514
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99730293988007e-05,
+      "loss": 1.1701,
+      "step": 1515
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972946640802192e-05,
+      "loss": 1.1208,
+      "step": 1516
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997286375620061e-05,
+      "loss": 1.0321,
+      "step": 1517
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997278074499701e-05,
+      "loss": 1.1977,
+      "step": 1518
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972697607192442e-05,
+      "loss": 1.2313,
+      "step": 1519
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972614342787967e-05,
+      "loss": 1.1233,
+      "step": 1520
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972530951784635e-05,
+      "loss": 1.083,
+      "step": 1521
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997244743418351e-05,
+      "loss": 1.1706,
+      "step": 1522
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972363789985652e-05,
+      "loss": 1.0676,
+      "step": 1523
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972280019192117e-05,
+      "loss": 1.1633,
+      "step": 1524
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997219612180398e-05,
+      "loss": 1.1016,
+      "step": 1525
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9972112097822294e-05,
+      "loss": 1.2909,
+      "step": 1526
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997202794724813e-05,
+      "loss": 1.073,
+      "step": 1527
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997194367008256e-05,
+      "loss": 1.0267,
+      "step": 1528
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997185926632665e-05,
+      "loss": 1.0117,
+      "step": 1529
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997177473598147e-05,
+      "loss": 1.2778,
+      "step": 1530
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99716900790481e-05,
+      "loss": 1.0818,
+      "step": 1531
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971605295527604e-05,
+      "loss": 1.0376,
+      "step": 1532
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971520385421067e-05,
+      "loss": 1.2159,
+      "step": 1533
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971435348729562e-05,
+      "loss": 1.0962,
+      "step": 1534
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971350185454174e-05,
+      "loss": 1.2478,
+      "step": 1535
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971264895595977e-05,
+      "loss": 1.1886,
+      "step": 1536
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.997117947915606e-05,
+      "loss": 1.0742,
+      "step": 1537
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971093936135504e-05,
+      "loss": 1.0816,
+      "step": 1538
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9971008266535398e-05,
+      "loss": 1.1299,
+      "step": 1539
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970922470356825e-05,
+      "loss": 1.2308,
+      "step": 1540
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970836547600875e-05,
+      "loss": 1.2206,
+      "step": 1541
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970750498268644e-05,
+      "loss": 1.2857,
+      "step": 1542
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970664322361217e-05,
+      "loss": 1.2363,
+      "step": 1543
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970578019879695e-05,
+      "loss": 1.1796,
+      "step": 1544
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970491590825166e-05,
+      "loss": 1.1447,
+      "step": 1545
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970405035198733e-05,
+      "loss": 1.2141,
+      "step": 1546
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970318353001496e-05,
+      "loss": 1.1634,
+      "step": 1547
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970231544234548e-05,
+      "loss": 1.1447,
+      "step": 1548
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970144608899e-05,
+      "loss": 1.0421,
+      "step": 1549
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9970057546995947e-05,
+      "loss": 1.0277,
+      "step": 1550
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99699703585265e-05,
+      "loss": 1.205,
+      "step": 1551
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969883043491764e-05,
+      "loss": 1.3653,
+      "step": 1552
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996979560189285e-05,
+      "loss": 1.1992,
+      "step": 1553
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969708033730863e-05,
+      "loss": 1.1166,
+      "step": 1554
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969620339006916e-05,
+      "loss": 1.3044,
+      "step": 1555
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969532517722126e-05,
+      "loss": 1.1671,
+      "step": 1556
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969444569877603e-05,
+      "loss": 1.2389,
+      "step": 1557
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996935649547447e-05,
+      "loss": 1.2008,
+      "step": 1558
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969268294513838e-05,
+      "loss": 1.243,
+      "step": 1559
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996917996699683e-05,
+      "loss": 1.2249,
+      "step": 1560
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9969091512924566e-05,
+      "loss": 1.1717,
+      "step": 1561
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996900293229817e-05,
+      "loss": 1.1773,
+      "step": 1562
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968914225118768e-05,
+      "loss": 1.148,
+      "step": 1563
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968825391387483e-05,
+      "loss": 1.254,
+      "step": 1564
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968736431105448e-05,
+      "loss": 1.1898,
+      "step": 1565
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968647344273783e-05,
+      "loss": 1.1903,
+      "step": 1566
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968558130893627e-05,
+      "loss": 1.089,
+      "step": 1567
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968468790966108e-05,
+      "loss": 1.1341,
+      "step": 1568
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968379324492365e-05,
+      "loss": 1.0807,
+      "step": 1569
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996828973147353e-05,
+      "loss": 1.3401,
+      "step": 1570
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.996820001191074e-05,
+      "loss": 1.2653,
+      "step": 1571
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968110165805134e-05,
+      "loss": 1.1588,
+      "step": 1572
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9968020193157855e-05,
+      "loss": 1.0625,
+      "step": 1573
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967930093970046e-05,
+      "loss": 1.0894,
+      "step": 1574
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967839868242844e-05,
+      "loss": 1.2232,
+      "step": 1575
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.99677495159774e-05,
+      "loss": 1.1176,
+      "step": 1576
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967659037174863e-05,
+      "loss": 1.1088,
+      "step": 1577
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967568431836378e-05,
+      "loss": 1.1728,
+      "step": 1578
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967477699963096e-05,
+      "loss": 1.2066,
+      "step": 1579
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967386841556167e-05,
+      "loss": 1.1024,
+      "step": 1580
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.9967295856616746e-05,
+      "loss": 1.1984,
+      "step": 1581
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996720474514599e-05,
+      "loss": 1.2993,
+      "step": 1582
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996711350714505e-05,
+      "loss": 1.1864,
+      "step": 1583
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996702214261509e-05,
+      "loss": 1.149,
+      "step": 1584
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966930651557268e-05,
+      "loss": 1.27,
+      "step": 1585
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966839033972747e-05,
+      "loss": 1.1436,
+      "step": 1586
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966747289862686e-05,
+      "loss": 1.1395,
+      "step": 1587
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996665541922825e-05,
+      "loss": 1.138,
+      "step": 1588
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966563422070614e-05,
+      "loss": 1.2422,
+      "step": 1589
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966471298390934e-05,
+      "loss": 1.2189,
+      "step": 1590
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966379048190383e-05,
+      "loss": 1.1692,
+      "step": 1591
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966286671470137e-05,
+      "loss": 1.1154,
+      "step": 1592
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996619416823136e-05,
+      "loss": 1.1885,
+      "step": 1593
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966101538475238e-05,
+      "loss": 1.2246,
+      "step": 1594
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9966008782202936e-05,
+      "loss": 1.1141,
+      "step": 1595
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996591589941564e-05,
+      "loss": 1.4125,
+      "step": 1596
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965822890114525e-05,
+      "loss": 1.2725,
+      "step": 1597
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965729754300766e-05,
+      "loss": 1.1864,
+      "step": 1598
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965636491975556e-05,
+      "loss": 1.149,
+      "step": 1599
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965543103140073e-05,
+      "loss": 1.1376,
+      "step": 1600
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965449587795503e-05,
+      "loss": 1.144,
+      "step": 1601
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965355945943035e-05,
+      "loss": 1.1575,
+      "step": 1602
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9965262177583855e-05,
+      "loss": 1.2493,
+      "step": 1603
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996516828271916e-05,
+      "loss": 1.1423,
+      "step": 1604
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996507426135013e-05,
+      "loss": 1.1474,
+      "step": 1605
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964980113477968e-05,
+      "loss": 1.2326,
+      "step": 1606
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964885839103867e-05,
+      "loss": 1.2358,
+      "step": 1607
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996479143822902e-05,
+      "loss": 1.1092,
+      "step": 1608
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964696910854635e-05,
+      "loss": 1.1993,
+      "step": 1609
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964602256981903e-05,
+      "loss": 1.1435,
+      "step": 1610
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964507476612027e-05,
+      "loss": 1.227,
+      "step": 1611
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964412569746213e-05,
+      "loss": 1.1819,
+      "step": 1612
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996431753638566e-05,
+      "loss": 1.238,
+      "step": 1613
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964222376531584e-05,
+      "loss": 1.1708,
+      "step": 1614
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9964127090185185e-05,
+      "loss": 1.1747,
+      "step": 1615
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996403167734768e-05,
+      "loss": 1.162,
+      "step": 1616
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996393613802027e-05,
+      "loss": 1.1711,
+      "step": 1617
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963840472204178e-05,
+      "loss": 1.1366,
+      "step": 1618
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996374467990061e-05,
+      "loss": 1.1659,
+      "step": 1619
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963648761110787e-05,
+      "loss": 1.152,
+      "step": 1620
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996355271583593e-05,
+      "loss": 1.2856,
+      "step": 1621
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963456544077247e-05,
+      "loss": 1.1762,
+      "step": 1622
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996336024583597e-05,
+      "loss": 1.1333,
+      "step": 1623
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963263821113317e-05,
+      "loss": 1.1015,
+      "step": 1624
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9963167269910513e-05,
+      "loss": 1.1136,
+      "step": 1625
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996307059222878e-05,
+      "loss": 1.1837,
+      "step": 1626
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962973788069352e-05,
+      "loss": 1.0716,
+      "step": 1627
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996287685743345e-05,
+      "loss": 1.1772,
+      "step": 1628
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962779800322313e-05,
+      "loss": 1.2113,
+      "step": 1629
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962682616737163e-05,
+      "loss": 1.1785,
+      "step": 1630
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962585306679246e-05,
+      "loss": 1.0338,
+      "step": 1631
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962487870149786e-05,
+      "loss": 1.2038,
+      "step": 1632
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962390307150027e-05,
+      "loss": 1.3641,
+      "step": 1633
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962292617681204e-05,
+      "loss": 1.2262,
+      "step": 1634
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996219480174456e-05,
+      "loss": 1.1555,
+      "step": 1635
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9962096859341334e-05,
+      "loss": 1.2389,
+      "step": 1636
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996199879047277e-05,
+      "loss": 1.2427,
+      "step": 1637
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961900595140117e-05,
+      "loss": 1.2548,
+      "step": 1638
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961802273344615e-05,
+      "loss": 1.0963,
+      "step": 1639
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961703825087513e-05,
+      "loss": 1.1468,
+      "step": 1640
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961605250370067e-05,
+      "loss": 1.3431,
+      "step": 1641
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961506549193522e-05,
+      "loss": 1.233,
+      "step": 1642
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961407721559135e-05,
+      "loss": 1.2654,
+      "step": 1643
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961308767468158e-05,
+      "loss": 1.0823,
+      "step": 1644
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996120968692185e-05,
+      "loss": 1.123,
+      "step": 1645
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9961110479921463e-05,
+      "loss": 1.0923,
+      "step": 1646
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996101114646826e-05,
+      "loss": 1.3123,
+      "step": 1647
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960911686563508e-05,
+      "loss": 1.1708,
+      "step": 1648
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960812100208456e-05,
+      "loss": 1.1197,
+      "step": 1649
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960712387404382e-05,
+      "loss": 1.0887,
+      "step": 1650
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960612548152543e-05,
+      "loss": 1.257,
+      "step": 1651
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960512582454213e-05,
+      "loss": 1.2127,
+      "step": 1652
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996041249031065e-05,
+      "loss": 1.1946,
+      "step": 1653
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996031227172314e-05,
+      "loss": 1.066,
+      "step": 1654
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9960211926692943e-05,
+      "loss": 1.1245,
+      "step": 1655
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.996011145522134e-05,
+      "loss": 1.212,
+      "step": 1656
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99600108573096e-05,
+      "loss": 1.1425,
+      "step": 1657
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959910132959005e-05,
+      "loss": 1.328,
+      "step": 1658
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995980928217083e-05,
+      "loss": 1.1776,
+      "step": 1659
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959708304946362e-05,
+      "loss": 1.3515,
+      "step": 1660
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959607201286874e-05,
+      "loss": 1.2097,
+      "step": 1661
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959505971193657e-05,
+      "loss": 1.0999,
+      "step": 1662
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995940461466799e-05,
+      "loss": 1.0691,
+      "step": 1663
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959303131711162e-05,
+      "loss": 1.2928,
+      "step": 1664
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959201522324466e-05,
+      "loss": 1.1984,
+      "step": 1665
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9959099786509188e-05,
+      "loss": 1.1559,
+      "step": 1666
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958997924266616e-05,
+      "loss": 1.1625,
+      "step": 1667
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958895935598047e-05,
+      "loss": 1.1304,
+      "step": 1668
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958793820504777e-05,
+      "loss": 1.0239,
+      "step": 1669
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958691578988095e-05,
+      "loss": 1.1944,
+      "step": 1670
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995858921104931e-05,
+      "loss": 1.1642,
+      "step": 1671
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958486716689715e-05,
+      "loss": 1.1413,
+      "step": 1672
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958384095910613e-05,
+      "loss": 1.0509,
+      "step": 1673
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958281348713303e-05,
+      "loss": 1.2564,
+      "step": 1674
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958178475099094e-05,
+      "loss": 1.1966,
+      "step": 1675
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9958075475069287e-05,
+      "loss": 1.0474,
+      "step": 1676
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957972348625193e-05,
+      "loss": 1.1018,
+      "step": 1677
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957869095768126e-05,
+      "loss": 1.2234,
+      "step": 1678
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957765716499385e-05,
+      "loss": 1.0931,
+      "step": 1679
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957662210820293e-05,
+      "loss": 1.1585,
+      "step": 1680
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957558578732156e-05,
+      "loss": 1.0954,
+      "step": 1681
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99574548202363e-05,
+      "loss": 1.0718,
+      "step": 1682
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957350935334028e-05,
+      "loss": 1.1759,
+      "step": 1683
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995724692402667e-05,
+      "loss": 1.2021,
+      "step": 1684
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957142786315543e-05,
+      "loss": 1.1218,
+      "step": 1685
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9957038522201968e-05,
+      "loss": 1.2442,
+      "step": 1686
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995693413168727e-05,
+      "loss": 1.1437,
+      "step": 1687
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995682961477277e-05,
+      "loss": 1.1074,
+      "step": 1688
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956724971459804e-05,
+      "loss": 1.1028,
+      "step": 1689
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995662020174969e-05,
+      "loss": 1.0925,
+      "step": 1690
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956515305643767e-05,
+      "loss": 1.2519,
+      "step": 1691
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995641028314336e-05,
+      "loss": 1.131,
+      "step": 1692
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956305134249804e-05,
+      "loss": 1.259,
+      "step": 1693
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9956199858964434e-05,
+      "loss": 1.1445,
+      "step": 1694
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995609445728859e-05,
+      "loss": 1.0491,
+      "step": 1695
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955988929223604e-05,
+      "loss": 1.201,
+      "step": 1696
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955883274770817e-05,
+      "loss": 1.0968,
+      "step": 1697
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955777493931573e-05,
+      "loss": 1.0255,
+      "step": 1698
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955671586707213e-05,
+      "loss": 1.1342,
+      "step": 1699
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955565553099084e-05,
+      "loss": 1.2179,
+      "step": 1700
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955459393108533e-05,
+      "loss": 1.2995,
+      "step": 1701
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955353106736902e-05,
+      "loss": 1.0759,
+      "step": 1702
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955246693985543e-05,
+      "loss": 1.1041,
+      "step": 1703
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955140154855804e-05,
+      "loss": 1.1263,
+      "step": 1704
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9955033489349045e-05,
+      "loss": 1.1487,
+      "step": 1705
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954926697466616e-05,
+      "loss": 1.1582,
+      "step": 1706
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995481977920987e-05,
+      "loss": 1.0262,
+      "step": 1707
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954712734580165e-05,
+      "loss": 1.1042,
+      "step": 1708
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995460556357886e-05,
+      "loss": 0.9176,
+      "step": 1709
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954498266207322e-05,
+      "loss": 1.2498,
+      "step": 1710
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954390842466904e-05,
+      "loss": 1.0605,
+      "step": 1711
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954283292358975e-05,
+      "loss": 0.9554,
+      "step": 1712
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954175615884897e-05,
+      "loss": 1.1789,
+      "step": 1713
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9954067813046044e-05,
+      "loss": 1.2067,
+      "step": 1714
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953959883843773e-05,
+      "loss": 1.1767,
+      "step": 1715
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953851828279466e-05,
+      "loss": 1.2147,
+      "step": 1716
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953743646354487e-05,
+      "loss": 1.1244,
+      "step": 1717
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953635338070213e-05,
+      "loss": 1.0869,
+      "step": 1718
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953526903428017e-05,
+      "loss": 1.2058,
+      "step": 1719
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953418342429275e-05,
+      "loss": 1.2762,
+      "step": 1720
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953309655075368e-05,
+      "loss": 1.0928,
+      "step": 1721
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995320084136767e-05,
+      "loss": 1.1429,
+      "step": 1722
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9953091901307572e-05,
+      "loss": 1.1831,
+      "step": 1723
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952982834896447e-05,
+      "loss": 1.2616,
+      "step": 1724
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952873642135687e-05,
+      "loss": 1.0768,
+      "step": 1725
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995276432302667e-05,
+      "loss": 1.2884,
+      "step": 1726
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952654877570793e-05,
+      "loss": 0.9873,
+      "step": 1727
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995254530576944e-05,
+      "loss": 1.2327,
+      "step": 1728
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952435607624002e-05,
+      "loss": 1.1717,
+      "step": 1729
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9952325783135876e-05,
+      "loss": 1.1531,
+      "step": 1730
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995221583230645e-05,
+      "loss": 1.1937,
+      "step": 1731
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995210575513712e-05,
+      "loss": 1.1063,
+      "step": 1732
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951995551629293e-05,
+      "loss": 1.1935,
+      "step": 1733
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951885221784356e-05,
+      "loss": 1.1398,
+      "step": 1734
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951774765603716e-05,
+      "loss": 1.1322,
+      "step": 1735
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951664183088776e-05,
+      "loss": 1.176,
+      "step": 1736
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951553474240935e-05,
+      "loss": 1.122,
+      "step": 1737
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951442639061603e-05,
+      "loss": 1.2231,
+      "step": 1738
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951331677552183e-05,
+      "loss": 1.1744,
+      "step": 1739
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951220589714087e-05,
+      "loss": 1.1611,
+      "step": 1740
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9951109375548724e-05,
+      "loss": 1.0383,
+      "step": 1741
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99509980350575e-05,
+      "loss": 1.1799,
+      "step": 1742
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950886568241845e-05,
+      "loss": 1.2141,
+      "step": 1743
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950774975103154e-05,
+      "loss": 1.2199,
+      "step": 1744
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995066325564286e-05,
+      "loss": 1.0799,
+      "step": 1745
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950551409862368e-05,
+      "loss": 1.0564,
+      "step": 1746
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.995043943776311e-05,
+      "loss": 1.0672,
+      "step": 1747
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950327339346496e-05,
+      "loss": 1.2265,
+      "step": 1748
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950215114613954e-05,
+      "loss": 1.1375,
+      "step": 1749
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9950102763566913e-05,
+      "loss": 1.1584,
+      "step": 1750
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949990286206795e-05,
+      "loss": 1.179,
+      "step": 1751
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994987768253503e-05,
+      "loss": 1.1963,
+      "step": 1752
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994976495255304e-05,
+      "loss": 1.0942,
+      "step": 1753
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949652096262264e-05,
+      "loss": 1.2927,
+      "step": 1754
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949539113664133e-05,
+      "loss": 1.1364,
+      "step": 1755
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949426004760078e-05,
+      "loss": 1.111,
+      "step": 1756
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949312769551542e-05,
+      "loss": 1.1032,
+      "step": 1757
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9949199408039958e-05,
+      "loss": 1.2163,
+      "step": 1758
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994908592022676e-05,
+      "loss": 1.3198,
+      "step": 1759
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99489723061134e-05,
+      "loss": 1.1466,
+      "step": 1760
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994885856570131e-05,
+      "loss": 1.3639,
+      "step": 1761
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994874469899194e-05,
+      "loss": 1.115,
+      "step": 1762
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994863070598673e-05,
+      "loss": 1.1936,
+      "step": 1763
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948516586687133e-05,
+      "loss": 1.1989,
+      "step": 1764
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948402341094597e-05,
+      "loss": 1.0786,
+      "step": 1765
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994828796921057e-05,
+      "loss": 1.1421,
+      "step": 1766
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948173471036502e-05,
+      "loss": 1.2527,
+      "step": 1767
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9948058846573853e-05,
+      "loss": 1.1893,
+      "step": 1768
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994794409582407e-05,
+      "loss": 1.1065,
+      "step": 1769
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947829218788615e-05,
+      "loss": 0.9938,
+      "step": 1770
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947714215468946e-05,
+      "loss": 1.1924,
+      "step": 1771
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994759908586652e-05,
+      "loss": 1.0358,
+      "step": 1772
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947483829982803e-05,
+      "loss": 1.1281,
+      "step": 1773
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994736844781926e-05,
+      "loss": 1.2334,
+      "step": 1774
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947252939377343e-05,
+      "loss": 1.1772,
+      "step": 1775
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9947137304658532e-05,
+      "loss": 1.271,
+      "step": 1776
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994702154366429e-05,
+      "loss": 1.1942,
+      "step": 1777
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946905656396084e-05,
+      "loss": 1.2047,
+      "step": 1778
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946789642855388e-05,
+      "loss": 1.1883,
+      "step": 1779
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946673503043675e-05,
+      "loss": 1.3018,
+      "step": 1780
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994655723696242e-05,
+      "loss": 1.1052,
+      "step": 1781
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994644084461309e-05,
+      "loss": 1.1387,
+      "step": 1782
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946324325997182e-05,
+      "loss": 1.2647,
+      "step": 1783
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9946207681116155e-05,
+      "loss": 1.0269,
+      "step": 1784
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99460909099715e-05,
+      "loss": 1.2134,
+      "step": 1785
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945974012564698e-05,
+      "loss": 1.1486,
+      "step": 1786
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945856988897235e-05,
+      "loss": 1.0447,
+      "step": 1787
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994573983897059e-05,
+      "loss": 1.1995,
+      "step": 1788
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945622562786252e-05,
+      "loss": 1.1202,
+      "step": 1789
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945505160345716e-05,
+      "loss": 1.2159,
+      "step": 1790
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945387631650468e-05,
+      "loss": 1.169,
+      "step": 1791
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945269976702e-05,
+      "loss": 1.124,
+      "step": 1792
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9945152195501806e-05,
+      "loss": 1.1663,
+      "step": 1793
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994503428805138e-05,
+      "loss": 1.2428,
+      "step": 1794
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994491625435222e-05,
+      "loss": 1.0637,
+      "step": 1795
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944798094405827e-05,
+      "loss": 1.2254,
+      "step": 1796
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944679808213693e-05,
+      "loss": 1.2421,
+      "step": 1797
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944561395777333e-05,
+      "loss": 1.1823,
+      "step": 1798
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944442857098233e-05,
+      "loss": 1.1529,
+      "step": 1799
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944324192177912e-05,
+      "loss": 1.2978,
+      "step": 1800
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994420540101787e-05,
+      "loss": 1.131,
+      "step": 1801
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9944086483619617e-05,
+      "loss": 1.1042,
+      "step": 1802
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994396743998466e-05,
+      "loss": 1.1351,
+      "step": 1803
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943848270114516e-05,
+      "loss": 1.1273,
+      "step": 1804
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994372897401069e-05,
+      "loss": 1.0769,
+      "step": 1805
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943609551674704e-05,
+      "loss": 1.0039,
+      "step": 1806
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943490003108068e-05,
+      "loss": 1.218,
+      "step": 1807
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943370328312302e-05,
+      "loss": 1.1085,
+      "step": 1808
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943250527288925e-05,
+      "loss": 1.0835,
+      "step": 1809
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994313060003946e-05,
+      "loss": 1.144,
+      "step": 1810
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9943010546565424e-05,
+      "loss": 1.189,
+      "step": 1811
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942890366868348e-05,
+      "loss": 1.1378,
+      "step": 1812
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942770060949753e-05,
+      "loss": 1.1704,
+      "step": 1813
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942649628811167e-05,
+      "loss": 1.0669,
+      "step": 1814
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942529070454117e-05,
+      "loss": 1.0942,
+      "step": 1815
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994240838588014e-05,
+      "loss": 1.0772,
+      "step": 1816
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994228757509076e-05,
+      "loss": 1.116,
+      "step": 1817
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942166638087518e-05,
+      "loss": 1.1529,
+      "step": 1818
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9942045574871947e-05,
+      "loss": 1.1255,
+      "step": 1819
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941924385445577e-05,
+      "loss": 1.2234,
+      "step": 1820
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941803069809956e-05,
+      "loss": 0.9843,
+      "step": 1821
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994168162796662e-05,
+      "loss": 1.145,
+      "step": 1822
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941560059917108e-05,
+      "loss": 1.1183,
+      "step": 1823
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941438365662967e-05,
+      "loss": 1.2805,
+      "step": 1824
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941316545205742e-05,
+      "loss": 1.014,
+      "step": 1825
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9941194598546977e-05,
+      "loss": 1.173,
+      "step": 1826
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994107252568822e-05,
+      "loss": 1.2624,
+      "step": 1827
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940950326631027e-05,
+      "loss": 1.2503,
+      "step": 1828
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.994082800137694e-05,
+      "loss": 1.1275,
+      "step": 1829
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940705549927516e-05,
+      "loss": 1.2367,
+      "step": 1830
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940582972284313e-05,
+      "loss": 1.0441,
+      "step": 1831
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940460268448882e-05,
+      "loss": 1.2116,
+      "step": 1832
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940337438422784e-05,
+      "loss": 1.1352,
+      "step": 1833
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940214482207577e-05,
+      "loss": 1.1995,
+      "step": 1834
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9940091399804817e-05,
+      "loss": 1.094,
+      "step": 1835
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939968191216076e-05,
+      "loss": 1.0824,
+      "step": 1836
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993984485644291e-05,
+      "loss": 1.29,
+      "step": 1837
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993972139548689e-05,
+      "loss": 1.0934,
+      "step": 1838
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993959780834958e-05,
+      "loss": 1.1443,
+      "step": 1839
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993947409503255e-05,
+      "loss": 1.1284,
+      "step": 1840
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993935025553737e-05,
+      "loss": 1.0426,
+      "step": 1841
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939226289865613e-05,
+      "loss": 1.274,
+      "step": 1842
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9939102198018853e-05,
+      "loss": 1.1727,
+      "step": 1843
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938977979998665e-05,
+      "loss": 1.1631,
+      "step": 1844
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938853635806625e-05,
+      "loss": 1.1911,
+      "step": 1845
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938729165444314e-05,
+      "loss": 1.1749,
+      "step": 1846
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938604568913308e-05,
+      "loss": 1.1084,
+      "step": 1847
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993847984621519e-05,
+      "loss": 1.0416,
+      "step": 1848
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938354997351547e-05,
+      "loss": 1.1918,
+      "step": 1849
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938230022323957e-05,
+      "loss": 1.1802,
+      "step": 1850
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9938104921134018e-05,
+      "loss": 1.1823,
+      "step": 1851
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.99379796937833e-05,
+      "loss": 1.0984,
+      "step": 1852
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993785434027341e-05,
+      "loss": 1.1176,
+      "step": 1853
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937728860605932e-05,
+      "loss": 1.1057,
+      "step": 1854
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937603254782464e-05,
+      "loss": 1.1991,
+      "step": 1855
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937477522804592e-05,
+      "loss": 1.1218,
+      "step": 1856
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937351664673915e-05,
+      "loss": 1.0299,
+      "step": 1857
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937225680392033e-05,
+      "loss": 1.2909,
+      "step": 1858
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9937099569960545e-05,
+      "loss": 1.0245,
+      "step": 1859
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936973333381053e-05,
+      "loss": 1.0169,
+      "step": 1860
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936846970655157e-05,
+      "loss": 1.0401,
+      "step": 1861
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993672048178446e-05,
+      "loss": 1.0617,
+      "step": 1862
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936593866770572e-05,
+      "loss": 0.9924,
+      "step": 1863
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936467125615096e-05,
+      "loss": 1.276,
+      "step": 1864
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936340258319645e-05,
+      "loss": 1.2318,
+      "step": 1865
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936213264885827e-05,
+      "loss": 1.1224,
+      "step": 1866
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.9936086145315253e-05,
+      "loss": 1.1805,
+      "step": 1867
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 1.993595889960954e-05,
+      "loss": 0.9676,
+      "step": 1868
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935831527770303e-05,
+      "loss": 1.1077,
+      "step": 1869
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935704029799155e-05,
+      "loss": 1.25,
+      "step": 1870
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993557640569772e-05,
+      "loss": 1.1449,
+      "step": 1871
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935448655467614e-05,
+      "loss": 1.1509,
+      "step": 1872
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993532077911046e-05,
+      "loss": 1.2045,
+      "step": 1873
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993519277662788e-05,
+      "loss": 1.1797,
+      "step": 1874
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9935064648021502e-05,
+      "loss": 1.1482,
+      "step": 1875
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993493639329295e-05,
+      "loss": 1.1234,
+      "step": 1876
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934808012443852e-05,
+      "loss": 1.1434,
+      "step": 1877
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934679505475842e-05,
+      "loss": 1.3194,
+      "step": 1878
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993455087239055e-05,
+      "loss": 1.2031,
+      "step": 1879
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934422113189603e-05,
+      "loss": 1.2247,
+      "step": 1880
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993429322787464e-05,
+      "loss": 1.1779,
+      "step": 1881
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9934164216447297e-05,
+      "loss": 1.1356,
+      "step": 1882
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993403507890921e-05,
+      "loss": 1.2135,
+      "step": 1883
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933905815262025e-05,
+      "loss": 1.2113,
+      "step": 1884
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933776425507374e-05,
+      "loss": 1.1366,
+      "step": 1885
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933646909646908e-05,
+      "loss": 0.9868,
+      "step": 1886
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993351726768226e-05,
+      "loss": 1.0142,
+      "step": 1887
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993338749961509e-05,
+      "loss": 1.1198,
+      "step": 1888
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9933257605447028e-05,
+      "loss": 1.1352,
+      "step": 1889
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993312758517974e-05,
+      "loss": 1.1515,
+      "step": 1890
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932997438814867e-05,
+      "loss": 1.2558,
+      "step": 1891
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932867166354066e-05,
+      "loss": 1.1831,
+      "step": 1892
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932736767798987e-05,
+      "loss": 1.0717,
+      "step": 1893
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932606243151283e-05,
+      "loss": 1.1276,
+      "step": 1894
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932475592412618e-05,
+      "loss": 1.1484,
+      "step": 1895
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932344815584646e-05,
+      "loss": 1.2527,
+      "step": 1896
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9932213912669033e-05,
+      "loss": 1.2881,
+      "step": 1897
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993208288366743e-05,
+      "loss": 1.1947,
+      "step": 1898
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931951728581506e-05,
+      "loss": 1.1063,
+      "step": 1899
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993182044741293e-05,
+      "loss": 1.1702,
+      "step": 1900
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931689040163366e-05,
+      "loss": 1.2666,
+      "step": 1901
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.993155750683448e-05,
+      "loss": 1.2547,
+      "step": 1902
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931425847427938e-05,
+      "loss": 1.1867,
+      "step": 1903
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931294061945417e-05,
+      "loss": 1.111,
+      "step": 1904
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931162150388595e-05,
+      "loss": 1.0392,
+      "step": 1905
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9931030112759138e-05,
+      "loss": 1.0362,
+      "step": 1906
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930897949058724e-05,
+      "loss": 1.2538,
+      "step": 1907
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930765659289033e-05,
+      "loss": 1.1528,
+      "step": 1908
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930633243451742e-05,
+      "loss": 1.1377,
+      "step": 1909
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930500701548535e-05,
+      "loss": 1.1398,
+      "step": 1910
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930368033581093e-05,
+      "loss": 1.2231,
+      "step": 1911
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930235239551096e-05,
+      "loss": 1.1383,
+      "step": 1912
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9930102319460238e-05,
+      "loss": 1.2626,
+      "step": 1913
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99299692733102e-05,
+      "loss": 1.091,
+      "step": 1914
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992983610110267e-05,
+      "loss": 1.0879,
+      "step": 1915
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929702802839344e-05,
+      "loss": 1.1669,
+      "step": 1916
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929569378521916e-05,
+      "loss": 1.2787,
+      "step": 1917
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992943582815207e-05,
+      "loss": 1.133,
+      "step": 1918
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992930215173151e-05,
+      "loss": 1.0415,
+      "step": 1919
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929168349261928e-05,
+      "loss": 1.3329,
+      "step": 1920
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9929034420745024e-05,
+      "loss": 1.1374,
+      "step": 1921
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99289003661825e-05,
+      "loss": 1.184,
+      "step": 1922
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928766185576057e-05,
+      "loss": 1.0296,
+      "step": 1923
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928631878927396e-05,
+      "loss": 1.0674,
+      "step": 1924
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928497446238223e-05,
+      "loss": 1.1594,
+      "step": 1925
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928362887510246e-05,
+      "loss": 1.1982,
+      "step": 1926
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928228202745173e-05,
+      "loss": 1.2235,
+      "step": 1927
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9928093391944714e-05,
+      "loss": 1.2572,
+      "step": 1928
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992795845511058e-05,
+      "loss": 1.2497,
+      "step": 1929
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992782339224448e-05,
+      "loss": 1.1627,
+      "step": 1930
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927688203348134e-05,
+      "loss": 1.0956,
+      "step": 1931
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927552888423257e-05,
+      "loss": 1.0595,
+      "step": 1932
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927417447471566e-05,
+      "loss": 1.0977,
+      "step": 1933
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927281880494783e-05,
+      "loss": 1.1268,
+      "step": 1934
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9927146187494622e-05,
+      "loss": 1.1014,
+      "step": 1935
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992701036847281e-05,
+      "loss": 1.1391,
+      "step": 1936
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992687442343108e-05,
+      "loss": 1.1634,
+      "step": 1937
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926738352371138e-05,
+      "loss": 1.2259,
+      "step": 1938
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926602155294728e-05,
+      "loss": 1.1097,
+      "step": 1939
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926465832203573e-05,
+      "loss": 1.1802,
+      "step": 1940
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99263293830994e-05,
+      "loss": 1.106,
+      "step": 1941
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992619280798395e-05,
+      "loss": 1.127,
+      "step": 1942
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9926056106858948e-05,
+      "loss": 0.9992,
+      "step": 1943
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925919279726136e-05,
+      "loss": 1.1946,
+      "step": 1944
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992578232658725e-05,
+      "loss": 1.2176,
+      "step": 1945
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992564524744402e-05,
+      "loss": 1.0454,
+      "step": 1946
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99255080422982e-05,
+      "loss": 1.0704,
+      "step": 1947
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992537071115152e-05,
+      "loss": 1.0004,
+      "step": 1948
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992523325400573e-05,
+      "loss": 1.2545,
+      "step": 1949
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9925095670862572e-05,
+      "loss": 1.1423,
+      "step": 1950
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924957961723796e-05,
+      "loss": 1.2306,
+      "step": 1951
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924820126591148e-05,
+      "loss": 1.1362,
+      "step": 1952
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924682165466375e-05,
+      "loss": 1.1693,
+      "step": 1953
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992454407835123e-05,
+      "loss": 1.1095,
+      "step": 1954
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924405865247468e-05,
+      "loss": 1.2578,
+      "step": 1955
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9924267526156848e-05,
+      "loss": 1.1821,
+      "step": 1956
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992412906108111e-05,
+      "loss": 1.0738,
+      "step": 1957
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923990470022032e-05,
+      "loss": 1.1669,
+      "step": 1958
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992385175298136e-05,
+      "loss": 1.0781,
+      "step": 1959
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992371290996086e-05,
+      "loss": 1.1539,
+      "step": 1960
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923573940962294e-05,
+      "loss": 1.0934,
+      "step": 1961
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992343484598742e-05,
+      "loss": 1.1331,
+      "step": 1962
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923295625038015e-05,
+      "loss": 1.1796,
+      "step": 1963
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9923156278115837e-05,
+      "loss": 1.086,
+      "step": 1964
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992301680522266e-05,
+      "loss": 1.1563,
+      "step": 1965
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922877206360254e-05,
+      "loss": 1.1514,
+      "step": 1966
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992273748153039e-05,
+      "loss": 1.0983,
+      "step": 1967
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992259763073484e-05,
+      "loss": 1.2707,
+      "step": 1968
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922457653975385e-05,
+      "loss": 1.05,
+      "step": 1969
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922317551253796e-05,
+      "loss": 1.3255,
+      "step": 1970
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9922177322571855e-05,
+      "loss": 1.0878,
+      "step": 1971
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992203696793134e-05,
+      "loss": 1.0065,
+      "step": 1972
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921896487334035e-05,
+      "loss": 1.1811,
+      "step": 1973
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921755880781718e-05,
+      "loss": 1.2242,
+      "step": 1974
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921615148276184e-05,
+      "loss": 1.1328,
+      "step": 1975
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992147428981921e-05,
+      "loss": 1.0958,
+      "step": 1976
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921333305412588e-05,
+      "loss": 1.2305,
+      "step": 1977
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921192195058108e-05,
+      "loss": 1.2475,
+      "step": 1978
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9921050958757563e-05,
+      "loss": 1.1312,
+      "step": 1979
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992090959651274e-05,
+      "loss": 1.2208,
+      "step": 1980
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920768108325438e-05,
+      "loss": 1.1791,
+      "step": 1981
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920626494197454e-05,
+      "loss": 1.0746,
+      "step": 1982
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920484754130587e-05,
+      "loss": 1.199,
+      "step": 1983
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992034288812663e-05,
+      "loss": 1.1843,
+      "step": 1984
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.992020089618739e-05,
+      "loss": 1.1731,
+      "step": 1985
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9920058778314666e-05,
+      "loss": 1.1563,
+      "step": 1986
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991991653451026e-05,
+      "loss": 1.1722,
+      "step": 1987
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919774164775983e-05,
+      "loss": 1.2441,
+      "step": 1988
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919631669113642e-05,
+      "loss": 1.1204,
+      "step": 1989
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919489047525045e-05,
+      "loss": 1.0428,
+      "step": 1990
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9919346300012e-05,
+      "loss": 1.0395,
+      "step": 1991
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991920342657632e-05,
+      "loss": 1.2975,
+      "step": 1992
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991906042721982e-05,
+      "loss": 1.1601,
+      "step": 1993
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918917301944315e-05,
+      "loss": 1.1719,
+      "step": 1994
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918774050751624e-05,
+      "loss": 1.1746,
+      "step": 1995
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918630673643563e-05,
+      "loss": 1.1122,
+      "step": 1996
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991848717062195e-05,
+      "loss": 1.1339,
+      "step": 1997
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918343541688608e-05,
+      "loss": 1.229,
+      "step": 1998
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918199786845367e-05,
+      "loss": 1.1939,
+      "step": 1999
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9918055906094043e-05,
+      "loss": 1.0895,
+      "step": 2000
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917911899436465e-05,
+      "loss": 1.1535,
+      "step": 2001
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917767766874466e-05,
+      "loss": 1.169,
+      "step": 2002
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991762350840987e-05,
+      "loss": 1.1913,
+      "step": 2003
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991747912404451e-05,
+      "loss": 1.0971,
+      "step": 2004
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9917334613780216e-05,
+      "loss": 1.157,
+      "step": 2005
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991718997761883e-05,
+      "loss": 1.0449,
+      "step": 2006
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991704521556218e-05,
+      "loss": 1.038,
+      "step": 2007
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991690032761211e-05,
+      "loss": 1.1136,
+      "step": 2008
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916755313770456e-05,
+      "loss": 1.3263,
+      "step": 2009
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991661017403906e-05,
+      "loss": 1.0705,
+      "step": 2010
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916464908419762e-05,
+      "loss": 1.0701,
+      "step": 2011
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916319516914412e-05,
+      "loss": 1.0473,
+      "step": 2012
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916173999524848e-05,
+      "loss": 1.1748,
+      "step": 2013
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9916028356252924e-05,
+      "loss": 1.1421,
+      "step": 2014
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915882587100485e-05,
+      "loss": 1.0758,
+      "step": 2015
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991573669206938e-05,
+      "loss": 1.1513,
+      "step": 2016
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915590671161466e-05,
+      "loss": 1.2079,
+      "step": 2017
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9915444524378595e-05,
+      "loss": 1.1915,
+      "step": 2018
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991529825172262e-05,
+      "loss": 1.0881,
+      "step": 2019
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99151518531954e-05,
+      "loss": 1.1271,
+      "step": 2020
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991500532879879e-05,
+      "loss": 1.1423,
+      "step": 2021
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991485867853466e-05,
+      "loss": 1.1187,
+      "step": 2022
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991471190240486e-05,
+      "loss": 1.1498,
+      "step": 2023
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991456500041126e-05,
+      "loss": 1.1275,
+      "step": 2024
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991441797255572e-05,
+      "loss": 1.1642,
+      "step": 2025
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991427081884011e-05,
+      "loss": 1.0653,
+      "step": 2026
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.99141235392663e-05,
+      "loss": 1.1311,
+      "step": 2027
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913976133836156e-05,
+      "loss": 1.1398,
+      "step": 2028
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991382860255155e-05,
+      "loss": 1.1358,
+      "step": 2029
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913680945414354e-05,
+      "loss": 1.0956,
+      "step": 2030
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913533162426444e-05,
+      "loss": 1.1392,
+      "step": 2031
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913385253589698e-05,
+      "loss": 1.0728,
+      "step": 2032
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913237218905984e-05,
+      "loss": 1.2073,
+      "step": 2033
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9913089058377196e-05,
+      "loss": 1.1131,
+      "step": 2034
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912940772005205e-05,
+      "loss": 1.1427,
+      "step": 2035
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912792359791896e-05,
+      "loss": 1.1722,
+      "step": 2036
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991264382173915e-05,
+      "loss": 1.0468,
+      "step": 2037
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991249515784886e-05,
+      "loss": 1.1914,
+      "step": 2038
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991234636812291e-05,
+      "loss": 1.201,
+      "step": 2039
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912197452563182e-05,
+      "loss": 1.0377,
+      "step": 2040
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9912048411171575e-05,
+      "loss": 1.103,
+      "step": 2041
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991189924394998e-05,
+      "loss": 1.1984,
+      "step": 2042
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911749950900283e-05,
+      "loss": 1.0438,
+      "step": 2043
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991160053202439e-05,
+      "loss": 1.1612,
+      "step": 2044
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991145098732419e-05,
+      "loss": 1.2644,
+      "step": 2045
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911301316801584e-05,
+      "loss": 1.238,
+      "step": 2046
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9911151520458472e-05,
+      "loss": 1.1529,
+      "step": 2047
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991100159829676e-05,
+      "loss": 1.1244,
+      "step": 2048
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991085155031834e-05,
+      "loss": 1.154,
+      "step": 2049
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910701376525134e-05,
+      "loss": 1.1848,
+      "step": 2050
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991055107691903e-05,
+      "loss": 1.1802,
+      "step": 2051
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910400651501946e-05,
+      "loss": 1.0469,
+      "step": 2052
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.991025010027579e-05,
+      "loss": 1.029,
+      "step": 2053
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9910099423242474e-05,
+      "loss": 1.1058,
+      "step": 2054
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909948620403904e-05,
+      "loss": 1.0846,
+      "step": 2055
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990979769176201e-05,
+      "loss": 1.2085,
+      "step": 2056
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990964663731869e-05,
+      "loss": 1.083,
+      "step": 2057
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909495457075874e-05,
+      "loss": 1.1602,
+      "step": 2058
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909344151035474e-05,
+      "loss": 1.2351,
+      "step": 2059
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909192719199416e-05,
+      "loss": 1.1923,
+      "step": 2060
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9909041161569618e-05,
+      "loss": 1.0813,
+      "step": 2061
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908889478148007e-05,
+      "loss": 1.176,
+      "step": 2062
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908737668936507e-05,
+      "loss": 1.2239,
+      "step": 2063
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990858573393705e-05,
+      "loss": 1.2604,
+      "step": 2064
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908433673151555e-05,
+      "loss": 1.3102,
+      "step": 2065
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990828148658196e-05,
+      "loss": 1.1729,
+      "step": 2066
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9908129174230194e-05,
+      "loss": 1.1219,
+      "step": 2067
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907976736098195e-05,
+      "loss": 1.1243,
+      "step": 2068
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907824172187888e-05,
+      "loss": 1.1972,
+      "step": 2069
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907671482501222e-05,
+      "loss": 1.0998,
+      "step": 2070
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907518667040128e-05,
+      "loss": 1.1827,
+      "step": 2071
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907365725806547e-05,
+      "loss": 1.138,
+      "step": 2072
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990721265880242e-05,
+      "loss": 1.162,
+      "step": 2073
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9907059466029693e-05,
+      "loss": 1.1615,
+      "step": 2074
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906906147490307e-05,
+      "loss": 1.1334,
+      "step": 2075
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906752703186213e-05,
+      "loss": 1.0865,
+      "step": 2076
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906599133119357e-05,
+      "loss": 1.1933,
+      "step": 2077
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906445437291685e-05,
+      "loss": 1.1484,
+      "step": 2078
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990629161570515e-05,
+      "loss": 1.2304,
+      "step": 2079
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9906137668361708e-05,
+      "loss": 1.1312,
+      "step": 2080
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990598359526331e-05,
+      "loss": 1.2406,
+      "step": 2081
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990582939641191e-05,
+      "loss": 1.0653,
+      "step": 2082
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905675071809474e-05,
+      "loss": 1.0998,
+      "step": 2083
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905520621457952e-05,
+      "loss": 1.1643,
+      "step": 2084
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990536604535931e-05,
+      "loss": 1.0996,
+      "step": 2085
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905211343515503e-05,
+      "loss": 1.0381,
+      "step": 2086
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9905056515928507e-05,
+      "loss": 1.1169,
+      "step": 2087
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904901562600275e-05,
+      "loss": 1.1396,
+      "step": 2088
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904746483532785e-05,
+      "loss": 1.0737,
+      "step": 2089
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904591278727997e-05,
+      "loss": 1.1346,
+      "step": 2090
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904435948187886e-05,
+      "loss": 1.1707,
+      "step": 2091
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9904280491914422e-05,
+      "loss": 1.3344,
+      "step": 2092
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990412490990958e-05,
+      "loss": 1.0718,
+      "step": 2093
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903969202175333e-05,
+      "loss": 1.1805,
+      "step": 2094
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990381336871366e-05,
+      "loss": 1.124,
+      "step": 2095
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903657409526535e-05,
+      "loss": 1.1389,
+      "step": 2096
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903501324615944e-05,
+      "loss": 1.0942,
+      "step": 2097
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903345113983864e-05,
+      "loss": 1.147,
+      "step": 2098
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903188777632282e-05,
+      "loss": 1.2363,
+      "step": 2099
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9903032315563178e-05,
+      "loss": 1.1315,
+      "step": 2100
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902875727778535e-05,
+      "loss": 1.1788,
+      "step": 2101
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902719014280352e-05,
+      "loss": 1.1243,
+      "step": 2102
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902562175070615e-05,
+      "loss": 1.1198,
+      "step": 2103
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990240521015131e-05,
+      "loss": 1.1169,
+      "step": 2104
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9902248119524428e-05,
+      "loss": 1.0923,
+      "step": 2105
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990209090319197e-05,
+      "loss": 1.151,
+      "step": 2106
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990193356115593e-05,
+      "loss": 1.0838,
+      "step": 2107
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901776093418306e-05,
+      "loss": 1.1029,
+      "step": 2108
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990161849998109e-05,
+      "loss": 1.0277,
+      "step": 2109
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901460780846294e-05,
+      "loss": 1.2503,
+      "step": 2110
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9901302936015914e-05,
+      "loss": 1.1329,
+      "step": 2111
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990114496549195e-05,
+      "loss": 1.0296,
+      "step": 2112
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990098686927641e-05,
+      "loss": 1.0234,
+      "step": 2113
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.990082864737131e-05,
+      "loss": 1.0837,
+      "step": 2114
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900670299778647e-05,
+      "loss": 1.2425,
+      "step": 2115
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900511826500435e-05,
+      "loss": 1.1158,
+      "step": 2116
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900353227538686e-05,
+      "loss": 1.1085,
+      "step": 2117
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900194502895417e-05,
+      "loss": 1.2738,
+      "step": 2118
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9900035652572633e-05,
+      "loss": 1.1026,
+      "step": 2119
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989987667657236e-05,
+      "loss": 1.0264,
+      "step": 2120
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989971757489661e-05,
+      "loss": 1.1678,
+      "step": 2121
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899558347547413e-05,
+      "loss": 1.1719,
+      "step": 2122
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899398994526777e-05,
+      "loss": 1.3301,
+      "step": 2123
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989923951583673e-05,
+      "loss": 1.1796,
+      "step": 2124
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9899079911479303e-05,
+      "loss": 1.3407,
+      "step": 2125
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898920181456512e-05,
+      "loss": 1.1581,
+      "step": 2126
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989876032577039e-05,
+      "loss": 1.1285,
+      "step": 2127
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898600344422966e-05,
+      "loss": 1.2123,
+      "step": 2128
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898440237416273e-05,
+      "loss": 1.0757,
+      "step": 2129
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898280004752337e-05,
+      "loss": 1.0516,
+      "step": 2130
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9898119646433198e-05,
+      "loss": 1.0677,
+      "step": 2131
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897959162460888e-05,
+      "loss": 1.2243,
+      "step": 2132
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989779855283745e-05,
+      "loss": 1.37,
+      "step": 2133
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897637817564917e-05,
+      "loss": 1.1662,
+      "step": 2134
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989747695664533e-05,
+      "loss": 1.0906,
+      "step": 2135
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897315970080737e-05,
+      "loss": 1.0795,
+      "step": 2136
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9897154857873174e-05,
+      "loss": 1.236,
+      "step": 2137
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989699362002469e-05,
+      "loss": 1.2123,
+      "step": 2138
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989683225653733e-05,
+      "loss": 1.0359,
+      "step": 2139
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896670767413148e-05,
+      "loss": 1.0227,
+      "step": 2140
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896509152654185e-05,
+      "loss": 1.2309,
+      "step": 2141
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98963474122625e-05,
+      "loss": 1.1707,
+      "step": 2142
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989618554624014e-05,
+      "loss": 1.1532,
+      "step": 2143
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9896023554589168e-05,
+      "loss": 1.2227,
+      "step": 2144
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895861437311635e-05,
+      "loss": 1.1285,
+      "step": 2145
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.98956991944096e-05,
+      "loss": 1.0935,
+      "step": 2146
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.989553682588512e-05,
+      "loss": 1.0211,
+      "step": 2147
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895374331740266e-05,
+      "loss": 1.2016,
+      "step": 2148
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895211711977083e-05,
+      "loss": 1.148,
+      "step": 2149
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9895048966597655e-05,
+      "loss": 1.1004,
+      "step": 2150
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894886095604036e-05,
+      "loss": 1.1163,
+      "step": 2151
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894723098998298e-05,
+      "loss": 1.2098,
+      "step": 2152
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894559976782507e-05,
+      "loss": 1.1896,
+      "step": 2153
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894396728958732e-05,
+      "loss": 1.0956,
+      "step": 2154
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.9894233355529056e-05,
+      "loss": 1.0198,
+      "step": 2155
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9894069856495544e-05,
+      "loss": 1.041,
+      "step": 2156
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893906231860273e-05,
+      "loss": 1.0783,
+      "step": 2157
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893742481625316e-05,
+      "loss": 1.1743,
+      "step": 2158
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989357860579276e-05,
+      "loss": 1.3273,
+      "step": 2159
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893414604364686e-05,
+      "loss": 1.1993,
+      "step": 2160
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9893250477343166e-05,
+      "loss": 1.1912,
+      "step": 2161
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989308622473029e-05,
+      "loss": 1.1265,
+      "step": 2162
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892921846528147e-05,
+      "loss": 1.2102,
+      "step": 2163
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892757342738817e-05,
+      "loss": 1.0172,
+      "step": 2164
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892592713364387e-05,
+      "loss": 1.2139,
+      "step": 2165
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989242795840695e-05,
+      "loss": 1.0635,
+      "step": 2166
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892263077868604e-05,
+      "loss": 1.2387,
+      "step": 2167
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9892098071751433e-05,
+      "loss": 1.0146,
+      "step": 2168
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891932940057534e-05,
+      "loss": 1.1136,
+      "step": 2169
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891767682789e-05,
+      "loss": 1.1568,
+      "step": 2170
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989160229994794e-05,
+      "loss": 1.2043,
+      "step": 2171
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891436791536442e-05,
+      "loss": 1.2443,
+      "step": 2172
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989127115755661e-05,
+      "loss": 1.2078,
+      "step": 2173
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9891105398010554e-05,
+      "loss": 1.1473,
+      "step": 2174
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890939512900367e-05,
+      "loss": 1.1948,
+      "step": 2175
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989077350222816e-05,
+      "loss": 1.1937,
+      "step": 2176
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890607365996043e-05,
+      "loss": 1.1218,
+      "step": 2177
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.989044110420612e-05,
+      "loss": 1.0119,
+      "step": 2178
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890274716860508e-05,
+      "loss": 1.1457,
+      "step": 2179
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9890108203961315e-05,
+      "loss": 1.2691,
+      "step": 2180
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988994156551065e-05,
+      "loss": 0.9953,
+      "step": 2181
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889774801510638e-05,
+      "loss": 1.163,
+      "step": 2182
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889607911963392e-05,
+      "loss": 1.1666,
+      "step": 2183
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988944089687103e-05,
+      "loss": 1.0785,
+      "step": 2184
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9889273756235673e-05,
+      "loss": 1.1879,
+      "step": 2185
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988910649005944e-05,
+      "loss": 1.0586,
+      "step": 2186
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988893909834446e-05,
+      "loss": 1.1455,
+      "step": 2187
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988877158109285e-05,
+      "loss": 1.4292,
+      "step": 2188
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888603938306747e-05,
+      "loss": 1.1487,
+      "step": 2189
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988843616998827e-05,
+      "loss": 1.1467,
+      "step": 2190
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888268276139554e-05,
+      "loss": 1.0255,
+      "step": 2191
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9888100256762727e-05,
+      "loss": 1.2715,
+      "step": 2192
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887932111859924e-05,
+      "loss": 1.0906,
+      "step": 2193
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887763841433275e-05,
+      "loss": 1.0959,
+      "step": 2194
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887595445484926e-05,
+      "loss": 1.1616,
+      "step": 2195
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887426924017007e-05,
+      "loss": 1.2232,
+      "step": 2196
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988725827703166e-05,
+      "loss": 1.1567,
+      "step": 2197
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9887089504531026e-05,
+      "loss": 1.2481,
+      "step": 2198
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886920606517243e-05,
+      "loss": 1.1098,
+      "step": 2199
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988675158299246e-05,
+      "loss": 1.2639,
+      "step": 2200
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886582433958822e-05,
+      "loss": 1.0908,
+      "step": 2201
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886413159418477e-05,
+      "loss": 1.1088,
+      "step": 2202
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886243759373572e-05,
+      "loss": 1.0924,
+      "step": 2203
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9886074233826255e-05,
+      "loss": 1.2282,
+      "step": 2204
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885904582778685e-05,
+      "loss": 1.19,
+      "step": 2205
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988573480623301e-05,
+      "loss": 1.1733,
+      "step": 2206
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885564904191388e-05,
+      "loss": 1.1024,
+      "step": 2207
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988539487665597e-05,
+      "loss": 1.0183,
+      "step": 2208
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9885224723628927e-05,
+      "loss": 1.0497,
+      "step": 2209
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988505444511241e-05,
+      "loss": 1.1405,
+      "step": 2210
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988488404110858e-05,
+      "loss": 1.1473,
+      "step": 2211
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98847135116196e-05,
+      "loss": 1.2166,
+      "step": 2212
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988454285664764e-05,
+      "loss": 1.2015,
+      "step": 2213
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9884372076194865e-05,
+      "loss": 1.2207,
+      "step": 2214
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988420117026344e-05,
+      "loss": 1.2526,
+      "step": 2215
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988403013885553e-05,
+      "loss": 1.0261,
+      "step": 2216
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883858981973322e-05,
+      "loss": 1.1752,
+      "step": 2217
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883687699618974e-05,
+      "loss": 1.1089,
+      "step": 2218
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883516291794665e-05,
+      "loss": 1.0878,
+      "step": 2219
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988334475850257e-05,
+      "loss": 1.1295,
+      "step": 2220
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988317309974487e-05,
+      "loss": 1.2142,
+      "step": 2221
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9883001315523746e-05,
+      "loss": 1.0337,
+      "step": 2222
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882829405841367e-05,
+      "loss": 1.2238,
+      "step": 2223
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882657370699928e-05,
+      "loss": 1.1952,
+      "step": 2224
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988248521010161e-05,
+      "loss": 1.1692,
+      "step": 2225
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882312924048592e-05,
+      "loss": 1.2625,
+      "step": 2226
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9882140512543065e-05,
+      "loss": 1.2322,
+      "step": 2227
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881967975587222e-05,
+      "loss": 1.078,
+      "step": 2228
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988179531318325e-05,
+      "loss": 1.0814,
+      "step": 2229
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881622525333337e-05,
+      "loss": 1.2354,
+      "step": 2230
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9881449612039682e-05,
+      "loss": 1.1179,
+      "step": 2231
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988127657330448e-05,
+      "loss": 1.155,
+      "step": 2232
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988110340912992e-05,
+      "loss": 1.2494,
+      "step": 2233
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880930119518212e-05,
+      "loss": 1.0936,
+      "step": 2234
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880756704471547e-05,
+      "loss": 1.1224,
+      "step": 2235
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988058316399213e-05,
+      "loss": 1.1397,
+      "step": 2236
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988040949808216e-05,
+      "loss": 1.2875,
+      "step": 2237
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.988023570674385e-05,
+      "loss": 1.0889,
+      "step": 2238
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9880061789979403e-05,
+      "loss": 1.0321,
+      "step": 2239
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987988774779102e-05,
+      "loss": 1.0855,
+      "step": 2240
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879713580180917e-05,
+      "loss": 1.1729,
+      "step": 2241
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879539287151303e-05,
+      "loss": 1.1688,
+      "step": 2242
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879364868704393e-05,
+      "loss": 1.054,
+      "step": 2243
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9879190324842397e-05,
+      "loss": 1.0902,
+      "step": 2244
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987901565556753e-05,
+      "loss": 1.1486,
+      "step": 2245
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878840860882017e-05,
+      "loss": 0.9823,
+      "step": 2246
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987866594078807e-05,
+      "loss": 1.0916,
+      "step": 2247
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878490895287915e-05,
+      "loss": 1.0811,
+      "step": 2248
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987831572438377e-05,
+      "loss": 1.0747,
+      "step": 2249
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9878140428077856e-05,
+      "loss": 1.2308,
+      "step": 2250
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877965006372408e-05,
+      "loss": 1.1987,
+      "step": 2251
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987778945926964e-05,
+      "loss": 1.0743,
+      "step": 2252
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987761378677179e-05,
+      "loss": 1.0394,
+      "step": 2253
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877437988881088e-05,
+      "loss": 1.2363,
+      "step": 2254
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9877262065599764e-05,
+      "loss": 1.1069,
+      "step": 2255
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987708601693005e-05,
+      "loss": 1.0768,
+      "step": 2256
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876909842874175e-05,
+      "loss": 1.1571,
+      "step": 2257
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876733543434386e-05,
+      "loss": 1.1293,
+      "step": 2258
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987655711861292e-05,
+      "loss": 1.128,
+      "step": 2259
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876380568412014e-05,
+      "loss": 1.1341,
+      "step": 2260
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9876203892833902e-05,
+      "loss": 1.0878,
+      "step": 2261
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987602709188084e-05,
+      "loss": 1.1635,
+      "step": 2262
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875850165555063e-05,
+      "loss": 1.1648,
+      "step": 2263
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987567311385882e-05,
+      "loss": 1.172,
+      "step": 2264
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987549593679436e-05,
+      "loss": 1.1092,
+      "step": 2265
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987531863436393e-05,
+      "loss": 1.2526,
+      "step": 2266
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9875141206569785e-05,
+      "loss": 1.1871,
+      "step": 2267
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987496365341417e-05,
+      "loss": 1.1598,
+      "step": 2268
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874785974899344e-05,
+      "loss": 1.2263,
+      "step": 2269
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874608171027563e-05,
+      "loss": 1.1845,
+      "step": 2270
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9874430241801082e-05,
+      "loss": 1.1252,
+      "step": 2271
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987425218722216e-05,
+      "loss": 1.0661,
+      "step": 2272
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987407400729306e-05,
+      "loss": 1.0152,
+      "step": 2273
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987389570201604e-05,
+      "loss": 1.1991,
+      "step": 2274
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873717271393363e-05,
+      "loss": 1.1459,
+      "step": 2275
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98735387154273e-05,
+      "loss": 1.0702,
+      "step": 2276
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873360034120116e-05,
+      "loss": 1.1751,
+      "step": 2277
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873181227474072e-05,
+      "loss": 1.0265,
+      "step": 2278
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9873002295491445e-05,
+      "loss": 1.14,
+      "step": 2279
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872823238174508e-05,
+      "loss": 1.1123,
+      "step": 2280
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987264405552553e-05,
+      "loss": 1.1795,
+      "step": 2281
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872464747546784e-05,
+      "loss": 1.0972,
+      "step": 2282
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987228531424055e-05,
+      "loss": 1.0953,
+      "step": 2283
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9872105755609106e-05,
+      "loss": 1.2075,
+      "step": 2284
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871926071654728e-05,
+      "loss": 1.12,
+      "step": 2285
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98717462623797e-05,
+      "loss": 1.152,
+      "step": 2286
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871566327786306e-05,
+      "loss": 1.0679,
+      "step": 2287
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871386267876825e-05,
+      "loss": 1.1384,
+      "step": 2288
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871206082653546e-05,
+      "loss": 1.1342,
+      "step": 2289
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9871025772118756e-05,
+      "loss": 1.0982,
+      "step": 2290
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870845336274746e-05,
+      "loss": 1.105,
+      "step": 2291
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98706647751238e-05,
+      "loss": 1.1567,
+      "step": 2292
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.987048408866822e-05,
+      "loss": 1.1392,
+      "step": 2293
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870303276910293e-05,
+      "loss": 1.2121,
+      "step": 2294
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9870122339852317e-05,
+      "loss": 1.0564,
+      "step": 2295
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869941277496586e-05,
+      "loss": 1.1043,
+      "step": 2296
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98697600898454e-05,
+      "loss": 1.1465,
+      "step": 2297
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986957877690106e-05,
+      "loss": 1.1433,
+      "step": 2298
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869397338665868e-05,
+      "loss": 1.1699,
+      "step": 2299
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9869215775142127e-05,
+      "loss": 1.1313,
+      "step": 2300
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986903408633214e-05,
+      "loss": 1.1541,
+      "step": 2301
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868852272238216e-05,
+      "loss": 1.2241,
+      "step": 2302
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986867033286266e-05,
+      "loss": 1.1184,
+      "step": 2303
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9868488268207787e-05,
+      "loss": 1.1036,
+      "step": 2304
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98683060782759e-05,
+      "loss": 1.1856,
+      "step": 2305
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986812376306932e-05,
+      "loss": 1.2488,
+      "step": 2306
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867941322590362e-05,
+      "loss": 1.2578,
+      "step": 2307
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986775875684133e-05,
+      "loss": 1.2588,
+      "step": 2308
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867576065824554e-05,
+      "loss": 1.2606,
+      "step": 2309
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867393249542347e-05,
+      "loss": 1.0734,
+      "step": 2310
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867210307997035e-05,
+      "loss": 1.2171,
+      "step": 2311
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9867027241190936e-05,
+      "loss": 0.9956,
+      "step": 2312
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866844049126374e-05,
+      "loss": 1.077,
+      "step": 2313
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866660731805676e-05,
+      "loss": 1.0693,
+      "step": 2314
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9866477289231168e-05,
+      "loss": 1.1804,
+      "step": 2315
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986629372140518e-05,
+      "loss": 1.1794,
+      "step": 2316
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986611002833004e-05,
+      "loss": 1.0824,
+      "step": 2317
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865926210008087e-05,
+      "loss": 1.1804,
+      "step": 2318
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865742266441644e-05,
+      "loss": 0.9906,
+      "step": 2319
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986555819763305e-05,
+      "loss": 1.2969,
+      "step": 2320
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986537400358465e-05,
+      "loss": 1.1225,
+      "step": 2321
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9865189684298772e-05,
+      "loss": 1.1708,
+      "step": 2322
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986500523977776e-05,
+      "loss": 1.029,
+      "step": 2323
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986482067002395e-05,
+      "loss": 1.1753,
+      "step": 2324
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9864635975039695e-05,
+      "loss": 1.1469,
+      "step": 2325
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986445115482733e-05,
+      "loss": 1.2455,
+      "step": 2326
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986426620938921e-05,
+      "loss": 1.2772,
+      "step": 2327
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986408113872768e-05,
+      "loss": 1.1155,
+      "step": 2328
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986389594284508e-05,
+      "loss": 1.0624,
+      "step": 2329
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863710621743774e-05,
+      "loss": 1.1742,
+      "step": 2330
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986352517542611e-05,
+      "loss": 1.104,
+      "step": 2331
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9863339603894437e-05,
+      "loss": 1.0973,
+      "step": 2332
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986315390715112e-05,
+      "loss": 1.2083,
+      "step": 2333
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986296808519851e-05,
+      "loss": 1.1389,
+      "step": 2334
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986278213803897e-05,
+      "loss": 1.007,
+      "step": 2335
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9862596065674853e-05,
+      "loss": 1.1044,
+      "step": 2336
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986240986810853e-05,
+      "loss": 0.9862,
+      "step": 2337
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986222354534236e-05,
+      "loss": 1.2109,
+      "step": 2338
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986203709737871e-05,
+      "loss": 1.1694,
+      "step": 2339
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861850524219947e-05,
+      "loss": 1.1264,
+      "step": 2340
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861663825868435e-05,
+      "loss": 1.1036,
+      "step": 2341
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986147700232655e-05,
+      "loss": 1.1345,
+      "step": 2342
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861290053596662e-05,
+      "loss": 1.1549,
+      "step": 2343
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9861102979681143e-05,
+      "loss": 1.1376,
+      "step": 2344
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986091578058237e-05,
+      "loss": 1.2609,
+      "step": 2345
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860728456302714e-05,
+      "loss": 0.9975,
+      "step": 2346
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986054100684456e-05,
+      "loss": 1.2302,
+      "step": 2347
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9860353432210283e-05,
+      "loss": 1.1618,
+      "step": 2348
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.986016573240227e-05,
+      "loss": 1.1656,
+      "step": 2349
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859977907422893e-05,
+      "loss": 1.1951,
+      "step": 2350
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859789957274545e-05,
+      "loss": 1.1551,
+      "step": 2351
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859601881959614e-05,
+      "loss": 1.0686,
+      "step": 2352
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985941368148048e-05,
+      "loss": 1.1471,
+      "step": 2353
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859225355839537e-05,
+      "loss": 1.0707,
+      "step": 2354
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9859036905039172e-05,
+      "loss": 1.0459,
+      "step": 2355
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985884832908178e-05,
+      "loss": 1.1464,
+      "step": 2356
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858659627969757e-05,
+      "loss": 1.2489,
+      "step": 2357
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858470801705495e-05,
+      "loss": 1.1185,
+      "step": 2358
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985828185029139e-05,
+      "loss": 1.1187,
+      "step": 2359
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9858092773729846e-05,
+      "loss": 1.1454,
+      "step": 2360
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857903572023257e-05,
+      "loss": 1.1351,
+      "step": 2361
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985771424517403e-05,
+      "loss": 1.1068,
+      "step": 2362
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857524793184562e-05,
+      "loss": 1.1789,
+      "step": 2363
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857335216057264e-05,
+      "loss": 0.9749,
+      "step": 2364
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9857145513794546e-05,
+      "loss": 1.1278,
+      "step": 2365
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856955686398807e-05,
+      "loss": 1.0653,
+      "step": 2366
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985676573387246e-05,
+      "loss": 1.1109,
+      "step": 2367
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985657565621792e-05,
+      "loss": 1.1245,
+      "step": 2368
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856385453437598e-05,
+      "loss": 1.1768,
+      "step": 2369
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9856195125533902e-05,
+      "loss": 1.0459,
+      "step": 2370
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985600467250926e-05,
+      "loss": 1.2146,
+      "step": 2371
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855814094366083e-05,
+      "loss": 1.1808,
+      "step": 2372
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855623391106787e-05,
+      "loss": 1.2231,
+      "step": 2373
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98554325627338e-05,
+      "loss": 1.231,
+      "step": 2374
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855241609249543e-05,
+      "loss": 1.1036,
+      "step": 2375
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9855050530656433e-05,
+      "loss": 1.1347,
+      "step": 2376
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985485932695691e-05,
+      "loss": 1.1329,
+      "step": 2377
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854667998153384e-05,
+      "loss": 1.2117,
+      "step": 2378
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854476544248297e-05,
+      "loss": 1.1603,
+      "step": 2379
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854284965244073e-05,
+      "loss": 1.0064,
+      "step": 2380
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9854093261143145e-05,
+      "loss": 1.2564,
+      "step": 2381
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985390143194795e-05,
+      "loss": 1.2025,
+      "step": 2382
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853709477660918e-05,
+      "loss": 1.1341,
+      "step": 2383
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985351739828449e-05,
+      "loss": 1.1457,
+      "step": 2384
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98533251938211e-05,
+      "loss": 1.1632,
+      "step": 2385
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9853132864273194e-05,
+      "loss": 1.0344,
+      "step": 2386
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985294040964321e-05,
+      "loss": 1.0842,
+      "step": 2387
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985274782993359e-05,
+      "loss": 1.0712,
+      "step": 2388
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852555125146784e-05,
+      "loss": 1.25,
+      "step": 2389
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985236229528523e-05,
+      "loss": 1.1609,
+      "step": 2390
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9852169340351382e-05,
+      "loss": 1.0599,
+      "step": 2391
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985197626034769e-05,
+      "loss": 1.1669,
+      "step": 2392
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98517830552766e-05,
+      "loss": 1.0848,
+      "step": 2393
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851589725140567e-05,
+      "loss": 1.2073,
+      "step": 2394
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9851396269942048e-05,
+      "loss": 1.1634,
+      "step": 2395
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98512026896835e-05,
+      "loss": 1.259,
+      "step": 2396
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985100898436737e-05,
+      "loss": 1.2605,
+      "step": 2397
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850815153996127e-05,
+      "loss": 1.0862,
+      "step": 2398
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985062119857223e-05,
+      "loss": 1.1223,
+      "step": 2399
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.985042711809814e-05,
+      "loss": 1.2562,
+      "step": 2400
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850232912576322e-05,
+      "loss": 1.1331,
+      "step": 2401
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9850038582009235e-05,
+      "loss": 1.2008,
+      "step": 2402
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849844126399357e-05,
+      "loss": 1.1384,
+      "step": 2403
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984964954574915e-05,
+      "loss": 1.1144,
+      "step": 2404
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9849454840061085e-05,
+      "loss": 1.2003,
+      "step": 2405
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984926000933763e-05,
+      "loss": 1.1042,
+      "step": 2406
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984906505358126e-05,
+      "loss": 1.0039,
+      "step": 2407
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848869972794457e-05,
+      "loss": 1.0205,
+      "step": 2408
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984867476697969e-05,
+      "loss": 1.0964,
+      "step": 2409
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848479436139444e-05,
+      "loss": 1.1083,
+      "step": 2410
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9848283980276188e-05,
+      "loss": 1.1459,
+      "step": 2411
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984808839939241e-05,
+      "loss": 1.0372,
+      "step": 2412
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847892693490592e-05,
+      "loss": 1.1116,
+      "step": 2413
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847696862573222e-05,
+      "loss": 1.0897,
+      "step": 2414
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847500906642782e-05,
+      "loss": 1.171,
+      "step": 2415
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9847304825701757e-05,
+      "loss": 1.0681,
+      "step": 2416
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984710861975264e-05,
+      "loss": 1.1736,
+      "step": 2417
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984691228879792e-05,
+      "loss": 1.1592,
+      "step": 2418
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846715832840092e-05,
+      "loss": 1.2159,
+      "step": 2419
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846519251881648e-05,
+      "loss": 1.1388,
+      "step": 2420
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9846322545925084e-05,
+      "loss": 1.2415,
+      "step": 2421
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.98461257149729e-05,
+      "loss": 1.2451,
+      "step": 2422
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845928759027585e-05,
+      "loss": 1.1696,
+      "step": 2423
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984573167809165e-05,
+      "loss": 1.147,
+      "step": 2424
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845534472167594e-05,
+      "loss": 1.0929,
+      "step": 2425
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845337141257915e-05,
+      "loss": 1.1508,
+      "step": 2426
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9845139685365126e-05,
+      "loss": 1.1666,
+      "step": 2427
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844942104491728e-05,
+      "loss": 1.2489,
+      "step": 2428
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984474439864023e-05,
+      "loss": 1.0827,
+      "step": 2429
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844546567813146e-05,
+      "loss": 0.9224,
+      "step": 2430
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844348612012984e-05,
+      "loss": 0.9821,
+      "step": 2431
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9844150531242258e-05,
+      "loss": 1.1636,
+      "step": 2432
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984395232550348e-05,
+      "loss": 1.1478,
+      "step": 2433
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984375399479917e-05,
+      "loss": 1.1942,
+      "step": 2434
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984355553913184e-05,
+      "loss": 1.1581,
+      "step": 2435
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843356958504018e-05,
+      "loss": 1.1513,
+      "step": 2436
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9843158252918212e-05,
+      "loss": 1.2013,
+      "step": 2437
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984295942237696e-05,
+      "loss": 1.1917,
+      "step": 2438
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842760466882778e-05,
+      "loss": 1.0142,
+      "step": 2439
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.984256138643819e-05,
+      "loss": 1.1514,
+      "step": 2440
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842362181045724e-05,
+      "loss": 1.1624,
+      "step": 2441
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9842162850707915e-05,
+      "loss": 1.014,
+      "step": 2442
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 1.9841963395427285e-05,
+      "loss": 1.2028,
+      "step": 2443
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984176381520637e-05,
+      "loss": 1.0939,
+      "step": 2444
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841564110047708e-05,
+      "loss": 1.184,
+      "step": 2445
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841364279953826e-05,
+      "loss": 1.0174,
+      "step": 2446
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9841164324927264e-05,
+      "loss": 1.1332,
+      "step": 2447
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840964244970558e-05,
+      "loss": 1.1844,
+      "step": 2448
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840764040086257e-05,
+      "loss": 1.2308,
+      "step": 2449
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840563710276896e-05,
+      "loss": 1.1242,
+      "step": 2450
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9840363255545015e-05,
+      "loss": 1.073,
+      "step": 2451
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.984016267589316e-05,
+      "loss": 1.0779,
+      "step": 2452
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983996197132388e-05,
+      "loss": 1.1877,
+      "step": 2453
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839761141839725e-05,
+      "loss": 1.1475,
+      "step": 2454
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983956018744324e-05,
+      "loss": 1.1536,
+      "step": 2455
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839359108136983e-05,
+      "loss": 1.0992,
+      "step": 2456
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9839157903923495e-05,
+      "loss": 1.0937,
+      "step": 2457
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838956574805337e-05,
+      "loss": 1.0944,
+      "step": 2458
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838755120785066e-05,
+      "loss": 1.1858,
+      "step": 2459
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983855354186524e-05,
+      "loss": 1.0652,
+      "step": 2460
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838351838048414e-05,
+      "loss": 1.0264,
+      "step": 2461
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9838150009337148e-05,
+      "loss": 1.2157,
+      "step": 2462
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837948055734005e-05,
+      "loss": 1.2344,
+      "step": 2463
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837745977241557e-05,
+      "loss": 1.1254,
+      "step": 2464
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837543773862357e-05,
+      "loss": 1.165,
+      "step": 2465
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9837341445598978e-05,
+      "loss": 1.1916,
+      "step": 2466
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983713899245399e-05,
+      "loss": 1.019,
+      "step": 2467
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983693641442996e-05,
+      "loss": 0.9706,
+      "step": 2468
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836733711529458e-05,
+      "loss": 1.1343,
+      "step": 2469
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836530883755066e-05,
+      "loss": 1.2612,
+      "step": 2470
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836327931109345e-05,
+      "loss": 1.1168,
+      "step": 2471
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9836124853594885e-05,
+      "loss": 1.2109,
+      "step": 2472
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835921651214253e-05,
+      "loss": 1.3125,
+      "step": 2473
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835718323970038e-05,
+      "loss": 1.146,
+      "step": 2474
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835514871864818e-05,
+      "loss": 1.0966,
+      "step": 2475
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983531129490117e-05,
+      "loss": 1.0413,
+      "step": 2476
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9835107593081684e-05,
+      "loss": 1.1525,
+      "step": 2477
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983490376640895e-05,
+      "loss": 1.1809,
+      "step": 2478
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834699814885542e-05,
+      "loss": 1.1144,
+      "step": 2479
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834495738514065e-05,
+      "loss": 1.1317,
+      "step": 2480
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834291537297098e-05,
+      "loss": 1.0508,
+      "step": 2481
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9834087211237235e-05,
+      "loss": 1.0817,
+      "step": 2482
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833882760337076e-05,
+      "loss": 1.0012,
+      "step": 2483
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9833678184599212e-05,
+      "loss": 1.139,
+      "step": 2484
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983347348402624e-05,
+      "loss": 1.2069,
+      "step": 2485
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983326865862076e-05,
+      "loss": 1.1423,
+      "step": 2486
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983306370838537e-05,
+      "loss": 1.134,
+      "step": 2487
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832858633322675e-05,
+      "loss": 1.1311,
+      "step": 2488
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983265343343528e-05,
+      "loss": 1.009,
+      "step": 2489
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983244810872578e-05,
+      "loss": 1.0689,
+      "step": 2490
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832242659196792e-05,
+      "loss": 1.0521,
+      "step": 2491
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9832037084850916e-05,
+      "loss": 1.0806,
+      "step": 2492
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983183138569077e-05,
+      "loss": 0.9884,
+      "step": 2493
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983162556171896e-05,
+      "loss": 1.2514,
+      "step": 2494
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98314196129381e-05,
+      "loss": 1.0205,
+      "step": 2495
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9831213539350804e-05,
+      "loss": 1.1281,
+      "step": 2496
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983100734095969e-05,
+      "loss": 1.2459,
+      "step": 2497
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830801017767373e-05,
+      "loss": 1.1778,
+      "step": 2498
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9830594569776475e-05,
+      "loss": 1.1347,
+      "step": 2499
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983038799698961e-05,
+      "loss": 1.1741,
+      "step": 2500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.983018129940941e-05,
+      "loss": 1.106,
+      "step": 2501
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829974477038494e-05,
+      "loss": 1.2599,
+      "step": 2502
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829767529879486e-05,
+      "loss": 1.1494,
+      "step": 2503
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982956045793502e-05,
+      "loss": 1.1033,
+      "step": 2504
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829353261207714e-05,
+      "loss": 1.1714,
+      "step": 2505
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9829145939700204e-05,
+      "loss": 1.1123,
+      "step": 2506
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828938493415123e-05,
+      "loss": 1.3136,
+      "step": 2507
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828730922355106e-05,
+      "loss": 1.2582,
+      "step": 2508
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982852322652278e-05,
+      "loss": 1.1757,
+      "step": 2509
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828315405920796e-05,
+      "loss": 1.2485,
+      "step": 2510
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9828107460551776e-05,
+      "loss": 0.9746,
+      "step": 2511
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982789939041837e-05,
+      "loss": 1.1161,
+      "step": 2512
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827691195523216e-05,
+      "loss": 1.1174,
+      "step": 2513
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827482875868958e-05,
+      "loss": 1.1253,
+      "step": 2514
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982727443145824e-05,
+      "loss": 1.1889,
+      "step": 2515
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9827065862293706e-05,
+      "loss": 1.0824,
+      "step": 2516
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826857168378007e-05,
+      "loss": 1.0705,
+      "step": 2517
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982664834971379e-05,
+      "loss": 1.1863,
+      "step": 2518
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826439406303708e-05,
+      "loss": 1.0397,
+      "step": 2519
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9826230338150415e-05,
+      "loss": 1.2201,
+      "step": 2520
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982602114525656e-05,
+      "loss": 1.0703,
+      "step": 2521
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825811827624796e-05,
+      "loss": 1.0665,
+      "step": 2522
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825602385257792e-05,
+      "loss": 1.1168,
+      "step": 2523
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9825392818158197e-05,
+      "loss": 1.0794,
+      "step": 2524
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982518312632868e-05,
+      "loss": 1.0723,
+      "step": 2525
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824973309771892e-05,
+      "loss": 1.1438,
+      "step": 2526
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824763368490502e-05,
+      "loss": 1.1908,
+      "step": 2527
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824553302487176e-05,
+      "loss": 1.335,
+      "step": 2528
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982434311176458e-05,
+      "loss": 1.1187,
+      "step": 2529
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9824132796325384e-05,
+      "loss": 1.1109,
+      "step": 2530
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823922356172257e-05,
+      "loss": 1.0154,
+      "step": 2531
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823711791307864e-05,
+      "loss": 1.1312,
+      "step": 2532
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982350110173489e-05,
+      "loss": 1.0644,
+      "step": 2533
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823290287456e-05,
+      "loss": 1.1686,
+      "step": 2534
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9823079348473873e-05,
+      "loss": 1.0679,
+      "step": 2535
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822868284791185e-05,
+      "loss": 1.1632,
+      "step": 2536
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982265709641062e-05,
+      "loss": 1.1045,
+      "step": 2537
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822445783334858e-05,
+      "loss": 1.1123,
+      "step": 2538
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822234345566577e-05,
+      "loss": 1.2098,
+      "step": 2539
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9822022783108466e-05,
+      "loss": 1.1673,
+      "step": 2540
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821811095963213e-05,
+      "loss": 1.1536,
+      "step": 2541
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821599284133496e-05,
+      "loss": 1.0727,
+      "step": 2542
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982138734762201e-05,
+      "loss": 1.1631,
+      "step": 2543
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9821175286431446e-05,
+      "loss": 1.1067,
+      "step": 2544
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.982096310056449e-05,
+      "loss": 1.1098,
+      "step": 2545
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820750790023844e-05,
+      "loss": 1.1886,
+      "step": 2546
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98205383548122e-05,
+      "loss": 1.1048,
+      "step": 2547
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9820325794932255e-05,
+      "loss": 1.098,
+      "step": 2548
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98201131103867e-05,
+      "loss": 1.2104,
+      "step": 2549
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981990030117825e-05,
+      "loss": 1.0937,
+      "step": 2550
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819687367309594e-05,
+      "loss": 1.0435,
+      "step": 2551
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981947430878344e-05,
+      "loss": 1.1155,
+      "step": 2552
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819261125602492e-05,
+      "loss": 1.1436,
+      "step": 2553
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9819047817769456e-05,
+      "loss": 1.0842,
+      "step": 2554
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981883438528704e-05,
+      "loss": 1.0706,
+      "step": 2555
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981862082815795e-05,
+      "loss": 1.0467,
+      "step": 2556
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818407146384907e-05,
+      "loss": 1.1781,
+      "step": 2557
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9818193339970612e-05,
+      "loss": 1.1576,
+      "step": 2558
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817979408917788e-05,
+      "loss": 1.0988,
+      "step": 2559
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817765353229147e-05,
+      "loss": 1.0628,
+      "step": 2560
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9817551172907404e-05,
+      "loss": 1.2691,
+      "step": 2561
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981733686795528e-05,
+      "loss": 1.2151,
+      "step": 2562
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98171224383755e-05,
+      "loss": 0.9843,
+      "step": 2563
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981690788417078e-05,
+      "loss": 1.1229,
+      "step": 2564
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816693205343843e-05,
+      "loss": 1.1864,
+      "step": 2565
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981647840189742e-05,
+      "loss": 1.2607,
+      "step": 2566
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816263473834233e-05,
+      "loss": 1.0465,
+      "step": 2567
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9816048421157015e-05,
+      "loss": 1.0285,
+      "step": 2568
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981583324386849e-05,
+      "loss": 1.0611,
+      "step": 2569
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815617941971395e-05,
+      "loss": 1.1812,
+      "step": 2570
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815402515468465e-05,
+      "loss": 1.244,
+      "step": 2571
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9815186964362424e-05,
+      "loss": 1.1774,
+      "step": 2572
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981497128865602e-05,
+      "loss": 1.1228,
+      "step": 2573
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814755488351983e-05,
+      "loss": 1.105,
+      "step": 2574
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981453956345306e-05,
+      "loss": 1.1965,
+      "step": 2575
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814323513961985e-05,
+      "loss": 1.028,
+      "step": 2576
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9814107339881504e-05,
+      "loss": 1.1273,
+      "step": 2577
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813891041214355e-05,
+      "loss": 1.2004,
+      "step": 2578
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98136746179633e-05,
+      "loss": 1.0847,
+      "step": 2579
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981345807013107e-05,
+      "loss": 1.2258,
+      "step": 2580
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9813241397720418e-05,
+      "loss": 1.254,
+      "step": 2581
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98130246007341e-05,
+      "loss": 1.0338,
+      "step": 2582
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812807679174863e-05,
+      "loss": 1.1104,
+      "step": 2583
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812590633045462e-05,
+      "loss": 1.2,
+      "step": 2584
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9812373462348655e-05,
+      "loss": 1.1935,
+      "step": 2585
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.98121561670872e-05,
+      "loss": 1.1703,
+      "step": 2586
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811938747263847e-05,
+      "loss": 1.1051,
+      "step": 2587
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981172120288136e-05,
+      "loss": 1.1705,
+      "step": 2588
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811503533942505e-05,
+      "loss": 1.0265,
+      "step": 2589
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811285740450043e-05,
+      "loss": 1.0978,
+      "step": 2590
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9811067822406732e-05,
+      "loss": 1.1818,
+      "step": 2591
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.981084977981535e-05,
+      "loss": 1.1126,
+      "step": 2592
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810631612678658e-05,
+      "loss": 1.0387,
+      "step": 2593
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810413320999427e-05,
+      "loss": 1.1867,
+      "step": 2594
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9810194904780427e-05,
+      "loss": 1.2178,
+      "step": 2595
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809976364024433e-05,
+      "loss": 1.1321,
+      "step": 2596
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809757698734216e-05,
+      "loss": 0.9393,
+      "step": 2597
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809538908912554e-05,
+      "loss": 1.0443,
+      "step": 2598
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9809319994562226e-05,
+      "loss": 1.0461,
+      "step": 2599
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980910095568601e-05,
+      "loss": 1.0633,
+      "step": 2600
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980888179228668e-05,
+      "loss": 1.145,
+      "step": 2601
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9808662504367034e-05,
+      "loss": 1.1608,
+      "step": 2602
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980844309192984e-05,
+      "loss": 1.1948,
+      "step": 2603
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980822355497789e-05,
+      "loss": 1.1734,
+      "step": 2604
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980800389351397e-05,
+      "loss": 1.1396,
+      "step": 2605
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980778410754087e-05,
+      "loss": 0.9994,
+      "step": 2606
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980756419706138e-05,
+      "loss": 1.2463,
+      "step": 2607
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807344162078287e-05,
+      "loss": 1.1364,
+      "step": 2608
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9807124002594392e-05,
+      "loss": 1.1438,
+      "step": 2609
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806903718612483e-05,
+      "loss": 1.2029,
+      "step": 2610
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980668331013536e-05,
+      "loss": 1.0803,
+      "step": 2611
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980646277716582e-05,
+      "loss": 1.2059,
+      "step": 2612
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9806242119706662e-05,
+      "loss": 1.1585,
+      "step": 2613
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980602133776069e-05,
+      "loss": 1.2191,
+      "step": 2614
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805800431330704e-05,
+      "loss": 1.095,
+      "step": 2615
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980557940041951e-05,
+      "loss": 1.0231,
+      "step": 2616
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805358245029908e-05,
+      "loss": 1.1721,
+      "step": 2617
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9805136965164715e-05,
+      "loss": 1.06,
+      "step": 2618
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804915560826733e-05,
+      "loss": 1.1458,
+      "step": 2619
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804694032018778e-05,
+      "loss": 1.159,
+      "step": 2620
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804472378743657e-05,
+      "loss": 1.1576,
+      "step": 2621
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804250601004187e-05,
+      "loss": 1.1368,
+      "step": 2622
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9804028698803183e-05,
+      "loss": 1.1784,
+      "step": 2623
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980380667214346e-05,
+      "loss": 1.1179,
+      "step": 2624
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803584521027842e-05,
+      "loss": 1.1572,
+      "step": 2625
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980336224545914e-05,
+      "loss": 1.0319,
+      "step": 2626
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9803139845440183e-05,
+      "loss": 1.1301,
+      "step": 2627
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980291732097379e-05,
+      "loss": 1.062,
+      "step": 2628
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802694672062794e-05,
+      "loss": 1.2093,
+      "step": 2629
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980247189871001e-05,
+      "loss": 1.0428,
+      "step": 2630
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802249000918274e-05,
+      "loss": 1.1162,
+      "step": 2631
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9802025978690412e-05,
+      "loss": 1.0339,
+      "step": 2632
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980180283202926e-05,
+      "loss": 1.1662,
+      "step": 2633
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980157956093764e-05,
+      "loss": 1.2056,
+      "step": 2634
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9801356165418402e-05,
+      "loss": 1.0466,
+      "step": 2635
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980113264547437e-05,
+      "loss": 1.1633,
+      "step": 2636
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980090900110838e-05,
+      "loss": 1.1989,
+      "step": 2637
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800685232323283e-05,
+      "loss": 1.3151,
+      "step": 2638
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.980046133912191e-05,
+      "loss": 1.0477,
+      "step": 2639
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800237321507105e-05,
+      "loss": 1.1432,
+      "step": 2640
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9800013179481713e-05,
+      "loss": 1.0897,
+      "step": 2641
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979978891304858e-05,
+      "loss": 1.2054,
+      "step": 2642
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979956452221055e-05,
+      "loss": 1.2479,
+      "step": 2643
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9799340006970476e-05,
+      "loss": 1.1315,
+      "step": 2644
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979911536733121e-05,
+      "loss": 1.0717,
+      "step": 2645
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798890603295592e-05,
+      "loss": 1.0838,
+      "step": 2646
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9798665714866485e-05,
+      "loss": 1.0085,
+      "step": 2647
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979844070204674e-05,
+      "loss": 1.048,
+      "step": 2648
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979821556483922e-05,
+      "loss": 1.0415,
+      "step": 2649
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797990303246776e-05,
+      "loss": 1.0098,
+      "step": 2650
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797764917272267e-05,
+      "loss": 1.1964,
+      "step": 2651
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797539406918563e-05,
+      "loss": 1.0547,
+      "step": 2652
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797313772188516e-05,
+      "loss": 1.173,
+      "step": 2653
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9797088013084997e-05,
+      "loss": 1.3349,
+      "step": 2654
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796862129610874e-05,
+      "loss": 0.9971,
+      "step": 2655
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796636121769004e-05,
+      "loss": 1.1722,
+      "step": 2656
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796409989562263e-05,
+      "loss": 1.3686,
+      "step": 2657
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9796183732993522e-05,
+      "loss": 1.0555,
+      "step": 2658
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795957352065657e-05,
+      "loss": 1.209,
+      "step": 2659
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795730846781534e-05,
+      "loss": 1.1937,
+      "step": 2660
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795504217144033e-05,
+      "loss": 1.1042,
+      "step": 2661
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795277463156027e-05,
+      "loss": 1.0829,
+      "step": 2662
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9795050584820403e-05,
+      "loss": 1.1355,
+      "step": 2663
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794823582140034e-05,
+      "loss": 1.1084,
+      "step": 2664
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97945964551178e-05,
+      "loss": 0.9951,
+      "step": 2665
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979436920375659e-05,
+      "loss": 1.218,
+      "step": 2666
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9794141828059287e-05,
+      "loss": 1.1815,
+      "step": 2667
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793914328028774e-05,
+      "loss": 1.1385,
+      "step": 2668
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793686703667944e-05,
+      "loss": 1.1338,
+      "step": 2669
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793458954979685e-05,
+      "loss": 1.1771,
+      "step": 2670
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793231081966887e-05,
+      "loss": 1.0673,
+      "step": 2671
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9793003084632442e-05,
+      "loss": 1.1378,
+      "step": 2672
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792774962979246e-05,
+      "loss": 0.9838,
+      "step": 2673
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792546717010195e-05,
+      "loss": 1.0817,
+      "step": 2674
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979231834672819e-05,
+      "loss": 1.1646,
+      "step": 2675
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9792089852136124e-05,
+      "loss": 1.1874,
+      "step": 2676
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791861233236898e-05,
+      "loss": 1.215,
+      "step": 2677
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791632490033418e-05,
+      "loss": 1.0875,
+      "step": 2678
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791403622528586e-05,
+      "loss": 1.1074,
+      "step": 2679
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9791174630725307e-05,
+      "loss": 1.1658,
+      "step": 2680
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979094551462649e-05,
+      "loss": 1.0593,
+      "step": 2681
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979071627423504e-05,
+      "loss": 1.1904,
+      "step": 2682
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979048690955387e-05,
+      "loss": 1.2262,
+      "step": 2683
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.979025742058589e-05,
+      "loss": 1.1269,
+      "step": 2684
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9790027807334016e-05,
+      "loss": 1.1034,
+      "step": 2685
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978979806980116e-05,
+      "loss": 1.083,
+      "step": 2686
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789568207990238e-05,
+      "loss": 1.0158,
+      "step": 2687
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978933822190417e-05,
+      "loss": 1.1437,
+      "step": 2688
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9789108111545876e-05,
+      "loss": 1.2464,
+      "step": 2689
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978887787691828e-05,
+      "loss": 1.1263,
+      "step": 2690
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788647518024294e-05,
+      "loss": 1.0654,
+      "step": 2691
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788417034866853e-05,
+      "loss": 1.0947,
+      "step": 2692
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9788186427448878e-05,
+      "loss": 1.1404,
+      "step": 2693
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787955695773295e-05,
+      "loss": 0.9916,
+      "step": 2694
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978772483984304e-05,
+      "loss": 1.0962,
+      "step": 2695
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787493859661037e-05,
+      "loss": 1.1334,
+      "step": 2696
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978726275523022e-05,
+      "loss": 1.1629,
+      "step": 2697
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9787031526553526e-05,
+      "loss": 1.217,
+      "step": 2698
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786800173633885e-05,
+      "loss": 1.089,
+      "step": 2699
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978656869647424e-05,
+      "loss": 1.1826,
+      "step": 2700
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786337095077522e-05,
+      "loss": 1.0893,
+      "step": 2701
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9786105369446678e-05,
+      "loss": 0.974,
+      "step": 2702
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978587351958465e-05,
+      "loss": 1.1054,
+      "step": 2703
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9785641545494373e-05,
+      "loss": 1.0196,
+      "step": 2704
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97854094471788e-05,
+      "loss": 1.1868,
+      "step": 2705
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978517722464087e-05,
+      "loss": 1.2326,
+      "step": 2706
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978494487788354e-05,
+      "loss": 1.186,
+      "step": 2707
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784712406909752e-05,
+      "loss": 1.0675,
+      "step": 2708
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978447981172246e-05,
+      "loss": 1.1025,
+      "step": 2709
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9784247092324617e-05,
+      "loss": 1.0426,
+      "step": 2710
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978401424871918e-05,
+      "loss": 1.0899,
+      "step": 2711
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9783781280909102e-05,
+      "loss": 1.0786,
+      "step": 2712
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978354818889734e-05,
+      "loss": 1.1014,
+      "step": 2713
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978331497268685e-05,
+      "loss": 1.0581,
+      "step": 2714
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97830816322806e-05,
+      "loss": 1.1506,
+      "step": 2715
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782848167681547e-05,
+      "loss": 1.1354,
+      "step": 2716
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782614578892654e-05,
+      "loss": 1.0113,
+      "step": 2717
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9782380865916887e-05,
+      "loss": 1.1468,
+      "step": 2718
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978214702875722e-05,
+      "loss": 1.0512,
+      "step": 2719
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781913067416613e-05,
+      "loss": 1.1375,
+      "step": 2720
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781678981898035e-05,
+      "loss": 1.2035,
+      "step": 2721
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781444772204468e-05,
+      "loss": 1.1329,
+      "step": 2722
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9781210438338876e-05,
+      "loss": 1.2173,
+      "step": 2723
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780975980304234e-05,
+      "loss": 1.1485,
+      "step": 2724
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.978074139810352e-05,
+      "loss": 1.013,
+      "step": 2725
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780506691739717e-05,
+      "loss": 1.0811,
+      "step": 2726
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.97802718612158e-05,
+      "loss": 1.0173,
+      "step": 2727
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9780036906534747e-05,
+      "loss": 1.1534,
+      "step": 2728
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.9779801827699545e-05,
+      "loss": 1.0303,
+      "step": 2729
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 1.977956662471318e-05,
+      "loss": 1.1047,
+      "step": 2730
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779331297578633e-05,
+      "loss": 1.1507,
+      "step": 2731
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9779095846298893e-05,
+      "loss": 1.0366,
+      "step": 2732
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778860270876947e-05,
+      "loss": 1.2397,
+      "step": 2733
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977862457131579e-05,
+      "loss": 1.0676,
+      "step": 2734
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9778388747618416e-05,
+      "loss": 1.1292,
+      "step": 2735
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977815279978781e-05,
+      "loss": 1.1698,
+      "step": 2736
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777916727826975e-05,
+      "loss": 1.2944,
+      "step": 2737
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777680531738903e-05,
+      "loss": 0.9988,
+      "step": 2738
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777444211526592e-05,
+      "loss": 1.2283,
+      "step": 2739
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9777207767193048e-05,
+      "loss": 1.1767,
+      "step": 2740
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977697119874127e-05,
+      "loss": 1.1837,
+      "step": 2741
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776734506174254e-05,
+      "loss": 1.0116,
+      "step": 2742
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776497689495015e-05,
+      "loss": 1.0295,
+      "step": 2743
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9776260748706554e-05,
+      "loss": 1.2516,
+      "step": 2744
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977602368381188e-05,
+      "loss": 1.0129,
+      "step": 2745
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775786494814008e-05,
+      "loss": 1.236,
+      "step": 2746
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775549181715934e-05,
+      "loss": 1.1548,
+      "step": 2747
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9775311744520687e-05,
+      "loss": 0.9713,
+      "step": 2748
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977507418323127e-05,
+      "loss": 1.0565,
+      "step": 2749
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774836497850707e-05,
+      "loss": 1.2189,
+      "step": 2750
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774598688382006e-05,
+      "loss": 1.1553,
+      "step": 2751
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9774360754828195e-05,
+      "loss": 1.2004,
+      "step": 2752
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977412269719229e-05,
+      "loss": 1.2175,
+      "step": 2753
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773884515477313e-05,
+      "loss": 1.2178,
+      "step": 2754
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977364620968629e-05,
+      "loss": 1.1236,
+      "step": 2755
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9773407779822243e-05,
+      "loss": 1.0653,
+      "step": 2756
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97731692258882e-05,
+      "loss": 1.1894,
+      "step": 2757
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772930547887196e-05,
+      "loss": 1.1193,
+      "step": 2758
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772691745822244e-05,
+      "loss": 1.0892,
+      "step": 2759
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9772452819696395e-05,
+      "loss": 1.1921,
+      "step": 2760
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977221376951267e-05,
+      "loss": 1.2952,
+      "step": 2761
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771974595274107e-05,
+      "loss": 1.0364,
+      "step": 2762
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771735296983742e-05,
+      "loss": 1.2796,
+      "step": 2763
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771495874644614e-05,
+      "loss": 1.29,
+      "step": 2764
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977125632825976e-05,
+      "loss": 1.0901,
+      "step": 2765
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9771016657832222e-05,
+      "loss": 1.1228,
+      "step": 2766
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770776863365045e-05,
+      "loss": 1.1618,
+      "step": 2767
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977053694486127e-05,
+      "loss": 1.0155,
+      "step": 2768
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9770296902323943e-05,
+      "loss": 1.0518,
+      "step": 2769
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.977005673575611e-05,
+      "loss": 1.1009,
+      "step": 2770
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769816445160824e-05,
+      "loss": 0.967,
+      "step": 2771
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769576030541134e-05,
+      "loss": 1.1426,
+      "step": 2772
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976933549190009e-05,
+      "loss": 1.0765,
+      "step": 2773
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9769094829240746e-05,
+      "loss": 1.231,
+      "step": 2774
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976885404256616e-05,
+      "loss": 1.2333,
+      "step": 2775
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768613131879386e-05,
+      "loss": 1.1419,
+      "step": 2776
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976837209718348e-05,
+      "loss": 1.1539,
+      "step": 2777
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9768130938481508e-05,
+      "loss": 1.2372,
+      "step": 2778
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767889655776527e-05,
+      "loss": 0.9911,
+      "step": 2779
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97676482490716e-05,
+      "loss": 1.2064,
+      "step": 2780
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9767406718369794e-05,
+      "loss": 1.0348,
+      "step": 2781
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976716506367417e-05,
+      "loss": 1.2197,
+      "step": 2782
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766923284987808e-05,
+      "loss": 1.2529,
+      "step": 2783
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976668138231376e-05,
+      "loss": 1.1634,
+      "step": 2784
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976643935565511e-05,
+      "loss": 1.1796,
+      "step": 2785
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9766197205014925e-05,
+      "loss": 1.0708,
+      "step": 2786
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976595493039628e-05,
+      "loss": 1.1844,
+      "step": 2787
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765712531802252e-05,
+      "loss": 1.1404,
+      "step": 2788
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9765470009235915e-05,
+      "loss": 1.219,
+      "step": 2789
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976522736270035e-05,
+      "loss": 1.171,
+      "step": 2790
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976498459219864e-05,
+      "loss": 1.2057,
+      "step": 2791
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976474169773386e-05,
+      "loss": 1.0546,
+      "step": 2792
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764498679309095e-05,
+      "loss": 1.0231,
+      "step": 2793
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764255536927436e-05,
+      "loss": 0.9537,
+      "step": 2794
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9764012270591968e-05,
+      "loss": 1.2102,
+      "step": 2795
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763768880305775e-05,
+      "loss": 1.1857,
+      "step": 2796
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763525366071946e-05,
+      "loss": 1.1284,
+      "step": 2797
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976328172789358e-05,
+      "loss": 1.3155,
+      "step": 2798
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9763037965773764e-05,
+      "loss": 1.076,
+      "step": 2799
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762794079715592e-05,
+      "loss": 1.0584,
+      "step": 2800
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9762550069722165e-05,
+      "loss": 1.0877,
+      "step": 2801
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976230593579657e-05,
+      "loss": 1.2201,
+      "step": 2802
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976206167794192e-05,
+      "loss": 0.9579,
+      "step": 2803
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761817296161307e-05,
+      "loss": 1.0163,
+      "step": 2804
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761572790457835e-05,
+      "loss": 1.1117,
+      "step": 2805
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761328160834606e-05,
+      "loss": 1.0246,
+      "step": 2806
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9761083407294735e-05,
+      "loss": 1.2495,
+      "step": 2807
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760838529841317e-05,
+      "loss": 1.1408,
+      "step": 2808
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760593528477465e-05,
+      "loss": 1.0876,
+      "step": 2809
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9760348403206293e-05,
+      "loss": 1.1054,
+      "step": 2810
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.976010315403091e-05,
+      "loss": 1.2473,
+      "step": 2811
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759857780954426e-05,
+      "loss": 1.1631,
+      "step": 2812
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759612283979958e-05,
+      "loss": 1.2041,
+      "step": 2813
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759366663110625e-05,
+      "loss": 1.0819,
+      "step": 2814
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9759120918349547e-05,
+      "loss": 1.0261,
+      "step": 2815
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758875049699834e-05,
+      "loss": 1.1881,
+      "step": 2816
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758629057164617e-05,
+      "loss": 1.1181,
+      "step": 2817
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758382940747014e-05,
+      "loss": 1.2602,
+      "step": 2818
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9758136700450153e-05,
+      "loss": 1.1972,
+      "step": 2819
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975789033627715e-05,
+      "loss": 1.1271,
+      "step": 2820
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757643848231146e-05,
+      "loss": 1.1211,
+      "step": 2821
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757397236315267e-05,
+      "loss": 1.1222,
+      "step": 2822
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9757150500532637e-05,
+      "loss": 1.0617,
+      "step": 2823
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975690364088639e-05,
+      "loss": 1.133,
+      "step": 2824
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9756656657379665e-05,
+      "loss": 1.2517,
+      "step": 2825
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975640955001559e-05,
+      "loss": 1.2516,
+      "step": 2826
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975616231879731e-05,
+      "loss": 1.0354,
+      "step": 2827
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755914963727955e-05,
+      "loss": 1.0632,
+      "step": 2828
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755667484810676e-05,
+      "loss": 1.161,
+      "step": 2829
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755419882048604e-05,
+      "loss": 1.2539,
+      "step": 2830
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9755172155444885e-05,
+      "loss": 1.2166,
+      "step": 2831
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754924305002668e-05,
+      "loss": 1.1031,
+      "step": 2832
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97546763307251e-05,
+      "loss": 1.1825,
+      "step": 2833
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9754428232615317e-05,
+      "loss": 1.3741,
+      "step": 2834
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975418001067648e-05,
+      "loss": 1.1422,
+      "step": 2835
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975393166491174e-05,
+      "loss": 1.1218,
+      "step": 2836
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753683195324246e-05,
+      "loss": 1.3624,
+      "step": 2837
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9753434601917156e-05,
+      "loss": 1.0573,
+      "step": 2838
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975318588469362e-05,
+      "loss": 1.2958,
+      "step": 2839
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752937043656796e-05,
+      "loss": 1.2287,
+      "step": 2840
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752688078809848e-05,
+      "loss": 1.0903,
+      "step": 2841
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752438990155933e-05,
+      "loss": 1.1965,
+      "step": 2842
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9752189777698215e-05,
+      "loss": 1.0397,
+      "step": 2843
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751940441439856e-05,
+      "loss": 1.1855,
+      "step": 2844
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975169098138402e-05,
+      "loss": 1.1091,
+      "step": 2845
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751441397533878e-05,
+      "loss": 1.1476,
+      "step": 2846
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9751191689892595e-05,
+      "loss": 1.1091,
+      "step": 2847
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750941858463338e-05,
+      "loss": 1.1094,
+      "step": 2848
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975069190324929e-05,
+      "loss": 1.15,
+      "step": 2849
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.975044182425361e-05,
+      "loss": 1.225,
+      "step": 2850
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9750191621479483e-05,
+      "loss": 1.1098,
+      "step": 2851
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749941294930083e-05,
+      "loss": 1.0182,
+      "step": 2852
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974969084460858e-05,
+      "loss": 1.0246,
+      "step": 2853
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9749440270518163e-05,
+      "loss": 1.1304,
+      "step": 2854
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974918957266201e-05,
+      "loss": 1.0793,
+      "step": 2855
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97489387510433e-05,
+      "loss": 1.1731,
+      "step": 2856
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748687805665224e-05,
+      "loss": 1.1612,
+      "step": 2857
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748436736530964e-05,
+      "loss": 1.0783,
+      "step": 2858
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9748185543643707e-05,
+      "loss": 1.173,
+      "step": 2859
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974793422700664e-05,
+      "loss": 1.0793,
+      "step": 2860
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747682786622955e-05,
+      "loss": 1.0723,
+      "step": 2861
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747431222495845e-05,
+      "loss": 1.1657,
+      "step": 2862
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9747179534628503e-05,
+      "loss": 1.3082,
+      "step": 2863
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974692772302412e-05,
+      "loss": 1.1392,
+      "step": 2864
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746675787685902e-05,
+      "loss": 1.1529,
+      "step": 2865
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974642372861704e-05,
+      "loss": 1.1086,
+      "step": 2866
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9746171545820735e-05,
+      "loss": 1.068,
+      "step": 2867
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974591923930019e-05,
+      "loss": 1.1095,
+      "step": 2868
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745666809058607e-05,
+      "loss": 1.1703,
+      "step": 2869
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974541425509919e-05,
+      "loss": 1.1275,
+      "step": 2870
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9745161577425144e-05,
+      "loss": 1.027,
+      "step": 2871
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744908776039677e-05,
+      "loss": 1.2602,
+      "step": 2872
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744655850946003e-05,
+      "loss": 1.0859,
+      "step": 2873
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9744402802147325e-05,
+      "loss": 1.1374,
+      "step": 2874
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974414962964686e-05,
+      "loss": 1.235,
+      "step": 2875
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743896333447826e-05,
+      "loss": 1.2566,
+      "step": 2876
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974364291355343e-05,
+      "loss": 1.0894,
+      "step": 2877
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743389369966893e-05,
+      "loss": 1.0791,
+      "step": 2878
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9743135702691435e-05,
+      "loss": 1.0619,
+      "step": 2879
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742881911730274e-05,
+      "loss": 1.1545,
+      "step": 2880
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742627997086633e-05,
+      "loss": 1.081,
+      "step": 2881
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742373958763734e-05,
+      "loss": 1.1093,
+      "step": 2882
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9742119796764805e-05,
+      "loss": 1.1141,
+      "step": 2883
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974186551109307e-05,
+      "loss": 1.1038,
+      "step": 2884
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741611101751755e-05,
+      "loss": 1.0228,
+      "step": 2885
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.974135656874409e-05,
+      "loss": 1.057,
+      "step": 2886
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9741101912073317e-05,
+      "loss": 1.1262,
+      "step": 2887
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740847131742654e-05,
+      "loss": 1.1446,
+      "step": 2888
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740592227755342e-05,
+      "loss": 1.2371,
+      "step": 2889
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740337200114616e-05,
+      "loss": 1.1375,
+      "step": 2890
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9740082048823716e-05,
+      "loss": 1.0225,
+      "step": 2891
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739826773885877e-05,
+      "loss": 1.1241,
+      "step": 2892
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973957137530434e-05,
+      "loss": 1.064,
+      "step": 2893
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739315853082352e-05,
+      "loss": 1.0485,
+      "step": 2894
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9739060207223154e-05,
+      "loss": 1.102,
+      "step": 2895
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738804437729987e-05,
+      "loss": 1.0789,
+      "step": 2896
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738548544606107e-05,
+      "loss": 1.0999,
+      "step": 2897
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9738292527854753e-05,
+      "loss": 1.0894,
+      "step": 2898
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973803638747918e-05,
+      "loss": 1.1174,
+      "step": 2899
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737780123482643e-05,
+      "loss": 1.1503,
+      "step": 2900
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737523735868386e-05,
+      "loss": 1.0683,
+      "step": 2901
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9737267224639668e-05,
+      "loss": 0.9924,
+      "step": 2902
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973701058979975e-05,
+      "loss": 0.9259,
+      "step": 2903
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9736753831351882e-05,
+      "loss": 1.2363,
+      "step": 2904
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973649694929933e-05,
+      "loss": 1.0545,
+      "step": 2905
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973623994364535e-05,
+      "loss": 1.1068,
+      "step": 2906
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735982814393213e-05,
+      "loss": 1.1213,
+      "step": 2907
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735725561546174e-05,
+      "loss": 1.1692,
+      "step": 2908
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97354681851075e-05,
+      "loss": 1.0843,
+      "step": 2909
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9735210685080462e-05,
+      "loss": 1.1398,
+      "step": 2910
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973495306146833e-05,
+      "loss": 1.1186,
+      "step": 2911
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734695314274366e-05,
+      "loss": 1.0258,
+      "step": 2912
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9734437443501847e-05,
+      "loss": 1.2116,
+      "step": 2913
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973417944915405e-05,
+      "loss": 1.2079,
+      "step": 2914
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733921331234248e-05,
+      "loss": 0.9636,
+      "step": 2915
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973366308974572e-05,
+      "loss": 1.1864,
+      "step": 2916
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733404724691738e-05,
+      "loss": 1.0595,
+      "step": 2917
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9733146236075585e-05,
+      "loss": 1.1346,
+      "step": 2918
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732887623900542e-05,
+      "loss": 1.1731,
+      "step": 2919
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732628888169897e-05,
+      "loss": 1.3862,
+      "step": 2920
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732370028886923e-05,
+      "loss": 1.1635,
+      "step": 2921
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9732111046054917e-05,
+      "loss": 1.1232,
+      "step": 2922
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973185193967716e-05,
+      "loss": 1.1813,
+      "step": 2923
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731592709756954e-05,
+      "loss": 1.1024,
+      "step": 2924
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973133335629757e-05,
+      "loss": 1.0945,
+      "step": 2925
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9731073879302315e-05,
+      "loss": 1.0264,
+      "step": 2926
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730814278774476e-05,
+      "loss": 1.1773,
+      "step": 2927
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730554554717355e-05,
+      "loss": 1.0804,
+      "step": 2928
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.973029470713424e-05,
+      "loss": 1.0341,
+      "step": 2929
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9730034736028438e-05,
+      "loss": 1.0349,
+      "step": 2930
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729774641403245e-05,
+      "loss": 1.3002,
+      "step": 2931
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9729514423261964e-05,
+      "loss": 1.2966,
+      "step": 2932
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97292540816079e-05,
+      "loss": 1.0685,
+      "step": 2933
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728993616444353e-05,
+      "loss": 1.0805,
+      "step": 2934
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728733027774637e-05,
+      "loss": 1.1289,
+      "step": 2935
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9728472315602053e-05,
+      "loss": 1.071,
+      "step": 2936
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972821147992991e-05,
+      "loss": 1.1063,
+      "step": 2937
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727950520761526e-05,
+      "loss": 1.0707,
+      "step": 2938
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727689438100213e-05,
+      "loss": 1.0932,
+      "step": 2939
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727428231949276e-05,
+      "loss": 1.1166,
+      "step": 2940
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9727166902312045e-05,
+      "loss": 1.0392,
+      "step": 2941
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726905449191827e-05,
+      "loss": 1.0392,
+      "step": 2942
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726643872591942e-05,
+      "loss": 1.2523,
+      "step": 2943
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9726382172515718e-05,
+      "loss": 1.0682,
+      "step": 2944
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972612034896647e-05,
+      "loss": 1.1706,
+      "step": 2945
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972585840194752e-05,
+      "loss": 1.2621,
+      "step": 2946
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.97255963314622e-05,
+      "loss": 1.2465,
+      "step": 2947
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9725334137513833e-05,
+      "loss": 1.1501,
+      "step": 2948
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972507182010575e-05,
+      "loss": 1.3288,
+      "step": 2949
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724809379241278e-05,
+      "loss": 1.0984,
+      "step": 2950
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724546814923754e-05,
+      "loss": 1.2215,
+      "step": 2951
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724284127156504e-05,
+      "loss": 1.1172,
+      "step": 2952
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9724021315942866e-05,
+      "loss": 1.1499,
+      "step": 2953
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723758381286176e-05,
+      "loss": 0.9653,
+      "step": 2954
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723495323189773e-05,
+      "loss": 1.0742,
+      "step": 2955
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9723232141656996e-05,
+      "loss": 1.0148,
+      "step": 2956
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722968836691186e-05,
+      "loss": 1.1254,
+      "step": 2957
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722705408295685e-05,
+      "loss": 1.0258,
+      "step": 2958
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722441856473836e-05,
+      "loss": 1.1049,
+      "step": 2959
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9722178181228986e-05,
+      "loss": 1.0843,
+      "step": 2960
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972191438256448e-05,
+      "loss": 1.1109,
+      "step": 2961
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721650460483676e-05,
+      "loss": 1.0866,
+      "step": 2962
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.972138641498991e-05,
+      "loss": 1.0391,
+      "step": 2963
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9721122246086544e-05,
+      "loss": 1.1685,
+      "step": 2964
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720857953776932e-05,
+      "loss": 1.179,
+      "step": 2965
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720593538064422e-05,
+      "loss": 1.0967,
+      "step": 2966
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720328998952377e-05,
+      "loss": 1.1004,
+      "step": 2967
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9720064336444152e-05,
+      "loss": 1.2219,
+      "step": 2968
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971979955054311e-05,
+      "loss": 1.1002,
+      "step": 2969
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719534641252608e-05,
+      "loss": 1.1416,
+      "step": 2970
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971926960857601e-05,
+      "loss": 1.0072,
+      "step": 2971
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9719004452516683e-05,
+      "loss": 1.1717,
+      "step": 2972
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971873917307799e-05,
+      "loss": 0.9721,
+      "step": 2973
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9718473770263306e-05,
+      "loss": 1.2071,
+      "step": 2974
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971820824407599e-05,
+      "loss": 1.114,
+      "step": 2975
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9717942594519417e-05,
+      "loss": 1.1731,
+      "step": 2976
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971767682159696e-05,
+      "loss": 1.0134,
+      "step": 2977
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971741092531199e-05,
+      "loss": 1.2069,
+      "step": 2978
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971714490566789e-05,
+      "loss": 1.1472,
+      "step": 2979
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971687876266803e-05,
+      "loss": 0.9816,
+      "step": 2980
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716612496315787e-05,
+      "loss": 1.2959,
+      "step": 2981
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971634610661455e-05,
+      "loss": 1.1122,
+      "step": 2982
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9716079593567694e-05,
+      "loss": 1.0868,
+      "step": 2983
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715812957178602e-05,
+      "loss": 1.1151,
+      "step": 2984
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715546197450658e-05,
+      "loss": 1.0476,
+      "step": 2985
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715279314387256e-05,
+      "loss": 1.0214,
+      "step": 2986
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9715012307991777e-05,
+      "loss": 1.1114,
+      "step": 2987
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714745178267614e-05,
+      "loss": 0.9901,
+      "step": 2988
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714477925218154e-05,
+      "loss": 1.0877,
+      "step": 2989
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9714210548846795e-05,
+      "loss": 1.2124,
+      "step": 2990
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713943049156926e-05,
+      "loss": 1.0928,
+      "step": 2991
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713675426151944e-05,
+      "loss": 1.1972,
+      "step": 2992
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713407679835253e-05,
+      "loss": 1.1223,
+      "step": 2993
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9713139810210246e-05,
+      "loss": 1.0335,
+      "step": 2994
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712871817280318e-05,
+      "loss": 1.2107,
+      "step": 2995
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712603701048882e-05,
+      "loss": 1.1273,
+      "step": 2996
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712335461519335e-05,
+      "loss": 0.9271,
+      "step": 2997
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9712067098695087e-05,
+      "loss": 1.1637,
+      "step": 2998
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971179861257954e-05,
+      "loss": 0.995,
+      "step": 2999
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9711530003176104e-05,
+      "loss": 1.1709,
+      "step": 3000
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971126127048819e-05,
+      "loss": 1.1383,
+      "step": 3001
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971099241451921e-05,
+      "loss": 1.1787,
+      "step": 3002
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710723435272573e-05,
+      "loss": 1.1652,
+      "step": 3003
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9710454332751697e-05,
+      "loss": 1.0768,
+      "step": 3004
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.971018510696e-05,
+      "loss": 1.0517,
+      "step": 3005
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709915757900893e-05,
+      "loss": 1.1527,
+      "step": 3006
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709646285577803e-05,
+      "loss": 0.9492,
+      "step": 3007
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9709376689994142e-05,
+      "loss": 1.1559,
+      "step": 3008
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970910697115334e-05,
+      "loss": 1.2289,
+      "step": 3009
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708837129058823e-05,
+      "loss": 1.1895,
+      "step": 3010
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970856716371401e-05,
+      "loss": 1.2237,
+      "step": 3011
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708297075122326e-05,
+      "loss": 1.0201,
+      "step": 3012
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9708026863287207e-05,
+      "loss": 1.1708,
+      "step": 3013
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970775652821208e-05,
+      "loss": 1.2895,
+      "step": 3014
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970748606990038e-05,
+      "loss": 0.994,
+      "step": 3015
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.9707215488355532e-05,
+      "loss": 1.0631,
+      "step": 3016
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970694478358098e-05,
+      "loss": 1.0631,
+      "step": 3017
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 1.970667395558016e-05,
+      "loss": 1.2397,
+      "step": 3018
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706403004356504e-05,
+      "loss": 0.9916,
+      "step": 3019
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9706131929913453e-05,
+      "loss": 1.0697,
+      "step": 3020
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705860732254448e-05,
+      "loss": 1.1417,
+      "step": 3021
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705589411382936e-05,
+      "loss": 1.0926,
+      "step": 3022
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970531796730236e-05,
+      "loss": 1.121,
+      "step": 3023
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9705046400016166e-05,
+      "loss": 1.081,
+      "step": 3024
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704774709527798e-05,
+      "loss": 1.1168,
+      "step": 3025
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704502895840706e-05,
+      "loss": 1.0351,
+      "step": 3026
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9704230958958345e-05,
+      "loss": 1.0769,
+      "step": 3027
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703958898884162e-05,
+      "loss": 1.0779,
+      "step": 3028
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703686715621614e-05,
+      "loss": 1.096,
+      "step": 3029
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703414409174153e-05,
+      "loss": 1.057,
+      "step": 3030
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9703141979545238e-05,
+      "loss": 1.104,
+      "step": 3031
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970286942673833e-05,
+      "loss": 1.1314,
+      "step": 3032
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702596750756884e-05,
+      "loss": 1.0555,
+      "step": 3033
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9702323951604364e-05,
+      "loss": 1.18,
+      "step": 3034
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970205102928423e-05,
+      "loss": 1.2243,
+      "step": 3035
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970177798379995e-05,
+      "loss": 1.0934,
+      "step": 3036
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701504815154996e-05,
+      "loss": 1.093,
+      "step": 3037
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9701231523352824e-05,
+      "loss": 1.1738,
+      "step": 3038
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970095810839691e-05,
+      "loss": 1.1023,
+      "step": 3039
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.970068457029072e-05,
+      "loss": 1.1544,
+      "step": 3040
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700410909037737e-05,
+      "loss": 1.1405,
+      "step": 3041
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9700137124641423e-05,
+      "loss": 1.1644,
+      "step": 3042
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969986321710526e-05,
+      "loss": 1.1403,
+      "step": 3043
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699589186432724e-05,
+      "loss": 1.1363,
+      "step": 3044
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9699315032627293e-05,
+      "loss": 1.0288,
+      "step": 3045
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969904075569245e-05,
+      "loss": 1.0381,
+      "step": 3046
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698766355631674e-05,
+      "loss": 1.0665,
+      "step": 3047
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698491832448446e-05,
+      "loss": 1.3614,
+      "step": 3048
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9698217186146258e-05,
+      "loss": 1.0002,
+      "step": 3049
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697942416728594e-05,
+      "loss": 1.103,
+      "step": 3050
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9697667524198936e-05,
+      "loss": 1.0619,
+      "step": 3051
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969739250856078e-05,
+      "loss": 0.9907,
+      "step": 3052
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969711736981762e-05,
+      "loss": 1.2276,
+      "step": 3053
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969684210797294e-05,
+      "loss": 1.0416,
+      "step": 3054
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969656672303024e-05,
+      "loss": 1.1397,
+      "step": 3055
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696291214993013e-05,
+      "loss": 1.1696,
+      "step": 3056
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9696015583864764e-05,
+      "loss": 1.14,
+      "step": 3057
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969573982964898e-05,
+      "loss": 1.0966,
+      "step": 3058
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969546395234917e-05,
+      "loss": 1.1436,
+      "step": 3059
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9695187951968835e-05,
+      "loss": 1.0392,
+      "step": 3060
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694911828511474e-05,
+      "loss": 1.1995,
+      "step": 3061
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694635581980602e-05,
+      "loss": 1.0608,
+      "step": 3062
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694359212379715e-05,
+      "loss": 1.255,
+      "step": 3063
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9694082719712327e-05,
+      "loss": 0.9947,
+      "step": 3064
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969380610398195e-05,
+      "loss": 1.0855,
+      "step": 3065
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693529365192092e-05,
+      "loss": 1.1527,
+      "step": 3066
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9693252503346265e-05,
+      "loss": 1.2298,
+      "step": 3067
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692975518447987e-05,
+      "loss": 1.2589,
+      "step": 3068
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9692698410500772e-05,
+      "loss": 1.1279,
+      "step": 3069
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969242117950814e-05,
+      "loss": 1.1001,
+      "step": 3070
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969214382547361e-05,
+      "loss": 1.2066,
+      "step": 3071
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691866348400698e-05,
+      "loss": 1.0599,
+      "step": 3072
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691588748292934e-05,
+      "loss": 0.9623,
+      "step": 3073
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691311025153835e-05,
+      "loss": 1.0962,
+      "step": 3074
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9691033178986934e-05,
+      "loss": 1.224,
+      "step": 3075
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969075520979575e-05,
+      "loss": 1.1286,
+      "step": 3076
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.969047711758382e-05,
+      "loss": 1.1452,
+      "step": 3077
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9690198902354666e-05,
+      "loss": 1.0736,
+      "step": 3078
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689920564111828e-05,
+      "loss": 1.1284,
+      "step": 3079
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9689642102858833e-05,
+      "loss": 1.0762,
+      "step": 3080
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968936351859922e-05,
+      "loss": 0.9961,
+      "step": 3081
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968908481133652e-05,
+      "loss": 1.0279,
+      "step": 3082
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688805981074282e-05,
+      "loss": 1.1467,
+      "step": 3083
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9688527027816034e-05,
+      "loss": 1.0057,
+      "step": 3084
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968824795156532e-05,
+      "loss": 1.1573,
+      "step": 3085
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968796875232569e-05,
+      "loss": 1.1277,
+      "step": 3086
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968768943010068e-05,
+      "loss": 1.2141,
+      "step": 3087
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687409984893836e-05,
+      "loss": 1.0926,
+      "step": 3088
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9687130416708713e-05,
+      "loss": 1.1006,
+      "step": 3089
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686850725548856e-05,
+      "loss": 1.2351,
+      "step": 3090
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968657091141781e-05,
+      "loss": 1.2557,
+      "step": 3091
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9686290974319137e-05,
+      "loss": 1.1292,
+      "step": 3092
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968601091425638e-05,
+      "loss": 1.0142,
+      "step": 3093
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685730731233103e-05,
+      "loss": 1.0047,
+      "step": 3094
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685450425252864e-05,
+      "loss": 1.1808,
+      "step": 3095
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9685169996319212e-05,
+      "loss": 1.099,
+      "step": 3096
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684889444435712e-05,
+      "loss": 0.9306,
+      "step": 3097
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968460876960593e-05,
+      "loss": 1.191,
+      "step": 3098
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968432797183342e-05,
+      "loss": 1.0134,
+      "step": 3099
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9684047051121754e-05,
+      "loss": 1.0032,
+      "step": 3100
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96837660074745e-05,
+      "loss": 1.1772,
+      "step": 3101
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683484840895216e-05,
+      "loss": 1.0057,
+      "step": 3102
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9683203551387477e-05,
+      "loss": 1.2219,
+      "step": 3103
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682922138954855e-05,
+      "loss": 1.0461,
+      "step": 3104
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968264060360092e-05,
+      "loss": 1.0512,
+      "step": 3105
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9682358945329245e-05,
+      "loss": 1.169,
+      "step": 3106
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968207716414341e-05,
+      "loss": 1.0309,
+      "step": 3107
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968179526004699e-05,
+      "loss": 1.061,
+      "step": 3108
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968151323304356e-05,
+      "loss": 1.1169,
+      "step": 3109
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968123108313671e-05,
+      "loss": 1.266,
+      "step": 3110
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680948810330008e-05,
+      "loss": 1.1557,
+      "step": 3111
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.968066641462705e-05,
+      "loss": 1.2021,
+      "step": 3112
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680383896031414e-05,
+      "loss": 1.2025,
+      "step": 3113
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9680101254546686e-05,
+      "loss": 1.1309,
+      "step": 3114
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967981849017646e-05,
+      "loss": 1.2117,
+      "step": 3115
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967953560292432e-05,
+      "loss": 1.0803,
+      "step": 3116
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967925259279386e-05,
+      "loss": 1.0405,
+      "step": 3117
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678969459788665e-05,
+      "loss": 1.0858,
+      "step": 3118
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678686203912344e-05,
+      "loss": 1.241,
+      "step": 3119
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9678402825168482e-05,
+      "loss": 1.0026,
+      "step": 3120
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967811932356068e-05,
+      "loss": 1.2549,
+      "step": 3121
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967783569909253e-05,
+      "loss": 1.1007,
+      "step": 3122
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9677551951767645e-05,
+      "loss": 1.1664,
+      "step": 3123
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967726808158962e-05,
+      "loss": 1.0912,
+      "step": 3124
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967698408856206e-05,
+      "loss": 1.073,
+      "step": 3125
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676699972688565e-05,
+      "loss": 1.2671,
+      "step": 3126
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9676415733972753e-05,
+      "loss": 0.9853,
+      "step": 3127
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967613137241822e-05,
+      "loss": 1.0728,
+      "step": 3128
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967584688802858e-05,
+      "loss": 1.088,
+      "step": 3129
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967556228080745e-05,
+      "loss": 1.0542,
+      "step": 3130
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967527755075844e-05,
+      "loss": 1.0992,
+      "step": 3131
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967499269788516e-05,
+      "loss": 1.2109,
+      "step": 3132
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967470772219123e-05,
+      "loss": 1.1367,
+      "step": 3133
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967442262368027e-05,
+      "loss": 1.2115,
+      "step": 3134
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967413740235589e-05,
+      "loss": 1.1583,
+      "step": 3135
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967385205822172e-05,
+      "loss": 1.0756,
+      "step": 3136
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673566591281382e-05,
+      "loss": 1.218,
+      "step": 3137
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9673281001538495e-05,
+      "loss": 1.0064,
+      "step": 3138
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672995288996687e-05,
+      "loss": 1.1329,
+      "step": 3139
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672709453659588e-05,
+      "loss": 1.2749,
+      "step": 3140
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967242349553082e-05,
+      "loss": 1.1974,
+      "step": 3141
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9672137414614017e-05,
+      "loss": 1.2929,
+      "step": 3142
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967185121091281e-05,
+      "loss": 1.1403,
+      "step": 3143
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9671564884430832e-05,
+      "loss": 1.1542,
+      "step": 3144
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967127843517172e-05,
+      "loss": 1.1234,
+      "step": 3145
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967099186313911e-05,
+      "loss": 1.1551,
+      "step": 3146
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670705168336636e-05,
+      "loss": 1.1301,
+      "step": 3147
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.967041835076794e-05,
+      "loss": 1.2068,
+      "step": 3148
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9670131410436667e-05,
+      "loss": 1.1397,
+      "step": 3149
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669844347346453e-05,
+      "loss": 1.1195,
+      "step": 3150
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9669557161500944e-05,
+      "loss": 1.0606,
+      "step": 3151
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966926985290379e-05,
+      "loss": 1.1332,
+      "step": 3152
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966898242155863e-05,
+      "loss": 1.2414,
+      "step": 3153
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668694867469122e-05,
+      "loss": 1.0509,
+      "step": 3154
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9668407190638912e-05,
+      "loss": 1.1769,
+      "step": 3155
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966811939107165e-05,
+      "loss": 1.1441,
+      "step": 3156
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667831468770994e-05,
+      "loss": 1.0396,
+      "step": 3157
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9667543423740594e-05,
+      "loss": 1.2368,
+      "step": 3158
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966725525598411e-05,
+      "loss": 1.1372,
+      "step": 3159
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666966965505203e-05,
+      "loss": 1.2521,
+      "step": 3160
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666678552307527e-05,
+      "loss": 1.2121,
+      "step": 3161
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666390016394744e-05,
+      "loss": 1.0282,
+      "step": 3162
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9666101357770522e-05,
+      "loss": 1.117,
+      "step": 3163
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665812576438518e-05,
+      "loss": 1.1372,
+      "step": 3164
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665523672402404e-05,
+      "loss": 1.0358,
+      "step": 3165
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9665234645665845e-05,
+      "loss": 1.0939,
+      "step": 3166
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664945496232512e-05,
+      "loss": 1.0375,
+      "step": 3167
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9664656224106074e-05,
+      "loss": 1.0931,
+      "step": 3168
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96643668292902e-05,
+      "loss": 1.1747,
+      "step": 3169
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966407731178857e-05,
+      "loss": 1.2552,
+      "step": 3170
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663787671604856e-05,
+      "loss": 1.029,
+      "step": 3171
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663497908742734e-05,
+      "loss": 1.0268,
+      "step": 3172
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9663208023205887e-05,
+      "loss": 1.1218,
+      "step": 3173
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662918014997993e-05,
+      "loss": 1.1825,
+      "step": 3174
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662627884122725e-05,
+      "loss": 1.0679,
+      "step": 3175
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966233763058378e-05,
+      "loss": 0.9959,
+      "step": 3176
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9662047254384832e-05,
+      "loss": 1.1167,
+      "step": 3177
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661756755529575e-05,
+      "loss": 1.1784,
+      "step": 3178
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661466134021688e-05,
+      "loss": 1.0826,
+      "step": 3179
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9661175389864873e-05,
+      "loss": 1.122,
+      "step": 3180
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660884523062807e-05,
+      "loss": 1.1764,
+      "step": 3181
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966059353361919e-05,
+      "loss": 1.1541,
+      "step": 3182
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9660302421537718e-05,
+      "loss": 1.1297,
+      "step": 3183
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.966001118682208e-05,
+      "loss": 1.0294,
+      "step": 3184
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965971982947598e-05,
+      "loss": 1.1526,
+      "step": 3185
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659428349503108e-05,
+      "loss": 0.9776,
+      "step": 3186
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9659136746907173e-05,
+      "loss": 1.0174,
+      "step": 3187
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658845021691872e-05,
+      "loss": 1.1604,
+      "step": 3188
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658553173860912e-05,
+      "loss": 1.158,
+      "step": 3189
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9658261203417992e-05,
+      "loss": 1.1737,
+      "step": 3190
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657969110366823e-05,
+      "loss": 1.3693,
+      "step": 3191
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965767689471111e-05,
+      "loss": 1.2604,
+      "step": 3192
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965738455645457e-05,
+      "loss": 1.0634,
+      "step": 3193
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9657092095600906e-05,
+      "loss": 1.0633,
+      "step": 3194
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965679951215383e-05,
+      "loss": 1.1813,
+      "step": 3195
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9656506806117065e-05,
+      "loss": 1.1716,
+      "step": 3196
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965621397749432e-05,
+      "loss": 1.1535,
+      "step": 3197
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655921026289314e-05,
+      "loss": 1.0535,
+      "step": 3198
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655627952505763e-05,
+      "loss": 1.1153,
+      "step": 3199
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965533475614739e-05,
+      "loss": 1.1382,
+      "step": 3200
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9655041437217922e-05,
+      "loss": 1.1496,
+      "step": 3201
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654747995721072e-05,
+      "loss": 1.0946,
+      "step": 3202
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9654454431660574e-05,
+      "loss": 1.2346,
+      "step": 3203
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965416074504015e-05,
+      "loss": 1.1412,
+      "step": 3204
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965386693586353e-05,
+      "loss": 1.2074,
+      "step": 3205
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653573004134446e-05,
+      "loss": 1.0819,
+      "step": 3206
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9653278949856625e-05,
+      "loss": 1.1535,
+      "step": 3207
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652984773033804e-05,
+      "loss": 1.0873,
+      "step": 3208
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965269047366971e-05,
+      "loss": 1.0896,
+      "step": 3209
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9652396051768086e-05,
+      "loss": 1.2324,
+      "step": 3210
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965210150733267e-05,
+      "loss": 1.0482,
+      "step": 3211
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96518068403672e-05,
+      "loss": 1.0907,
+      "step": 3212
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9651512050875417e-05,
+      "loss": 1.0398,
+      "step": 3213
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.965121713886106e-05,
+      "loss": 1.0201,
+      "step": 3214
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650922104327878e-05,
+      "loss": 1.1024,
+      "step": 3215
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650626947279607e-05,
+      "loss": 1.1596,
+      "step": 3216
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650331667720005e-05,
+      "loss": 1.1938,
+      "step": 3217
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9650036265652814e-05,
+      "loss": 1.119,
+      "step": 3218
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649740741081786e-05,
+      "loss": 1.1301,
+      "step": 3219
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649445094010667e-05,
+      "loss": 1.1642,
+      "step": 3220
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9649149324443223e-05,
+      "loss": 1.2343,
+      "step": 3221
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96488534323832e-05,
+      "loss": 1.0745,
+      "step": 3222
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9648557417834352e-05,
+      "loss": 1.2192,
+      "step": 3223
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964826128080044e-05,
+      "loss": 1.0161,
+      "step": 3224
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647965021285226e-05,
+      "loss": 1.1313,
+      "step": 3225
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9647668639292466e-05,
+      "loss": 1.0489,
+      "step": 3226
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964737213482593e-05,
+      "loss": 1.1236,
+      "step": 3227
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964707550788937e-05,
+      "loss": 1.2203,
+      "step": 3228
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9646778758486565e-05,
+      "loss": 1.1796,
+      "step": 3229
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964648188662127e-05,
+      "loss": 1.022,
+      "step": 3230
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964618489229726e-05,
+      "loss": 1.1159,
+      "step": 3231
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645887775518307e-05,
+      "loss": 0.9509,
+      "step": 3232
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964559053628818e-05,
+      "loss": 1.1195,
+      "step": 3233
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9645293174610652e-05,
+      "loss": 1.0335,
+      "step": 3234
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96449956904895e-05,
+      "loss": 1.0823,
+      "step": 3235
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644698083928496e-05,
+      "loss": 1.1593,
+      "step": 3236
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964440035493142e-05,
+      "loss": 1.1424,
+      "step": 3237
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9644102503502055e-05,
+      "loss": 1.205,
+      "step": 3238
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964380452964418e-05,
+      "loss": 1.1397,
+      "step": 3239
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9643506433361577e-05,
+      "loss": 1.1696,
+      "step": 3240
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964320821465803e-05,
+      "loss": 1.1363,
+      "step": 3241
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642909873537326e-05,
+      "loss": 0.9401,
+      "step": 3242
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642611410003253e-05,
+      "loss": 1.0142,
+      "step": 3243
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9642312824059594e-05,
+      "loss": 1.1638,
+      "step": 3244
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964201411571015e-05,
+      "loss": 1.0616,
+      "step": 3245
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641715284958702e-05,
+      "loss": 1.1988,
+      "step": 3246
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9641416331809052e-05,
+      "loss": 1.152,
+      "step": 3247
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.964111725626499e-05,
+      "loss": 1.1015,
+      "step": 3248
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640818058330317e-05,
+      "loss": 1.0674,
+      "step": 3249
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640518738008825e-05,
+      "loss": 1.2021,
+      "step": 3250
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9640219295304316e-05,
+      "loss": 1.2291,
+      "step": 3251
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96399197302206e-05,
+      "loss": 1.1741,
+      "step": 3252
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639620042761468e-05,
+      "loss": 1.173,
+      "step": 3253
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639320232930728e-05,
+      "loss": 1.1643,
+      "step": 3254
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9639020300732188e-05,
+      "loss": 1.0111,
+      "step": 3255
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638720246169652e-05,
+      "loss": 1.2057,
+      "step": 3256
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638420069246935e-05,
+      "loss": 0.9776,
+      "step": 3257
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9638119769967843e-05,
+      "loss": 1.1493,
+      "step": 3258
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963781934833619e-05,
+      "loss": 1.0442,
+      "step": 3259
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637518804355787e-05,
+      "loss": 1.0694,
+      "step": 3260
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9637218138030454e-05,
+      "loss": 1.1158,
+      "step": 3261
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636917349364006e-05,
+      "loss": 1.1109,
+      "step": 3262
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636616438360257e-05,
+      "loss": 1.0869,
+      "step": 3263
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636315405023035e-05,
+      "loss": 1.075,
+      "step": 3264
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9636014249356154e-05,
+      "loss": 1.0946,
+      "step": 3265
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963571297136344e-05,
+      "loss": 1.2012,
+      "step": 3266
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963541157104872e-05,
+      "loss": 1.1746,
+      "step": 3267
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9635110048415816e-05,
+      "loss": 1.1266,
+      "step": 3268
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634808403468558e-05,
+      "loss": 1.1259,
+      "step": 3269
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9634506636210776e-05,
+      "loss": 1.1451,
+      "step": 3270
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.96342047466463e-05,
+      "loss": 1.1647,
+      "step": 3271
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963390273477896e-05,
+      "loss": 1.296,
+      "step": 3272
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633600600612595e-05,
+      "loss": 1.1107,
+      "step": 3273
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9633298344151037e-05,
+      "loss": 1.1472,
+      "step": 3274
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632995965398124e-05,
+      "loss": 1.0772,
+      "step": 3275
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632693464357695e-05,
+      "loss": 1.0651,
+      "step": 3276
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632390841033587e-05,
+      "loss": 1.119,
+      "step": 3277
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9632088095429647e-05,
+      "loss": 1.1077,
+      "step": 3278
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631785227549712e-05,
+      "loss": 1.0907,
+      "step": 3279
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9631482237397634e-05,
+      "loss": 1.1068,
+      "step": 3280
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963117912497725e-05,
+      "loss": 1.1534,
+      "step": 3281
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630875890292418e-05,
+      "loss": 1.1025,
+      "step": 3282
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9630572533346982e-05,
+      "loss": 1.1155,
+      "step": 3283
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.963026905414479e-05,
+      "loss": 1.0834,
+      "step": 3284
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629965452689706e-05,
+      "loss": 1.1942,
+      "step": 3285
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629661728985573e-05,
+      "loss": 1.0284,
+      "step": 3286
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962935788303625e-05,
+      "loss": 1.1453,
+      "step": 3287
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9629053914845595e-05,
+      "loss": 1.0416,
+      "step": 3288
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628749824417467e-05,
+      "loss": 1.2174,
+      "step": 3289
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9628445611755726e-05,
+      "loss": 1.2074,
+      "step": 3290
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962814127686423e-05,
+      "loss": 1.1706,
+      "step": 3291
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962783681974685e-05,
+      "loss": 1.1246,
+      "step": 3292
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962753224040744e-05,
+      "loss": 1.1661,
+      "step": 3293
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962722753884988e-05,
+      "loss": 1.1669,
+      "step": 3294
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626922715078028e-05,
+      "loss": 1.223,
+      "step": 3295
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962661776909576e-05,
+      "loss": 1.033,
+      "step": 3296
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626312700906945e-05,
+      "loss": 1.2566,
+      "step": 3297
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9626007510515452e-05,
+      "loss": 1.1249,
+      "step": 3298
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962570219792516e-05,
+      "loss": 1.2174,
+      "step": 3299
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962539676313994e-05,
+      "loss": 1.1678,
+      "step": 3300
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.9625091206163674e-05,
+      "loss": 1.1033,
+      "step": 3301
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962478552700024e-05,
+      "loss": 1.1471,
+      "step": 3302
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962447972565352e-05,
+      "loss": 1.0617,
+      "step": 3303
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962417380212739e-05,
+      "loss": 1.1551,
+      "step": 3304
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 1.962386775642574e-05,
+      "loss": 1.213,
+      "step": 3305
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623561588552454e-05,
+      "loss": 1.1137,
+      "step": 3306
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9623255298511418e-05,
+      "loss": 1.1232,
+      "step": 3307
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622948886306514e-05,
+      "loss": 1.0925,
+      "step": 3308
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622642351941643e-05,
+      "loss": 0.9296,
+      "step": 3309
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962233569542069e-05,
+      "loss": 1.1228,
+      "step": 3310
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9622028916747546e-05,
+      "loss": 0.9873,
+      "step": 3311
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962172201592611e-05,
+      "loss": 1.2396,
+      "step": 3312
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621414992960278e-05,
+      "loss": 1.0864,
+      "step": 3313
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9621107847853942e-05,
+      "loss": 1.0828,
+      "step": 3314
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620800580611006e-05,
+      "loss": 1.1021,
+      "step": 3315
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.962049319123537e-05,
+      "loss": 1.0216,
+      "step": 3316
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9620185679730938e-05,
+      "loss": 1.01,
+      "step": 3317
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619878046101606e-05,
+      "loss": 0.8963,
+      "step": 3318
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619570290351285e-05,
+      "loss": 1.2208,
+      "step": 3319
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9619262412483886e-05,
+      "loss": 1.1618,
+      "step": 3320
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961895441250331e-05,
+      "loss": 1.165,
+      "step": 3321
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618646290413474e-05,
+      "loss": 1.0128,
+      "step": 3322
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961833804621828e-05,
+      "loss": 1.1552,
+      "step": 3323
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9618029679921648e-05,
+      "loss": 1.0369,
+      "step": 3324
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617721191527493e-05,
+      "loss": 1.0664,
+      "step": 3325
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961741258103973e-05,
+      "loss": 0.9351,
+      "step": 3326
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9617103848462272e-05,
+      "loss": 1.1243,
+      "step": 3327
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616794993799046e-05,
+      "loss": 1.1124,
+      "step": 3328
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616486017053968e-05,
+      "loss": 1.0441,
+      "step": 3329
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9616176918230957e-05,
+      "loss": 1.2422,
+      "step": 3330
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615867697333944e-05,
+      "loss": 1.0201,
+      "step": 3331
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615558354366856e-05,
+      "loss": 1.1022,
+      "step": 3332
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9615248889333613e-05,
+      "loss": 1.1567,
+      "step": 3333
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614939302238145e-05,
+      "loss": 1.0155,
+      "step": 3334
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614629593084382e-05,
+      "loss": 1.0393,
+      "step": 3335
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9614319761876258e-05,
+      "loss": 1.1433,
+      "step": 3336
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961400980861771e-05,
+      "loss": 1.3608,
+      "step": 3337
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961369973331266e-05,
+      "loss": 1.0492,
+      "step": 3338
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961338953596506e-05,
+      "loss": 1.1291,
+      "step": 3339
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9613079216578833e-05,
+      "loss": 1.0679,
+      "step": 3340
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612768775157928e-05,
+      "loss": 1.1254,
+      "step": 3341
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9612458211706284e-05,
+      "loss": 1.2041,
+      "step": 3342
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961214752622784e-05,
+      "loss": 1.1356,
+      "step": 3343
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961183671872655e-05,
+      "loss": 1.0478,
+      "step": 3344
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961152578920635e-05,
+      "loss": 1.1803,
+      "step": 3345
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9611214737671186e-05,
+      "loss": 1.0538,
+      "step": 3346
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9610903564125014e-05,
+      "loss": 1.2261,
+      "step": 3347
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961059226857178e-05,
+      "loss": 0.9623,
+      "step": 3348
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.961028085101543e-05,
+      "loss": 1.1445,
+      "step": 3349
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609969311459933e-05,
+      "loss": 1.1697,
+      "step": 3350
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960965764990923e-05,
+      "loss": 1.0273,
+      "step": 3351
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9609345866367283e-05,
+      "loss": 1.1509,
+      "step": 3352
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960903396083805e-05,
+      "loss": 1.1888,
+      "step": 3353
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608721933325487e-05,
+      "loss": 1.0897,
+      "step": 3354
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960840978383356e-05,
+      "loss": 1.1123,
+      "step": 3355
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9608097512366226e-05,
+      "loss": 1.0622,
+      "step": 3356
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607785118927455e-05,
+      "loss": 1.0265,
+      "step": 3357
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9607472603521212e-05,
+      "loss": 1.0858,
+      "step": 3358
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960715996615146e-05,
+      "loss": 1.0306,
+      "step": 3359
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960684720682217e-05,
+      "loss": 1.1796,
+      "step": 3360
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606534325537315e-05,
+      "loss": 1.0538,
+      "step": 3361
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9606221322300864e-05,
+      "loss": 1.1148,
+      "step": 3362
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960590819711679e-05,
+      "loss": 1.1221,
+      "step": 3363
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605594949989068e-05,
+      "loss": 1.1623,
+      "step": 3364
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9605281580921675e-05,
+      "loss": 1.186,
+      "step": 3365
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604968089918594e-05,
+      "loss": 1.1418,
+      "step": 3366
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604654476983795e-05,
+      "loss": 1.0919,
+      "step": 3367
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960434074212127e-05,
+      "loss": 1.2106,
+      "step": 3368
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9604026885334993e-05,
+      "loss": 1.0866,
+      "step": 3369
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960371290662895e-05,
+      "loss": 1.0971,
+      "step": 3370
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960339880600713e-05,
+      "loss": 1.2073,
+      "step": 3371
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960308458347352e-05,
+      "loss": 1.3866,
+      "step": 3372
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960277023903211e-05,
+      "loss": 1.1703,
+      "step": 3373
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602455772686884e-05,
+      "loss": 1.1324,
+      "step": 3374
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9602141184441843e-05,
+      "loss": 1.1605,
+      "step": 3375
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601826474300975e-05,
+      "loss": 1.0385,
+      "step": 3376
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960151164226827e-05,
+      "loss": 1.0595,
+      "step": 3377
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9601196688347735e-05,
+      "loss": 1.1284,
+      "step": 3378
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9600881612543367e-05,
+      "loss": 1.2663,
+      "step": 3379
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960056641485916e-05,
+      "loss": 1.2511,
+      "step": 3380
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.960025109529912e-05,
+      "loss": 1.1816,
+      "step": 3381
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9599935653867247e-05,
+      "loss": 1.1596,
+      "step": 3382
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959962009056754e-05,
+      "loss": 1.0446,
+      "step": 3383
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959930440540402e-05,
+      "loss": 1.2577,
+      "step": 3384
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598988598380683e-05,
+      "loss": 1.2131,
+      "step": 3385
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598672669501543e-05,
+      "loss": 1.1948,
+      "step": 3386
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598356618770604e-05,
+      "loss": 1.0756,
+      "step": 3387
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9598040446191885e-05,
+      "loss": 1.0592,
+      "step": 3388
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597724151769398e-05,
+      "loss": 1.0974,
+      "step": 3389
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597407735507157e-05,
+      "loss": 1.0283,
+      "step": 3390
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9597091197409178e-05,
+      "loss": 1.0363,
+      "step": 3391
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596774537479485e-05,
+      "loss": 1.1181,
+      "step": 3392
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9596457755722088e-05,
+      "loss": 0.8826,
+      "step": 3393
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959614085214102e-05,
+      "loss": 1.1715,
+      "step": 3394
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595823826740294e-05,
+      "loss": 1.0961,
+      "step": 3395
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9595506679523937e-05,
+      "loss": 1.2331,
+      "step": 3396
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959518941049598e-05,
+      "loss": 1.2198,
+      "step": 3397
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594872019660446e-05,
+      "loss": 1.2003,
+      "step": 3398
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594554507021367e-05,
+      "loss": 1.1614,
+      "step": 3399
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9594236872582773e-05,
+      "loss": 1.1646,
+      "step": 3400
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959391911634869e-05,
+      "loss": 1.0228,
+      "step": 3401
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9593601238323163e-05,
+      "loss": 1.1527,
+      "step": 3402
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959328323851022e-05,
+      "loss": 1.1578,
+      "step": 3403
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95929651169139e-05,
+      "loss": 1.1122,
+      "step": 3404
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959264687353824e-05,
+      "loss": 1.1149,
+      "step": 3405
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959232850838728e-05,
+      "loss": 1.0481,
+      "step": 3406
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9592010021465064e-05,
+      "loss": 1.0704,
+      "step": 3407
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591691412775635e-05,
+      "loss": 1.2117,
+      "step": 3408
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959137268232303e-05,
+      "loss": 1.2379,
+      "step": 3409
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9591053830111307e-05,
+      "loss": 1.1266,
+      "step": 3410
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590734856144506e-05,
+      "loss": 1.1259,
+      "step": 3411
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.959041576042668e-05,
+      "loss": 1.3314,
+      "step": 3412
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9590096542961875e-05,
+      "loss": 1.1408,
+      "step": 3413
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958977720375415e-05,
+      "loss": 1.1019,
+      "step": 3414
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958945774280755e-05,
+      "loss": 1.1171,
+      "step": 3415
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9589138160126143e-05,
+      "loss": 1.1936,
+      "step": 3416
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588818455713975e-05,
+      "loss": 1.1074,
+      "step": 3417
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958849862957511e-05,
+      "loss": 1.2162,
+      "step": 3418
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9588178681713607e-05,
+      "loss": 1.0898,
+      "step": 3419
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958785861213352e-05,
+      "loss": 1.0446,
+      "step": 3420
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587538420838927e-05,
+      "loss": 1.1938,
+      "step": 3421
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9587218107833884e-05,
+      "loss": 1.0473,
+      "step": 3422
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586897673122456e-05,
+      "loss": 1.162,
+      "step": 3423
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586577116708713e-05,
+      "loss": 0.9529,
+      "step": 3424
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9586256438596725e-05,
+      "loss": 1.1008,
+      "step": 3425
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585935638790562e-05,
+      "loss": 1.0533,
+      "step": 3426
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585614717294296e-05,
+      "loss": 1.1642,
+      "step": 3427
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9585293674112002e-05,
+      "loss": 0.9056,
+      "step": 3428
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584972509247758e-05,
+      "loss": 1.0826,
+      "step": 3429
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584651222705635e-05,
+      "loss": 1.1828,
+      "step": 3430
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584329814489715e-05,
+      "loss": 1.1021,
+      "step": 3431
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9584008284604083e-05,
+      "loss": 1.1764,
+      "step": 3432
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583686633052806e-05,
+      "loss": 1.0397,
+      "step": 3433
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583364859839985e-05,
+      "loss": 1.0534,
+      "step": 3434
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9583042964969695e-05,
+      "loss": 1.0982,
+      "step": 3435
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582720948446026e-05,
+      "loss": 1.0559,
+      "step": 3436
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.958239881027306e-05,
+      "loss": 1.1418,
+      "step": 3437
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9582076550454892e-05,
+      "loss": 1.0919,
+      "step": 3438
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581754168995613e-05,
+      "loss": 1.1361,
+      "step": 3439
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581431665899312e-05,
+      "loss": 1.0201,
+      "step": 3440
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9581109041170086e-05,
+      "loss": 1.2136,
+      "step": 3441
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580786294812032e-05,
+      "loss": 1.1658,
+      "step": 3442
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580463426829242e-05,
+      "loss": 1.0811,
+      "step": 3443
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9580140437225816e-05,
+      "loss": 1.0865,
+      "step": 3444
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957981732600586e-05,
+      "loss": 1.0335,
+      "step": 3445
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9579494093173466e-05,
+      "loss": 1.0827,
+      "step": 3446
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957917073873275e-05,
+      "loss": 1.0859,
+      "step": 3447
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9578847262687802e-05,
+      "loss": 1.1028,
+      "step": 3448
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957852366504274e-05,
+      "loss": 1.1968,
+      "step": 3449
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957819994580167e-05,
+      "loss": 1.2414,
+      "step": 3450
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95778761049687e-05,
+      "loss": 1.1597,
+      "step": 3451
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9577552142547936e-05,
+      "loss": 0.9561,
+      "step": 3452
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95772280585435e-05,
+      "loss": 1.1238,
+      "step": 3453
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576903852959498e-05,
+      "loss": 1.2395,
+      "step": 3454
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957657952580005e-05,
+      "loss": 1.1861,
+      "step": 3455
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9576255077069273e-05,
+      "loss": 1.1203,
+      "step": 3456
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575930506771287e-05,
+      "loss": 1.0,
+      "step": 3457
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575605814910205e-05,
+      "loss": 1.2326,
+      "step": 3458
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9575281001490158e-05,
+      "loss": 1.1801,
+      "step": 3459
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574956066515262e-05,
+      "loss": 1.1669,
+      "step": 3460
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957463100998965e-05,
+      "loss": 1.0446,
+      "step": 3461
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9574305831917445e-05,
+      "loss": 1.2658,
+      "step": 3462
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573980532302766e-05,
+      "loss": 0.95,
+      "step": 3463
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957365511114976e-05,
+      "loss": 0.9798,
+      "step": 3464
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573329568462544e-05,
+      "loss": 1.1489,
+      "step": 3465
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9573003904245254e-05,
+      "loss": 1.3036,
+      "step": 3466
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957267811850203e-05,
+      "loss": 1.1899,
+      "step": 3467
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572352211236998e-05,
+      "loss": 1.1721,
+      "step": 3468
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9572026182454305e-05,
+      "loss": 1.0654,
+      "step": 3469
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571700032158085e-05,
+      "loss": 1.0147,
+      "step": 3470
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571373760352474e-05,
+      "loss": 1.1506,
+      "step": 3471
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9571047367041624e-05,
+      "loss": 1.0949,
+      "step": 3472
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.957072085222967e-05,
+      "loss": 1.0968,
+      "step": 3473
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570394215920762e-05,
+      "loss": 1.1692,
+      "step": 3474
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9570067458119047e-05,
+      "loss": 1.3258,
+      "step": 3475
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569740578828665e-05,
+      "loss": 1.0576,
+      "step": 3476
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569413578053774e-05,
+      "loss": 1.0333,
+      "step": 3477
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9569086455798523e-05,
+      "loss": 1.1036,
+      "step": 3478
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9568759212067064e-05,
+      "loss": 1.1265,
+      "step": 3479
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956843184686355e-05,
+      "loss": 1.0463,
+      "step": 3480
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956810436019214e-05,
+      "loss": 1.2049,
+      "step": 3481
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956777675205699e-05,
+      "loss": 1.2089,
+      "step": 3482
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9567449022462256e-05,
+      "loss": 1.1505,
+      "step": 3483
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95671211714121e-05,
+      "loss": 1.2535,
+      "step": 3484
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956679319891069e-05,
+      "loss": 0.9464,
+      "step": 3485
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956646510496218e-05,
+      "loss": 1.0121,
+      "step": 3486
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9566136889570743e-05,
+      "loss": 1.1625,
+      "step": 3487
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956580855274054e-05,
+      "loss": 1.1331,
+      "step": 3488
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565480094475737e-05,
+      "loss": 1.1004,
+      "step": 3489
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9565151514780512e-05,
+      "loss": 1.0455,
+      "step": 3490
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956482281365903e-05,
+      "loss": 1.0833,
+      "step": 3491
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564493991115467e-05,
+      "loss": 1.1454,
+      "step": 3492
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9564165047153995e-05,
+      "loss": 1.2123,
+      "step": 3493
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563835981778792e-05,
+      "loss": 1.147,
+      "step": 3494
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9563506794994035e-05,
+      "loss": 1.1292,
+      "step": 3495
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95631774868039e-05,
+      "loss": 1.1045,
+      "step": 3496
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956284805721257e-05,
+      "loss": 1.1284,
+      "step": 3497
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562518506224227e-05,
+      "loss": 1.1166,
+      "step": 3498
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9562188833843053e-05,
+      "loss": 1.1145,
+      "step": 3499
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561859040073235e-05,
+      "loss": 1.0746,
+      "step": 3500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9561529124918957e-05,
+      "loss": 1.1709,
+      "step": 3501
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956119908838441e-05,
+      "loss": 1.2495,
+      "step": 3502
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560868930473784e-05,
+      "loss": 1.0708,
+      "step": 3503
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9560538651191267e-05,
+      "loss": 1.0987,
+      "step": 3504
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.956020825054105e-05,
+      "loss": 1.1537,
+      "step": 3505
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955987772852734e-05,
+      "loss": 1.2497,
+      "step": 3506
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559547085154316e-05,
+      "loss": 1.1838,
+      "step": 3507
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9559216320426185e-05,
+      "loss": 1.1275,
+      "step": 3508
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558885434347144e-05,
+      "loss": 1.1495,
+      "step": 3509
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9558554426921397e-05,
+      "loss": 1.0634,
+      "step": 3510
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955822329815314e-05,
+      "loss": 0.9768,
+      "step": 3511
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955789204804658e-05,
+      "loss": 1.0535,
+      "step": 3512
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955756067660592e-05,
+      "loss": 1.1282,
+      "step": 3513
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9557229183835368e-05,
+      "loss": 1.1877,
+      "step": 3514
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556897569739135e-05,
+      "loss": 1.2234,
+      "step": 3515
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9556565834321425e-05,
+      "loss": 1.0762,
+      "step": 3516
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955623397758645e-05,
+      "loss": 1.2109,
+      "step": 3517
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955590199953843e-05,
+      "loss": 1.0918,
+      "step": 3518
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955556990018157e-05,
+      "loss": 1.2398,
+      "step": 3519
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9555237679520093e-05,
+      "loss": 1.0765,
+      "step": 3520
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554905337558213e-05,
+      "loss": 1.2452,
+      "step": 3521
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554572874300148e-05,
+      "loss": 1.0566,
+      "step": 3522
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9554240289750123e-05,
+      "loss": 1.1536,
+      "step": 3523
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553907583912352e-05,
+      "loss": 1.0396,
+      "step": 3524
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9553574756791068e-05,
+      "loss": 1.0443,
+      "step": 3525
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955324180839049e-05,
+      "loss": 1.0342,
+      "step": 3526
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552908738714847e-05,
+      "loss": 1.1386,
+      "step": 3527
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9552575547768365e-05,
+      "loss": 1.1287,
+      "step": 3528
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955224223555528e-05,
+      "loss": 1.1226,
+      "step": 3529
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551908802079815e-05,
+      "loss": 1.0863,
+      "step": 3530
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9551575247346207e-05,
+      "loss": 0.9906,
+      "step": 3531
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.955124157135869e-05,
+      "loss": 0.9676,
+      "step": 3532
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550907774121502e-05,
+      "loss": 1.0678,
+      "step": 3533
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550573855638875e-05,
+      "loss": 1.0993,
+      "step": 3534
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9550239815915052e-05,
+      "loss": 1.1472,
+      "step": 3535
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549905654954272e-05,
+      "loss": 1.2466,
+      "step": 3536
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954957137276078e-05,
+      "loss": 1.1671,
+      "step": 3537
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9549236969338814e-05,
+      "loss": 1.1052,
+      "step": 3538
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548902444692627e-05,
+      "loss": 1.0998,
+      "step": 3539
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9548567798826457e-05,
+      "loss": 1.1381,
+      "step": 3540
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954823303174456e-05,
+      "loss": 1.039,
+      "step": 3541
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954789814345118e-05,
+      "loss": 1.0835,
+      "step": 3542
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547563133950567e-05,
+      "loss": 1.2111,
+      "step": 3543
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9547228003246984e-05,
+      "loss": 1.1493,
+      "step": 3544
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546892751344674e-05,
+      "loss": 1.1742,
+      "step": 3545
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546557378247898e-05,
+      "loss": 1.3253,
+      "step": 3546
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9546221883960914e-05,
+      "loss": 1.0458,
+      "step": 3547
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954588626848798e-05,
+      "loss": 1.1703,
+      "step": 3548
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545550531833358e-05,
+      "loss": 1.0229,
+      "step": 3549
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9545214674001306e-05,
+      "loss": 1.0195,
+      "step": 3550
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9544878694996093e-05,
+      "loss": 1.0754,
+      "step": 3551
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954454259482198e-05,
+      "loss": 1.1808,
+      "step": 3552
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954420637348323e-05,
+      "loss": 1.1906,
+      "step": 3553
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954387003098412e-05,
+      "loss": 1.1769,
+      "step": 3554
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9543533567328916e-05,
+      "loss": 1.0908,
+      "step": 3555
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954319698252189e-05,
+      "loss": 0.9552,
+      "step": 3556
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542860276567316e-05,
+      "loss": 1.1082,
+      "step": 3557
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9542523449469462e-05,
+      "loss": 1.1026,
+      "step": 3558
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.954218650123261e-05,
+      "loss": 1.1036,
+      "step": 3559
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541849431861036e-05,
+      "loss": 1.0845,
+      "step": 3560
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541512241359018e-05,
+      "loss": 1.1076,
+      "step": 3561
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9541174929730835e-05,
+      "loss": 1.105,
+      "step": 3562
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540837496980775e-05,
+      "loss": 1.0796,
+      "step": 3563
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540499943113115e-05,
+      "loss": 1.0231,
+      "step": 3564
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9540162268132148e-05,
+      "loss": 1.0454,
+      "step": 3565
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953982447204215e-05,
+      "loss": 1.0923,
+      "step": 3566
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539486554847418e-05,
+      "loss": 1.2085,
+      "step": 3567
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9539148516552237e-05,
+      "loss": 1.1285,
+      "step": 3568
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538810357160903e-05,
+      "loss": 1.05,
+      "step": 3569
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9538472076677698e-05,
+      "loss": 1.4091,
+      "step": 3570
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953813367510693e-05,
+      "loss": 0.9616,
+      "step": 3571
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953779515245289e-05,
+      "loss": 1.1497,
+      "step": 3572
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953745650871987e-05,
+      "loss": 1.2017,
+      "step": 3573
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9537117743912178e-05,
+      "loss": 1.0426,
+      "step": 3574
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953677885803411e-05,
+      "loss": 1.1946,
+      "step": 3575
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9536439851089962e-05,
+      "loss": 1.032,
+      "step": 3576
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953610072308405e-05,
+      "loss": 1.1593,
+      "step": 3577
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535761474020668e-05,
+      "loss": 1.0941,
+      "step": 3578
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953542210390413e-05,
+      "loss": 1.1215,
+      "step": 3579
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9535082612738738e-05,
+      "loss": 1.0705,
+      "step": 3580
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534743000528808e-05,
+      "loss": 1.1022,
+      "step": 3581
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9534403267278647e-05,
+      "loss": 1.0492,
+      "step": 3582
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953406341299257e-05,
+      "loss": 1.2414,
+      "step": 3583
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953372343767489e-05,
+      "loss": 1.0125,
+      "step": 3584
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533383341329922e-05,
+      "loss": 1.1062,
+      "step": 3585
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9533043123961986e-05,
+      "loss": 1.1008,
+      "step": 3586
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.95327027855754e-05,
+      "loss": 1.1404,
+      "step": 3587
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9532362326174484e-05,
+      "loss": 1.0936,
+      "step": 3588
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953202174576356e-05,
+      "loss": 1.1324,
+      "step": 3589
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953168104434695e-05,
+      "loss": 1.108,
+      "step": 3590
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9531340221928985e-05,
+      "loss": 1.1372,
+      "step": 3591
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.9530999278513985e-05,
+      "loss": 1.1163,
+      "step": 3592
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 1.953065821410628e-05,
+      "loss": 1.0847,
+      "step": 3593
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95303170287102e-05,
+      "loss": 1.1158,
+      "step": 3594
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952997572233008e-05,
+      "loss": 1.0572,
+      "step": 3595
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529634294970245e-05,
+      "loss": 1.2107,
+      "step": 3596
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9529292746635034e-05,
+      "loss": 1.0534,
+      "step": 3597
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528951077328787e-05,
+      "loss": 1.2648,
+      "step": 3598
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952860928705583e-05,
+      "loss": 1.1476,
+      "step": 3599
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9528267375820516e-05,
+      "loss": 1.0703,
+      "step": 3600
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527925343627174e-05,
+      "loss": 0.9722,
+      "step": 3601
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527583190480152e-05,
+      "loss": 1.0314,
+      "step": 3602
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9527240916383788e-05,
+      "loss": 1.1404,
+      "step": 3603
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9526898521342432e-05,
+      "loss": 1.1241,
+      "step": 3604
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952655600536043e-05,
+      "loss": 1.1326,
+      "step": 3605
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952621336844213e-05,
+      "loss": 1.0979,
+      "step": 3606
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952587061059188e-05,
+      "loss": 1.1369,
+      "step": 3607
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9525527731814032e-05,
+      "loss": 1.0716,
+      "step": 3608
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952518473211294e-05,
+      "loss": 1.139,
+      "step": 3609
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524841611492958e-05,
+      "loss": 1.0561,
+      "step": 3610
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952449836995844e-05,
+      "loss": 1.1169,
+      "step": 3611
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9524155007513747e-05,
+      "loss": 0.9935,
+      "step": 3612
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952381152416323e-05,
+      "loss": 1.1012,
+      "step": 3613
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523467919911256e-05,
+      "loss": 1.0584,
+      "step": 3614
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9523124194762187e-05,
+      "loss": 1.0333,
+      "step": 3615
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522780348720385e-05,
+      "loss": 1.1535,
+      "step": 3616
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522436381790214e-05,
+      "loss": 0.9958,
+      "step": 3617
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9522092293976044e-05,
+      "loss": 1.0529,
+      "step": 3618
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.952174808528224e-05,
+      "loss": 1.222,
+      "step": 3619
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521403755713167e-05,
+      "loss": 1.0886,
+      "step": 3620
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9521059305273208e-05,
+      "loss": 1.0293,
+      "step": 3621
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520714733966726e-05,
+      "loss": 1.0864,
+      "step": 3622
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95203700417981e-05,
+      "loss": 1.0725,
+      "step": 3623
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9520025228771705e-05,
+      "loss": 1.2468,
+      "step": 3624
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519680294891913e-05,
+      "loss": 1.0153,
+      "step": 3625
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9519335240163108e-05,
+      "loss": 1.1475,
+      "step": 3626
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518990064589674e-05,
+      "loss": 1.1931,
+      "step": 3627
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518644768175986e-05,
+      "loss": 1.1115,
+      "step": 3628
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9518299350926427e-05,
+      "loss": 1.162,
+      "step": 3629
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517953812845386e-05,
+      "loss": 1.0629,
+      "step": 3630
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951760815393725e-05,
+      "loss": 1.0835,
+      "step": 3631
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9517262374206406e-05,
+      "loss": 1.1592,
+      "step": 3632
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516916473657243e-05,
+      "loss": 0.9968,
+      "step": 3633
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9516570452294147e-05,
+      "loss": 1.0617,
+      "step": 3634
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951622431012152e-05,
+      "loss": 1.1451,
+      "step": 3635
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515878047143747e-05,
+      "loss": 1.004,
+      "step": 3636
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951553166336523e-05,
+      "loss": 1.2731,
+      "step": 3637
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9515185158790368e-05,
+      "loss": 1.1604,
+      "step": 3638
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514838533423558e-05,
+      "loss": 1.1116,
+      "step": 3639
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514491787269192e-05,
+      "loss": 1.1403,
+      "step": 3640
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9514144920331682e-05,
+      "loss": 1.1317,
+      "step": 3641
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513797932615426e-05,
+      "loss": 1.1576,
+      "step": 3642
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951345082412483e-05,
+      "loss": 1.1218,
+      "step": 3643
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9513103594864303e-05,
+      "loss": 1.1395,
+      "step": 3644
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512756244838255e-05,
+      "loss": 1.0746,
+      "step": 3645
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512408774051086e-05,
+      "loss": 1.0788,
+      "step": 3646
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9512061182507213e-05,
+      "loss": 1.0897,
+      "step": 3647
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951171347021105e-05,
+      "loss": 1.0707,
+      "step": 3648
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511365637167008e-05,
+      "loss": 1.0302,
+      "step": 3649
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9511017683379504e-05,
+      "loss": 1.2271,
+      "step": 3650
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9510669608852956e-05,
+      "loss": 1.0167,
+      "step": 3651
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.951032141359178e-05,
+      "loss": 1.3536,
+      "step": 3652
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95099730976004e-05,
+      "loss": 1.1488,
+      "step": 3653
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509624660883233e-05,
+      "loss": 1.1304,
+      "step": 3654
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9509276103444706e-05,
+      "loss": 1.1731,
+      "step": 3655
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508927425289245e-05,
+      "loss": 1.1791,
+      "step": 3656
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950857862642127e-05,
+      "loss": 1.1102,
+      "step": 3657
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9508229706845213e-05,
+      "loss": 1.1769,
+      "step": 3658
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507880666565505e-05,
+      "loss": 1.3009,
+      "step": 3659
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507531505586577e-05,
+      "loss": 1.1217,
+      "step": 3660
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9507182223912857e-05,
+      "loss": 1.09,
+      "step": 3661
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506832821548784e-05,
+      "loss": 1.1172,
+      "step": 3662
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950648329849879e-05,
+      "loss": 1.1417,
+      "step": 3663
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9506133654767313e-05,
+      "loss": 1.1987,
+      "step": 3664
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505783890358793e-05,
+      "loss": 1.0873,
+      "step": 3665
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505434005277667e-05,
+      "loss": 1.2214,
+      "step": 3666
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9505083999528382e-05,
+      "loss": 1.1232,
+      "step": 3667
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950473387311538e-05,
+      "loss": 1.0818,
+      "step": 3668
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504383626043098e-05,
+      "loss": 1.1806,
+      "step": 3669
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9504033258315987e-05,
+      "loss": 1.0955,
+      "step": 3670
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.95036827699385e-05,
+      "loss": 1.1819,
+      "step": 3671
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9503332160915082e-05,
+      "loss": 1.1463,
+      "step": 3672
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502981431250183e-05,
+      "loss": 1.1631,
+      "step": 3673
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502630580948257e-05,
+      "loss": 1.082,
+      "step": 3674
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9502279610013756e-05,
+      "loss": 1.0266,
+      "step": 3675
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501928518451137e-05,
+      "loss": 1.1734,
+      "step": 3676
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950157730626486e-05,
+      "loss": 1.0616,
+      "step": 3677
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9501225973459378e-05,
+      "loss": 1.0287,
+      "step": 3678
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.950087452003915e-05,
+      "loss": 1.1017,
+      "step": 3679
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500522946008643e-05,
+      "loss": 1.0933,
+      "step": 3680
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9500171251372322e-05,
+      "loss": 1.1748,
+      "step": 3681
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949981943613464e-05,
+      "loss": 1.1387,
+      "step": 3682
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499467500300076e-05,
+      "loss": 1.0622,
+      "step": 3683
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9499115443873093e-05,
+      "loss": 1.0381,
+      "step": 3684
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949876326685816e-05,
+      "loss": 1.1874,
+      "step": 3685
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498410969259744e-05,
+      "loss": 1.1744,
+      "step": 3686
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9498058551082323e-05,
+      "loss": 1.1293,
+      "step": 3687
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497706012330368e-05,
+      "loss": 0.9241,
+      "step": 3688
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9497353353008356e-05,
+      "loss": 1.1453,
+      "step": 3689
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949700057312076e-05,
+      "loss": 1.1673,
+      "step": 3690
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496647672672066e-05,
+      "loss": 1.1007,
+      "step": 3691
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9496294651666745e-05,
+      "loss": 1.1496,
+      "step": 3692
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495941510109285e-05,
+      "loss": 1.1262,
+      "step": 3693
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9495588248004166e-05,
+      "loss": 1.1169,
+      "step": 3694
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949523486535587e-05,
+      "loss": 1.122,
+      "step": 3695
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494881362168893e-05,
+      "loss": 1.1567,
+      "step": 3696
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949452773844771e-05,
+      "loss": 1.0181,
+      "step": 3697
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9494173994196817e-05,
+      "loss": 1.1654,
+      "step": 3698
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949382012942071e-05,
+      "loss": 1.0997,
+      "step": 3699
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493466144123868e-05,
+      "loss": 1.0737,
+      "step": 3700
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9493112038310796e-05,
+      "loss": 1.0919,
+      "step": 3701
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949275781198598e-05,
+      "loss": 1.1977,
+      "step": 3702
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492403465153925e-05,
+      "loss": 1.1304,
+      "step": 3703
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9492048997819127e-05,
+      "loss": 1.1472,
+      "step": 3704
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.949169440998608e-05,
+      "loss": 1.0024,
+      "step": 3705
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9491339701659292e-05,
+      "loss": 1.0547,
+      "step": 3706
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490984872843267e-05,
+      "loss": 1.0209,
+      "step": 3707
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490629923542505e-05,
+      "loss": 1.1752,
+      "step": 3708
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9490274853761514e-05,
+      "loss": 1.1443,
+      "step": 3709
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9489919663504797e-05,
+      "loss": 1.1243,
+      "step": 3710
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948956435277687e-05,
+      "loss": 1.2406,
+      "step": 3711
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948920892158224e-05,
+      "loss": 1.1311,
+      "step": 3712
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948885336992542e-05,
+      "loss": 1.1333,
+      "step": 3713
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488497697810923e-05,
+      "loss": 1.3208,
+      "step": 3714
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9488141905243267e-05,
+      "loss": 1.087,
+      "step": 3715
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487785992226963e-05,
+      "loss": 1.1326,
+      "step": 3716
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487429958766532e-05,
+      "loss": 1.1202,
+      "step": 3717
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9487073804866493e-05,
+      "loss": 1.1711,
+      "step": 3718
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948671753053137e-05,
+      "loss": 1.0668,
+      "step": 3719
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9486361135765684e-05,
+      "loss": 1.1672,
+      "step": 3720
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948600462057396e-05,
+      "loss": 1.0803,
+      "step": 3721
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948564798496072e-05,
+      "loss": 1.2536,
+      "step": 3722
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9485291228930498e-05,
+      "loss": 1.1375,
+      "step": 3723
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484934352487818e-05,
+      "loss": 1.1346,
+      "step": 3724
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484577355637212e-05,
+      "loss": 1.0573,
+      "step": 3725
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9484220238383207e-05,
+      "loss": 1.0433,
+      "step": 3726
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483863000730346e-05,
+      "loss": 1.2188,
+      "step": 3727
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9483505642683156e-05,
+      "loss": 1.1987,
+      "step": 3728
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948314816424618e-05,
+      "loss": 1.0524,
+      "step": 3729
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482790565423954e-05,
+      "loss": 1.0669,
+      "step": 3730
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482432846221015e-05,
+      "loss": 1.1444,
+      "step": 3731
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9482075006641903e-05,
+      "loss": 1.198,
+      "step": 3732
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481717046691163e-05,
+      "loss": 1.1249,
+      "step": 3733
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948135896637334e-05,
+      "loss": 1.0583,
+      "step": 3734
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9481000765692982e-05,
+      "loss": 1.0481,
+      "step": 3735
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948064244465463e-05,
+      "loss": 1.0686,
+      "step": 3736
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.948028400326284e-05,
+      "loss": 1.183,
+      "step": 3737
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479925441522153e-05,
+      "loss": 1.1628,
+      "step": 3738
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479566759437126e-05,
+      "loss": 1.1823,
+      "step": 3739
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9479207957012316e-05,
+      "loss": 1.2897,
+      "step": 3740
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478849034252276e-05,
+      "loss": 1.0974,
+      "step": 3741
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9478489991161556e-05,
+      "loss": 1.2313,
+      "step": 3742
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947813082774472e-05,
+      "loss": 1.083,
+      "step": 3743
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477771544006327e-05,
+      "loss": 1.1218,
+      "step": 3744
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9477412139950938e-05,
+      "loss": 1.1162,
+      "step": 3745
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947705261558311e-05,
+      "loss": 1.0455,
+      "step": 3746
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9476692970907417e-05,
+      "loss": 0.9957,
+      "step": 3747
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947633320592842e-05,
+      "loss": 1.1065,
+      "step": 3748
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947597332065068e-05,
+      "loss": 1.0558,
+      "step": 3749
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475613315078773e-05,
+      "loss": 1.2,
+      "step": 3750
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9475253189217268e-05,
+      "loss": 1.1372,
+      "step": 3751
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474892943070736e-05,
+      "loss": 1.1406,
+      "step": 3752
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474532576643746e-05,
+      "loss": 1.0664,
+      "step": 3753
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9474172089940883e-05,
+      "loss": 1.0465,
+      "step": 3754
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947381148296671e-05,
+      "loss": 0.9311,
+      "step": 3755
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9473450755725817e-05,
+      "loss": 0.9247,
+      "step": 3756
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947308990822278e-05,
+      "loss": 1.0731,
+      "step": 3757
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947272894046217e-05,
+      "loss": 1.1152,
+      "step": 3758
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.947236785244858e-05,
+      "loss": 1.1523,
+      "step": 3759
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9472006644186594e-05,
+      "loss": 1.1323,
+      "step": 3760
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471645315680795e-05,
+      "loss": 1.2179,
+      "step": 3761
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9471283866935763e-05,
+      "loss": 1.0705,
+      "step": 3762
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470922297956098e-05,
+      "loss": 1.1345,
+      "step": 3763
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470560608746384e-05,
+      "loss": 1.161,
+      "step": 3764
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9470198799311213e-05,
+      "loss": 1.0393,
+      "step": 3765
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469836869655178e-05,
+      "loss": 1.0626,
+      "step": 3766
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469474819782874e-05,
+      "loss": 1.1055,
+      "step": 3767
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9469112649698898e-05,
+      "loss": 1.0608,
+      "step": 3768
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468750359407845e-05,
+      "loss": 1.1982,
+      "step": 3769
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946838794891432e-05,
+      "loss": 1.2124,
+      "step": 3770
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9468025418222916e-05,
+      "loss": 1.0862,
+      "step": 3771
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946766276733824e-05,
+      "loss": 1.1628,
+      "step": 3772
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9467299996264892e-05,
+      "loss": 1.0753,
+      "step": 3773
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466937105007484e-05,
+      "loss": 1.0553,
+      "step": 3774
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466574093570615e-05,
+      "loss": 1.1913,
+      "step": 3775
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9466210961958897e-05,
+      "loss": 1.0605,
+      "step": 3776
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946584771017694e-05,
+      "loss": 1.0353,
+      "step": 3777
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465484338229358e-05,
+      "loss": 1.1425,
+      "step": 3778
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9465120846120756e-05,
+      "loss": 1.1304,
+      "step": 3779
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464757233855757e-05,
+      "loss": 1.069,
+      "step": 3780
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464393501438972e-05,
+      "loss": 1.1597,
+      "step": 3781
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9464029648875022e-05,
+      "loss": 0.9828,
+      "step": 3782
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9463665676168518e-05,
+      "loss": 1.1021,
+      "step": 3783
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946330158332409e-05,
+      "loss": 1.0523,
+      "step": 3784
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462937370346356e-05,
+      "loss": 1.066,
+      "step": 3785
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462573037239937e-05,
+      "loss": 1.0899,
+      "step": 3786
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9462208584009464e-05,
+      "loss": 1.0551,
+      "step": 3787
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461844010659558e-05,
+      "loss": 1.088,
+      "step": 3788
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946147931719485e-05,
+      "loss": 1.0959,
+      "step": 3789
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9461114503619974e-05,
+      "loss": 1.1317,
+      "step": 3790
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946074956993955e-05,
+      "loss": 0.9675,
+      "step": 3791
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.946038451615822e-05,
+      "loss": 0.9832,
+      "step": 3792
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9460019342280615e-05,
+      "loss": 1.0971,
+      "step": 3793
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945965404831137e-05,
+      "loss": 1.1618,
+      "step": 3794
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9459288634255126e-05,
+      "loss": 1.1408,
+      "step": 3795
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458923100116517e-05,
+      "loss": 1.2269,
+      "step": 3796
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458557445900186e-05,
+      "loss": 1.0463,
+      "step": 3797
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9458191671610772e-05,
+      "loss": 1.0508,
+      "step": 3798
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457825777252927e-05,
+      "loss": 1.1401,
+      "step": 3799
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457459762831285e-05,
+      "loss": 1.1572,
+      "step": 3800
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9457093628350495e-05,
+      "loss": 1.2206,
+      "step": 3801
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456727373815213e-05,
+      "loss": 1.0746,
+      "step": 3802
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9456360999230076e-05,
+      "loss": 1.1058,
+      "step": 3803
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9455994504599748e-05,
+      "loss": 1.1782,
+      "step": 3804
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945562788992887e-05,
+      "loss": 1.1528,
+      "step": 3805
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94552611552221e-05,
+      "loss": 1.1675,
+      "step": 3806
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94548943004841e-05,
+      "loss": 1.0731,
+      "step": 3807
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945452732571952e-05,
+      "loss": 1.0245,
+      "step": 3808
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9454160230933018e-05,
+      "loss": 1.1104,
+      "step": 3809
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945379301612926e-05,
+      "loss": 1.1547,
+      "step": 3810
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94534256813129e-05,
+      "loss": 1.2335,
+      "step": 3811
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9453058226488604e-05,
+      "loss": 1.0977,
+      "step": 3812
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452690651661044e-05,
+      "loss": 1.217,
+      "step": 3813
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9452322956834875e-05,
+      "loss": 1.0226,
+      "step": 3814
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451955142014773e-05,
+      "loss": 1.0728,
+      "step": 3815
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9451587207205404e-05,
+      "loss": 1.1164,
+      "step": 3816
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945121915241144e-05,
+      "loss": 1.1551,
+      "step": 3817
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450850977637547e-05,
+      "loss": 1.0423,
+      "step": 3818
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.945048268288841e-05,
+      "loss": 1.042,
+      "step": 3819
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9450114268168697e-05,
+      "loss": 1.0461,
+      "step": 3820
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449745733483086e-05,
+      "loss": 1.0489,
+      "step": 3821
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449377078836254e-05,
+      "loss": 0.9589,
+      "step": 3822
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9449008304232888e-05,
+      "loss": 1.1913,
+      "step": 3823
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944863940967766e-05,
+      "loss": 1.0054,
+      "step": 3824
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944827039517526e-05,
+      "loss": 1.0895,
+      "step": 3825
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447901260730368e-05,
+      "loss": 1.217,
+      "step": 3826
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9447532006347676e-05,
+      "loss": 1.118,
+      "step": 3827
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944716263203186e-05,
+      "loss": 1.2219,
+      "step": 3828
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446793137787624e-05,
+      "loss": 1.1774,
+      "step": 3829
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446423523619648e-05,
+      "loss": 1.0134,
+      "step": 3830
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9446053789532625e-05,
+      "loss": 1.2197,
+      "step": 3831
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944568393553125e-05,
+      "loss": 1.1712,
+      "step": 3832
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9445313961620226e-05,
+      "loss": 1.0706,
+      "step": 3833
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9444943867804237e-05,
+      "loss": 1.0583,
+      "step": 3834
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944457365408799e-05,
+      "loss": 1.0548,
+      "step": 3835
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944420332047618e-05,
+      "loss": 1.0974,
+      "step": 3836
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944383286697351e-05,
+      "loss": 1.1228,
+      "step": 3837
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443462293584683e-05,
+      "loss": 1.1069,
+      "step": 3838
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9443091600314404e-05,
+      "loss": 1.1657,
+      "step": 3839
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442720787167377e-05,
+      "loss": 0.9948,
+      "step": 3840
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9442349854148308e-05,
+      "loss": 1.1947,
+      "step": 3841
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944197880126191e-05,
+      "loss": 1.1814,
+      "step": 3842
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944160762851289e-05,
+      "loss": 1.0855,
+      "step": 3843
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9441236335905963e-05,
+      "loss": 1.138,
+      "step": 3844
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440864923445843e-05,
+      "loss": 1.1054,
+      "step": 3845
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9440493391137242e-05,
+      "loss": 1.1092,
+      "step": 3846
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.944012173898487e-05,
+      "loss": 1.1873,
+      "step": 3847
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439749966993464e-05,
+      "loss": 1.0803,
+      "step": 3848
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439378075167722e-05,
+      "loss": 1.0042,
+      "step": 3849
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9439006063512378e-05,
+      "loss": 1.0029,
+      "step": 3850
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438633932032152e-05,
+      "loss": 1.0241,
+      "step": 3851
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9438261680731763e-05,
+      "loss": 1.2528,
+      "step": 3852
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943788930961595e-05,
+      "loss": 1.1201,
+      "step": 3853
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437516818689424e-05,
+      "loss": 1.1814,
+      "step": 3854
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9437144207956923e-05,
+      "loss": 1.0081,
+      "step": 3855
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436771477423172e-05,
+      "loss": 1.0878,
+      "step": 3856
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436398627092908e-05,
+      "loss": 1.0134,
+      "step": 3857
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9436025656970866e-05,
+      "loss": 1.2091,
+      "step": 3858
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435652567061768e-05,
+      "loss": 1.0154,
+      "step": 3859
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9435279357370366e-05,
+      "loss": 1.0817,
+      "step": 3860
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434906027901386e-05,
+      "loss": 1.0941,
+      "step": 3861
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9434532578659576e-05,
+      "loss": 0.9615,
+      "step": 3862
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943415900964967e-05,
+      "loss": 1.1289,
+      "step": 3863
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433785320876412e-05,
+      "loss": 1.066,
+      "step": 3864
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9433411512344547e-05,
+      "loss": 1.1438,
+      "step": 3865
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943303758405882e-05,
+      "loss": 1.1259,
+      "step": 3866
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943266353602398e-05,
+      "loss": 1.1744,
+      "step": 3867
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9432289368244772e-05,
+      "loss": 1.0522,
+      "step": 3868
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943191508072595e-05,
+      "loss": 1.0904,
+      "step": 3869
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943154067347226e-05,
+      "loss": 1.0904,
+      "step": 3870
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.943116614648846e-05,
+      "loss": 1.1111,
+      "step": 3871
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.94307914997793e-05,
+      "loss": 1.1394,
+      "step": 3872
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430416733349538e-05,
+      "loss": 1.1252,
+      "step": 3873
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9430041847203937e-05,
+      "loss": 1.1413,
+      "step": 3874
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429666841347244e-05,
+      "loss": 1.1823,
+      "step": 3875
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9429291715784232e-05,
+      "loss": 1.1484,
+      "step": 3876
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428916470519653e-05,
+      "loss": 1.0523,
+      "step": 3877
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.9428541105558277e-05,
+      "loss": 1.1083,
+      "step": 3878
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942816562090487e-05,
+      "loss": 1.1058,
+      "step": 3879
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 1.942779001656419e-05,
+      "loss": 1.0603,
+      "step": 3880
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427414292541017e-05,
+      "loss": 1.0337,
+      "step": 3881
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9427038448840112e-05,
+      "loss": 1.1261,
+      "step": 3882
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942666248546625e-05,
+      "loss": 1.2198,
+      "step": 3883
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94262864024242e-05,
+      "loss": 1.0853,
+      "step": 3884
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942591019971874e-05,
+      "loss": 1.0968,
+      "step": 3885
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425533877354646e-05,
+      "loss": 1.0536,
+      "step": 3886
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9425157435336694e-05,
+      "loss": 1.177,
+      "step": 3887
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424780873669663e-05,
+      "loss": 0.9547,
+      "step": 3888
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424404192358334e-05,
+      "loss": 1.1893,
+      "step": 3889
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9424027391407484e-05,
+      "loss": 1.0967,
+      "step": 3890
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423650470821903e-05,
+      "loss": 1.1566,
+      "step": 3891
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9423273430606377e-05,
+      "loss": 1.0518,
+      "step": 3892
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942289627076568e-05,
+      "loss": 1.1685,
+      "step": 3893
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422518991304615e-05,
+      "loss": 1.2088,
+      "step": 3894
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9422141592227965e-05,
+      "loss": 1.0931,
+      "step": 3895
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9421764073540518e-05,
+      "loss": 1.1276,
+      "step": 3896
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942138643524707e-05,
+      "loss": 1.08,
+      "step": 3897
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.942100867735241e-05,
+      "loss": 1.2015,
+      "step": 3898
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420630799861344e-05,
+      "loss": 1.141,
+      "step": 3899
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9420252802778664e-05,
+      "loss": 1.231,
+      "step": 3900
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419874686109165e-05,
+      "loss": 1.1222,
+      "step": 3901
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9419496449857648e-05,
+      "loss": 0.9487,
+      "step": 3902
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941911809402892e-05,
+      "loss": 1.1682,
+      "step": 3903
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418739618627773e-05,
+      "loss": 1.2051,
+      "step": 3904
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9418361023659026e-05,
+      "loss": 1.0007,
+      "step": 3905
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417982309127473e-05,
+      "loss": 1.096,
+      "step": 3906
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417603475037927e-05,
+      "loss": 1.078,
+      "step": 3907
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9417224521395196e-05,
+      "loss": 1.0424,
+      "step": 3908
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416845448204096e-05,
+      "loss": 1.009,
+      "step": 3909
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9416466255469433e-05,
+      "loss": 1.1329,
+      "step": 3910
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941608694319602e-05,
+      "loss": 0.9599,
+      "step": 3911
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415707511388678e-05,
+      "loss": 0.9849,
+      "step": 3912
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9415327960052215e-05,
+      "loss": 1.1104,
+      "step": 3913
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941494828919146e-05,
+      "loss": 1.1073,
+      "step": 3914
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9414568498811222e-05,
+      "loss": 1.1485,
+      "step": 3915
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941418858891633e-05,
+      "loss": 1.1235,
+      "step": 3916
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413808559511607e-05,
+      "loss": 0.9664,
+      "step": 3917
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941342841060187e-05,
+      "loss": 1.0771,
+      "step": 3918
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9413048142191956e-05,
+      "loss": 1.1308,
+      "step": 3919
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9412667754286683e-05,
+      "loss": 1.1354,
+      "step": 3920
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941228724689088e-05,
+      "loss": 1.1241,
+      "step": 3921
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9411906620009384e-05,
+      "loss": 1.1629,
+      "step": 3922
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941152587364702e-05,
+      "loss": 1.0832,
+      "step": 3923
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941114500780863e-05,
+      "loss": 0.9785,
+      "step": 3924
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410764022499037e-05,
+      "loss": 1.2046,
+      "step": 3925
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.941038291772309e-05,
+      "loss": 0.9556,
+      "step": 3926
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9410001693485618e-05,
+      "loss": 1.1042,
+      "step": 3927
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409620349791464e-05,
+      "loss": 1.115,
+      "step": 3928
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9409238886645467e-05,
+      "loss": 1.1168,
+      "step": 3929
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408857304052474e-05,
+      "loss": 1.0657,
+      "step": 3930
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408475602017328e-05,
+      "loss": 1.0168,
+      "step": 3931
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9408093780544868e-05,
+      "loss": 1.2328,
+      "step": 3932
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940771183963995e-05,
+      "loss": 1.1924,
+      "step": 3933
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9407329779307414e-05,
+      "loss": 1.1402,
+      "step": 3934
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940694759955212e-05,
+      "loss": 1.1943,
+      "step": 3935
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406565300378912e-05,
+      "loss": 1.1844,
+      "step": 3936
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9406182881792643e-05,
+      "loss": 1.1635,
+      "step": 3937
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405800343798174e-05,
+      "loss": 1.1161,
+      "step": 3938
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405417686400354e-05,
+      "loss": 1.1225,
+      "step": 3939
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9405034909604045e-05,
+      "loss": 1.1477,
+      "step": 3940
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9404652013414108e-05,
+      "loss": 1.1481,
+      "step": 3941
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94042689978354e-05,
+      "loss": 0.9424,
+      "step": 3942
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940388586287278e-05,
+      "loss": 1.0022,
+      "step": 3943
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940350260853112e-05,
+      "loss": 1.1246,
+      "step": 3944
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940311923481528e-05,
+      "loss": 1.0414,
+      "step": 3945
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940273574173013e-05,
+      "loss": 1.1119,
+      "step": 3946
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9402352129280532e-05,
+      "loss": 1.1695,
+      "step": 3947
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9401968397471363e-05,
+      "loss": 1.1164,
+      "step": 3948
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940158454630749e-05,
+      "loss": 1.1387,
+      "step": 3949
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940120057579379e-05,
+      "loss": 1.1226,
+      "step": 3950
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9400816485935134e-05,
+      "loss": 1.086,
+      "step": 3951
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.94004322767364e-05,
+      "loss": 0.9412,
+      "step": 3952
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.940004794820246e-05,
+      "loss": 1.1622,
+      "step": 3953
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93996635003382e-05,
+      "loss": 1.0417,
+      "step": 3954
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9399278933148496e-05,
+      "loss": 1.1238,
+      "step": 3955
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398894246638232e-05,
+      "loss": 1.2336,
+      "step": 3956
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939850944081229e-05,
+      "loss": 1.0371,
+      "step": 3957
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9398124515675554e-05,
+      "loss": 1.0777,
+      "step": 3958
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939773947123292e-05,
+      "loss": 1.0938,
+      "step": 3959
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939735430748926e-05,
+      "loss": 1.0007,
+      "step": 3960
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396969024449473e-05,
+      "loss": 1.1122,
+      "step": 3961
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396583622118452e-05,
+      "loss": 1.0097,
+      "step": 3962
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9396198100501086e-05,
+      "loss": 1.2257,
+      "step": 3963
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395812459602267e-05,
+      "loss": 1.1257,
+      "step": 3964
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395426699426894e-05,
+      "loss": 1.2168,
+      "step": 3965
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9395040819979863e-05,
+      "loss": 1.2862,
+      "step": 3966
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394654821266076e-05,
+      "loss": 1.0081,
+      "step": 3967
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9394268703290424e-05,
+      "loss": 0.9204,
+      "step": 3968
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393882466057817e-05,
+      "loss": 0.9476,
+      "step": 3969
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393496109573157e-05,
+      "loss": 0.9859,
+      "step": 3970
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9393109633841348e-05,
+      "loss": 1.1626,
+      "step": 3971
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939272303886729e-05,
+      "loss": 1.0355,
+      "step": 3972
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93923363246559e-05,
+      "loss": 1.1175,
+      "step": 3973
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391949491212086e-05,
+      "loss": 1.0678,
+      "step": 3974
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.939156253854075e-05,
+      "loss": 1.0975,
+      "step": 3975
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9391175466646815e-05,
+      "loss": 1.1185,
+      "step": 3976
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390788275535187e-05,
+      "loss": 1.0751,
+      "step": 3977
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390400965210787e-05,
+      "loss": 0.9921,
+      "step": 3978
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9390013535678528e-05,
+      "loss": 1.2251,
+      "step": 3979
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938962598694333e-05,
+      "loss": 1.1773,
+      "step": 3980
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938923831901011e-05,
+      "loss": 1.1939,
+      "step": 3981
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388850531883792e-05,
+      "loss": 1.1723,
+      "step": 3982
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93884626255693e-05,
+      "loss": 1.1636,
+      "step": 3983
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9388074600071554e-05,
+      "loss": 1.1115,
+      "step": 3984
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387686455395486e-05,
+      "loss": 1.0458,
+      "step": 3985
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9387298191546016e-05,
+      "loss": 1.1882,
+      "step": 3986
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386909808528077e-05,
+      "loss": 1.0125,
+      "step": 3987
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93865213063466e-05,
+      "loss": 1.0956,
+      "step": 3988
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9386132685006517e-05,
+      "loss": 1.0561,
+      "step": 3989
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938574394451276e-05,
+      "loss": 1.1455,
+      "step": 3990
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9385355084870265e-05,
+      "loss": 1.1234,
+      "step": 3991
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384966106083967e-05,
+      "loss": 1.1017,
+      "step": 3992
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384577008158802e-05,
+      "loss": 1.1252,
+      "step": 3993
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9384187791099717e-05,
+      "loss": 0.9936,
+      "step": 3994
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383798454911645e-05,
+      "loss": 1.0486,
+      "step": 3995
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383408999599532e-05,
+      "loss": 1.0917,
+      "step": 3996
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9383019425168323e-05,
+      "loss": 1.1066,
+      "step": 3997
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9382629731622956e-05,
+      "loss": 1.0004,
+      "step": 3998
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938223991896839e-05,
+      "loss": 1.2204,
+      "step": 3999
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938184998720957e-05,
+      "loss": 1.1262,
+      "step": 4000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381459936351435e-05,
+      "loss": 1.0319,
+      "step": 4001
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9381069766398953e-05,
+      "loss": 1.1597,
+      "step": 4002
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9380679477357067e-05,
+      "loss": 0.9777,
+      "step": 4003
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.938028906923073e-05,
+      "loss": 1.2099,
+      "step": 4004
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937989854202491e-05,
+      "loss": 1.1363,
+      "step": 4005
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937950789574455e-05,
+      "loss": 1.1332,
+      "step": 4006
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937911713039462e-05,
+      "loss": 1.1243,
+      "step": 4007
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378726245980078e-05,
+      "loss": 1.1174,
+      "step": 4008
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9378335242505878e-05,
+      "loss": 1.1042,
+      "step": 4009
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377944119976996e-05,
+      "loss": 0.9899,
+      "step": 4010
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377552878398385e-05,
+      "loss": 1.1013,
+      "step": 4011
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9377161517775023e-05,
+      "loss": 1.202,
+      "step": 4012
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376770038111873e-05,
+      "loss": 1.0582,
+      "step": 4013
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9376378439413903e-05,
+      "loss": 1.077,
+      "step": 4014
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375986721686085e-05,
+      "loss": 1.2128,
+      "step": 4015
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375594884933394e-05,
+      "loss": 1.1445,
+      "step": 4016
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9375202929160807e-05,
+      "loss": 1.0086,
+      "step": 4017
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937481085437329e-05,
+      "loss": 1.1597,
+      "step": 4018
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937441866057583e-05,
+      "loss": 1.1204,
+      "step": 4019
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93740263477734e-05,
+      "loss": 1.1217,
+      "step": 4020
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373633915970984e-05,
+      "loss": 1.1211,
+      "step": 4021
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9373241365173557e-05,
+      "loss": 1.1229,
+      "step": 4022
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372848695386114e-05,
+      "loss": 1.2221,
+      "step": 4023
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937245590661363e-05,
+      "loss": 1.1354,
+      "step": 4024
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9372062998861092e-05,
+      "loss": 1.1059,
+      "step": 4025
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937166997213349e-05,
+      "loss": 1.0163,
+      "step": 4026
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9371276826435816e-05,
+      "loss": 1.1532,
+      "step": 4027
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937088356177306e-05,
+      "loss": 1.1041,
+      "step": 4028
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937049017815021e-05,
+      "loss": 1.1825,
+      "step": 4029
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.937009667557226e-05,
+      "loss": 1.0875,
+      "step": 4030
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369703054044212e-05,
+      "loss": 1.1851,
+      "step": 4031
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9369309313571057e-05,
+      "loss": 1.137,
+      "step": 4032
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368915454157793e-05,
+      "loss": 1.0981,
+      "step": 4033
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368521475809428e-05,
+      "loss": 1.226,
+      "step": 4034
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9368127378530953e-05,
+      "loss": 1.075,
+      "step": 4035
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9367733162327377e-05,
+      "loss": 1.1298,
+      "step": 4036
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93673388272037e-05,
+      "loss": 1.2696,
+      "step": 4037
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366944373164935e-05,
+      "loss": 1.0346,
+      "step": 4038
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366549800216083e-05,
+      "loss": 1.1552,
+      "step": 4039
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9366155108362156e-05,
+      "loss": 0.9555,
+      "step": 4040
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365760297608164e-05,
+      "loss": 0.9804,
+      "step": 4041
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9365365367959115e-05,
+      "loss": 1.0876,
+      "step": 4042
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936497031942003e-05,
+      "loss": 1.0886,
+      "step": 4043
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936457515199592e-05,
+      "loss": 0.9622,
+      "step": 4044
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9364179865691805e-05,
+      "loss": 1.1062,
+      "step": 4045
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363784460512693e-05,
+      "loss": 1.1971,
+      "step": 4046
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9363388936463613e-05,
+      "loss": 1.2474,
+      "step": 4047
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362993293549586e-05,
+      "loss": 1.0992,
+      "step": 4048
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362597531775628e-05,
+      "loss": 1.1764,
+      "step": 4049
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9362201651146768e-05,
+      "loss": 1.225,
+      "step": 4050
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361805651668032e-05,
+      "loss": 1.2594,
+      "step": 4051
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361409533344445e-05,
+      "loss": 1.3104,
+      "step": 4052
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9361013296181034e-05,
+      "loss": 1.0567,
+      "step": 4053
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936061694018283e-05,
+      "loss": 1.1483,
+      "step": 4054
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.936022046535487e-05,
+      "loss": 1.0577,
+      "step": 4055
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359823871702182e-05,
+      "loss": 1.1212,
+      "step": 4056
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93594271592298e-05,
+      "loss": 1.106,
+      "step": 4057
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9359030327942762e-05,
+      "loss": 1.1243,
+      "step": 4058
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9358633377846105e-05,
+      "loss": 1.053,
+      "step": 4059
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935823630894487e-05,
+      "loss": 1.1064,
+      "step": 4060
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93578391212441e-05,
+      "loss": 1.1046,
+      "step": 4061
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9357441814748826e-05,
+      "loss": 1.152,
+      "step": 4062
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93570443894641e-05,
+      "loss": 0.9857,
+      "step": 4063
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356646845394968e-05,
+      "loss": 0.9795,
+      "step": 4064
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9356249182546473e-05,
+      "loss": 1.2636,
+      "step": 4065
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935585140092367e-05,
+      "loss": 0.9461,
+      "step": 4066
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.93554535005316e-05,
+      "loss": 1.0097,
+      "step": 4067
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9355055481375316e-05,
+      "loss": 1.0095,
+      "step": 4068
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9354657343459875e-05,
+      "loss": 1.1475,
+      "step": 4069
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935425908679033e-05,
+      "loss": 1.0887,
+      "step": 4070
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353860711371735e-05,
+      "loss": 1.109,
+      "step": 4071
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353462217209148e-05,
+      "loss": 1.1251,
+      "step": 4072
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9353063604307625e-05,
+      "loss": 0.9253,
+      "step": 4073
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352664872672228e-05,
+      "loss": 1.2497,
+      "step": 4074
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9352266022308024e-05,
+      "loss": 1.1908,
+      "step": 4075
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935186705322007e-05,
+      "loss": 1.1017,
+      "step": 4076
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351467965413427e-05,
+      "loss": 1.0921,
+      "step": 4077
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9351068758893173e-05,
+      "loss": 1.0462,
+      "step": 4078
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.935066943366437e-05,
+      "loss": 1.1467,
+      "step": 4079
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9350269989732085e-05,
+      "loss": 1.1559,
+      "step": 4080
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349870427101392e-05,
+      "loss": 1.1602,
+      "step": 4081
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934947074577736e-05,
+      "loss": 1.1211,
+      "step": 4082
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9349070945765068e-05,
+      "loss": 1.0704,
+      "step": 4083
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348671027069585e-05,
+      "loss": 1.1722,
+      "step": 4084
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9348270989695993e-05,
+      "loss": 1.0489,
+      "step": 4085
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934787083364937e-05,
+      "loss": 1.2347,
+      "step": 4086
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934747055893479e-05,
+      "loss": 1.0955,
+      "step": 4087
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934707016555734e-05,
+      "loss": 1.0921,
+      "step": 4088
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346669653522105e-05,
+      "loss": 1.1144,
+      "step": 4089
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9346269022834158e-05,
+      "loss": 1.1194,
+      "step": 4090
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345868273498597e-05,
+      "loss": 1.1437,
+      "step": 4091
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9345467405520506e-05,
+      "loss": 1.1423,
+      "step": 4092
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934506641890497e-05,
+      "loss": 1.0957,
+      "step": 4093
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934466531365709e-05,
+      "loss": 1.0896,
+      "step": 4094
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934426408978194e-05,
+      "loss": 0.9608,
+      "step": 4095
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934386274728463e-05,
+      "loss": 1.038,
+      "step": 4096
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343461286170246e-05,
+      "loss": 1.2486,
+      "step": 4097
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9343059706443888e-05,
+      "loss": 1.1639,
+      "step": 4098
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934265800811065e-05,
+      "loss": 1.0555,
+      "step": 4099
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9342256191175637e-05,
+      "loss": 1.1534,
+      "step": 4100
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934185425564395e-05,
+      "loss": 1.1101,
+      "step": 4101
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.934145220152068e-05,
+      "loss": 1.0313,
+      "step": 4102
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9341050028810945e-05,
+      "loss": 0.9085,
+      "step": 4103
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340647737519845e-05,
+      "loss": 1.2605,
+      "step": 4104
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9340245327652484e-05,
+      "loss": 1.0743,
+      "step": 4105
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933984279921398e-05,
+      "loss": 1.0857,
+      "step": 4106
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933944015220943e-05,
+      "loss": 1.1208,
+      "step": 4107
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9339037386643954e-05,
+      "loss": 1.0648,
+      "step": 4108
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933863450252266e-05,
+      "loss": 1.0592,
+      "step": 4109
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933823149985067e-05,
+      "loss": 1.1606,
+      "step": 4110
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933782837863309e-05,
+      "loss": 1.1049,
+      "step": 4111
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337425138875047e-05,
+      "loss": 1.1295,
+      "step": 4112
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9337021780581653e-05,
+      "loss": 1.1775,
+      "step": 4113
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336618303758032e-05,
+      "loss": 1.2116,
+      "step": 4114
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9336214708409305e-05,
+      "loss": 1.1525,
+      "step": 4115
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335810994540597e-05,
+      "loss": 1.0195,
+      "step": 4116
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9335407162157028e-05,
+      "loss": 1.0546,
+      "step": 4117
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933500321126373e-05,
+      "loss": 1.1668,
+      "step": 4118
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933459914186583e-05,
+      "loss": 1.2377,
+      "step": 4119
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9334194953968457e-05,
+      "loss": 1.1058,
+      "step": 4120
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333790647576743e-05,
+      "loss": 1.1051,
+      "step": 4121
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9333386222695816e-05,
+      "loss": 0.9686,
+      "step": 4122
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332981679330816e-05,
+      "loss": 1.1301,
+      "step": 4123
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332577017486872e-05,
+      "loss": 1.0611,
+      "step": 4124
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9332172237169128e-05,
+      "loss": 1.1482,
+      "step": 4125
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933176733838272e-05,
+      "loss": 1.0995,
+      "step": 4126
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9331362321132784e-05,
+      "loss": 1.0642,
+      "step": 4127
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330957185424468e-05,
+      "loss": 1.1102,
+      "step": 4128
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.933055193126291e-05,
+      "loss": 1.0588,
+      "step": 4129
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9330146558653258e-05,
+      "loss": 1.1526,
+      "step": 4130
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329741067600657e-05,
+      "loss": 1.1252,
+      "step": 4131
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9329335458110252e-05,
+      "loss": 1.0655,
+      "step": 4132
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328929730187195e-05,
+      "loss": 1.0897,
+      "step": 4133
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9328523883836636e-05,
+      "loss": 1.1088,
+      "step": 4134
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932811791906373e-05,
+      "loss": 1.023,
+      "step": 4135
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9327711835873625e-05,
+      "loss": 0.9918,
+      "step": 4136
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932730563427148e-05,
+      "loss": 0.9509,
+      "step": 4137
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326899314262447e-05,
+      "loss": 1.0352,
+      "step": 4138
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932649287585169e-05,
+      "loss": 1.153,
+      "step": 4139
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9326086319044365e-05,
+      "loss": 0.9515,
+      "step": 4140
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9325679643845633e-05,
+      "loss": 1.1974,
+      "step": 4141
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932527285026066e-05,
+      "loss": 1.0658,
+      "step": 4142
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324865938294604e-05,
+      "loss": 1.0138,
+      "step": 4143
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932445890795264e-05,
+      "loss": 1.1766,
+      "step": 4144
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9324051759239927e-05,
+      "loss": 1.2094,
+      "step": 4145
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323644492161633e-05,
+      "loss": 1.0667,
+      "step": 4146
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9323237106722935e-05,
+      "loss": 1.1248,
+      "step": 4147
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322829602929e-05,
+      "loss": 1.1406,
+      "step": 4148
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322421980784997e-05,
+      "loss": 1.05,
+      "step": 4149
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9322014240296112e-05,
+      "loss": 1.1635,
+      "step": 4150
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932160638146751e-05,
+      "loss": 0.9851,
+      "step": 4151
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9321198404304376e-05,
+      "loss": 0.9794,
+      "step": 4152
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9320790308811884e-05,
+      "loss": 1.007,
+      "step": 4153
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.932038209499522e-05,
+      "loss": 0.9679,
+      "step": 4154
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931997376285956e-05,
+      "loss": 1.097,
+      "step": 4155
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319565312410093e-05,
+      "loss": 1.2686,
+      "step": 4156
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9319156743652003e-05,
+      "loss": 1.0536,
+      "step": 4157
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318748056590474e-05,
+      "loss": 1.0303,
+      "step": 4158
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9318339251230695e-05,
+      "loss": 1.1896,
+      "step": 4159
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317930327577855e-05,
+      "loss": 1.1919,
+      "step": 4160
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931752128563715e-05,
+      "loss": 1.1279,
+      "step": 4161
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9317112125413766e-05,
+      "loss": 1.1138,
+      "step": 4162
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9316702846912902e-05,
+      "loss": 1.2455,
+      "step": 4163
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931629345013975e-05,
+      "loss": 1.1218,
+      "step": 4164
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931588393509951e-05,
+      "loss": 0.9927,
+      "step": 4165
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9315474301797382e-05,
+      "loss": 1.1451,
+      "step": 4166
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.931506455023856e-05,
+      "loss": 0.9991,
+      "step": 4167
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 1.9314654680428252e-05,
+      "loss": 1.0262,
+      "step": 4168
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931424469237166e-05,
+      "loss": 1.1477,
+      "step": 4169
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313834586073986e-05,
+      "loss": 0.9837,
+      "step": 4170
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313424361540434e-05,
+      "loss": 1.0335,
+      "step": 4171
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9313014018776218e-05,
+      "loss": 1.1378,
+      "step": 4172
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312603557786548e-05,
+      "loss": 0.9842,
+      "step": 4173
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9312192978576625e-05,
+      "loss": 1.1731,
+      "step": 4174
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931178228115167e-05,
+      "loss": 1.1411,
+      "step": 4175
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9311371465516894e-05,
+      "loss": 1.1171,
+      "step": 4176
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310960531677508e-05,
+      "loss": 1.1268,
+      "step": 4177
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9310549479638737e-05,
+      "loss": 1.1281,
+      "step": 4178
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.931013830940579e-05,
+      "loss": 1.0606,
+      "step": 4179
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9309727020983896e-05,
+      "loss": 1.2313,
+      "step": 4180
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930931561437827e-05,
+      "loss": 1.1356,
+      "step": 4181
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930890408959414e-05,
+      "loss": 1.056,
+      "step": 4182
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930849244663672e-05,
+      "loss": 1.1229,
+      "step": 4183
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9308080685511246e-05,
+      "loss": 0.9899,
+      "step": 4184
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9307668806222938e-05,
+      "loss": 1.0801,
+      "step": 4185
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930725680877703e-05,
+      "loss": 1.2068,
+      "step": 4186
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306844693178753e-05,
+      "loss": 1.0672,
+      "step": 4187
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930643245943333e-05,
+      "loss": 1.1151,
+      "step": 4188
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9306020107546004e-05,
+      "loss": 1.0322,
+      "step": 4189
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305607637522003e-05,
+      "loss": 0.97,
+      "step": 4190
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9305195049366567e-05,
+      "loss": 1.2418,
+      "step": 4191
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930478234308493e-05,
+      "loss": 1.2282,
+      "step": 4192
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930436951868234e-05,
+      "loss": 1.1094,
+      "step": 4193
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303956576164025e-05,
+      "loss": 1.004,
+      "step": 4194
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9303543515535236e-05,
+      "loss": 1.1547,
+      "step": 4195
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930313033680121e-05,
+      "loss": 1.0776,
+      "step": 4196
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.93027170399672e-05,
+      "loss": 0.9634,
+      "step": 4197
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9302303625038447e-05,
+      "loss": 0.9152,
+      "step": 4198
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301890092020204e-05,
+      "loss": 0.9615,
+      "step": 4199
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301476440917714e-05,
+      "loss": 1.1559,
+      "step": 4200
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9301062671736236e-05,
+      "loss": 0.9542,
+      "step": 4201
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9300648784481015e-05,
+      "loss": 1.1505,
+      "step": 4202
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.930023477915731e-05,
+      "loss": 1.079,
+      "step": 4203
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299820655770376e-05,
+      "loss": 1.1785,
+      "step": 4204
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9299406414325465e-05,
+      "loss": 1.1388,
+      "step": 4205
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298992054827844e-05,
+      "loss": 1.1057,
+      "step": 4206
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298577577282768e-05,
+      "loss": 1.0453,
+      "step": 4207
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9298162981695503e-05,
+      "loss": 1.0645,
+      "step": 4208
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297748268071306e-05,
+      "loss": 1.0447,
+      "step": 4209
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9297333436415443e-05,
+      "loss": 1.0071,
+      "step": 4210
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296918486733186e-05,
+      "loss": 1.0743,
+      "step": 4211
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296503419029797e-05,
+      "loss": 1.1243,
+      "step": 4212
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9296088233310543e-05,
+      "loss": 1.0031,
+      "step": 4213
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92956729295807e-05,
+      "loss": 1.1135,
+      "step": 4214
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9295257507845543e-05,
+      "loss": 1.1067,
+      "step": 4215
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294841968110333e-05,
+      "loss": 0.9785,
+      "step": 4216
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9294426310380357e-05,
+      "loss": 1.1271,
+      "step": 4217
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929401053466089e-05,
+      "loss": 1.0867,
+      "step": 4218
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293594640957205e-05,
+      "loss": 1.0623,
+      "step": 4219
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9293178629274587e-05,
+      "loss": 1.1066,
+      "step": 4220
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929276249961831e-05,
+      "loss": 1.0419,
+      "step": 4221
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9292346251993665e-05,
+      "loss": 1.2324,
+      "step": 4222
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291929886405932e-05,
+      "loss": 1.1383,
+      "step": 4223
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291513402860395e-05,
+      "loss": 1.0803,
+      "step": 4224
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9291096801362345e-05,
+      "loss": 1.2214,
+      "step": 4225
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.929068008191707e-05,
+      "loss": 0.8871,
+      "step": 4226
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9290263244529856e-05,
+      "loss": 1.048,
+      "step": 4227
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289846289206e-05,
+      "loss": 1.1376,
+      "step": 4228
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928942921595079e-05,
+      "loss": 1.1122,
+      "step": 4229
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9289012024769524e-05,
+      "loss": 1.1552,
+      "step": 4230
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92885947156675e-05,
+      "loss": 1.0597,
+      "step": 4231
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9288177288650007e-05,
+      "loss": 1.0625,
+      "step": 4232
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287759743722354e-05,
+      "loss": 1.1224,
+      "step": 4233
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9287342080889836e-05,
+      "loss": 0.9993,
+      "step": 4234
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286924300157755e-05,
+      "loss": 1.2865,
+      "step": 4235
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928650640153142e-05,
+      "loss": 1.0427,
+      "step": 4236
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9286088385016128e-05,
+      "loss": 1.0829,
+      "step": 4237
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9285670250617193e-05,
+      "loss": 1.0189,
+      "step": 4238
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928525199833992e-05,
+      "loss": 1.0841,
+      "step": 4239
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284833628189618e-05,
+      "loss": 1.0564,
+      "step": 4240
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9284415140171595e-05,
+      "loss": 1.0056,
+      "step": 4241
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.928399653429117e-05,
+      "loss": 1.1311,
+      "step": 4242
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283577810553654e-05,
+      "loss": 1.1246,
+      "step": 4243
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9283158968964364e-05,
+      "loss": 1.1334,
+      "step": 4244
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282740009528616e-05,
+      "loss": 1.1089,
+      "step": 4245
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9282320932251727e-05,
+      "loss": 1.0896,
+      "step": 4246
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281901737139017e-05,
+      "loss": 1.2189,
+      "step": 4247
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281482424195807e-05,
+      "loss": 1.1841,
+      "step": 4248
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9281062993427424e-05,
+      "loss": 1.1798,
+      "step": 4249
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280643444839194e-05,
+      "loss": 1.0516,
+      "step": 4250
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9280223778436436e-05,
+      "loss": 1.1291,
+      "step": 4251
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927980399422448e-05,
+      "loss": 1.2431,
+      "step": 4252
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9279384092208654e-05,
+      "loss": 1.1951,
+      "step": 4253
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9278964072394294e-05,
+      "loss": 1.0776,
+      "step": 4254
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927854393478673e-05,
+      "loss": 1.1808,
+      "step": 4255
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927812367939129e-05,
+      "loss": 1.0726,
+      "step": 4256
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9277703306213315e-05,
+      "loss": 1.1268,
+      "step": 4257
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927728281525814e-05,
+      "loss": 1.1988,
+      "step": 4258
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276862206531103e-05,
+      "loss": 1.0232,
+      "step": 4259
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276441480037538e-05,
+      "loss": 0.915,
+      "step": 4260
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9276020635782796e-05,
+      "loss": 1.1869,
+      "step": 4261
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927559967377221e-05,
+      "loss": 1.0645,
+      "step": 4262
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9275178594011136e-05,
+      "loss": 1.0366,
+      "step": 4263
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274757396504904e-05,
+      "loss": 1.1207,
+      "step": 4264
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9274336081258874e-05,
+      "loss": 1.0734,
+      "step": 4265
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273914648278384e-05,
+      "loss": 1.1903,
+      "step": 4266
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273493097568792e-05,
+      "loss": 1.1291,
+      "step": 4267
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9273071429135447e-05,
+      "loss": 1.1287,
+      "step": 4268
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92726496429837e-05,
+      "loss": 1.0018,
+      "step": 4269
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927222773911891e-05,
+      "loss": 1.0817,
+      "step": 4270
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9271805717546422e-05,
+      "loss": 1.0417,
+      "step": 4271
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927138357827161e-05,
+      "loss": 1.2034,
+      "step": 4272
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.927096132129982e-05,
+      "loss": 1.1835,
+      "step": 4273
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270538946636417e-05,
+      "loss": 1.1472,
+      "step": 4274
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9270116454286762e-05,
+      "loss": 1.1671,
+      "step": 4275
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269693844256225e-05,
+      "loss": 1.0732,
+      "step": 4276
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9269271116550157e-05,
+      "loss": 1.0581,
+      "step": 4277
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926884827117393e-05,
+      "loss": 1.0457,
+      "step": 4278
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268425308132922e-05,
+      "loss": 1.0815,
+      "step": 4279
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9268002227432488e-05,
+      "loss": 0.9613,
+      "step": 4280
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267579029078012e-05,
+      "loss": 1.167,
+      "step": 4281
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9267155713074857e-05,
+      "loss": 1.1351,
+      "step": 4282
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266732279428394e-05,
+      "loss": 0.9643,
+      "step": 4283
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9266308728144013e-05,
+      "loss": 1.0443,
+      "step": 4284
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265885059227075e-05,
+      "loss": 1.0706,
+      "step": 4285
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265461272682966e-05,
+      "loss": 1.0734,
+      "step": 4286
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9265037368517066e-05,
+      "loss": 1.0097,
+      "step": 4287
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264613346734754e-05,
+      "loss": 1.0628,
+      "step": 4288
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9264189207341415e-05,
+      "loss": 1.0843,
+      "step": 4289
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926376495034243e-05,
+      "loss": 1.1762,
+      "step": 4290
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926334057574319e-05,
+      "loss": 1.0573,
+      "step": 4291
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262916083549075e-05,
+      "loss": 1.0581,
+      "step": 4292
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262491473765483e-05,
+      "loss": 1.1765,
+      "step": 4293
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9262066746397796e-05,
+      "loss": 1.0841,
+      "step": 4294
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.926164190145141e-05,
+      "loss": 1.0916,
+      "step": 4295
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9261216938931717e-05,
+      "loss": 1.148,
+      "step": 4296
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260791858844113e-05,
+      "loss": 1.1378,
+      "step": 4297
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9260366661193993e-05,
+      "loss": 1.0436,
+      "step": 4298
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259941345986756e-05,
+      "loss": 1.1353,
+      "step": 4299
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259515913227798e-05,
+      "loss": 1.1483,
+      "step": 4300
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9259090362922524e-05,
+      "loss": 1.1794,
+      "step": 4301
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9258664695076335e-05,
+      "loss": 1.067,
+      "step": 4302
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925823890969463e-05,
+      "loss": 1.1251,
+      "step": 4303
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9257813006782822e-05,
+      "loss": 1.1134,
+      "step": 4304
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925738698634631e-05,
+      "loss": 1.1255,
+      "step": 4305
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925696084839051e-05,
+      "loss": 1.2079,
+      "step": 4306
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256534592920828e-05,
+      "loss": 1.0298,
+      "step": 4307
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9256108219942674e-05,
+      "loss": 1.1806,
+      "step": 4308
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9255681729461458e-05,
+      "loss": 1.0769,
+      "step": 4309
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92552551214826e-05,
+      "loss": 0.9766,
+      "step": 4310
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254828396011512e-05,
+      "loss": 1.2848,
+      "step": 4311
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9254401553053618e-05,
+      "loss": 1.1698,
+      "step": 4312
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253974592614324e-05,
+      "loss": 1.0402,
+      "step": 4313
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925354751469906e-05,
+      "loss": 1.1319,
+      "step": 4314
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9253120319313242e-05,
+      "loss": 1.1234,
+      "step": 4315
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92526930064623e-05,
+      "loss": 1.1324,
+      "step": 4316
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9252265576151654e-05,
+      "loss": 1.1683,
+      "step": 4317
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251838028386732e-05,
+      "loss": 0.9651,
+      "step": 4318
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9251410363172956e-05,
+      "loss": 1.1932,
+      "step": 4319
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925098258051576e-05,
+      "loss": 1.1524,
+      "step": 4320
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9250554680420573e-05,
+      "loss": 0.9872,
+      "step": 4321
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.925012666289283e-05,
+      "loss": 1.0876,
+      "step": 4322
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924969852793796e-05,
+      "loss": 1.2091,
+      "step": 4323
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9249270275561404e-05,
+      "loss": 1.1023,
+      "step": 4324
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248841905768594e-05,
+      "loss": 1.0681,
+      "step": 4325
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9248413418564965e-05,
+      "loss": 1.0388,
+      "step": 4326
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247984813955964e-05,
+      "loss": 1.0933,
+      "step": 4327
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9247556091947027e-05,
+      "loss": 1.0163,
+      "step": 4328
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92471272525436e-05,
+      "loss": 0.9619,
+      "step": 4329
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924669829575112e-05,
+      "loss": 1.1095,
+      "step": 4330
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9246269221575038e-05,
+      "loss": 1.1726,
+      "step": 4331
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9245840030020803e-05,
+      "loss": 1.122,
+      "step": 4332
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924541072109386e-05,
+      "loss": 1.0002,
+      "step": 4333
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244981294799657e-05,
+      "loss": 1.0799,
+      "step": 4334
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9244551751143652e-05,
+      "loss": 0.9943,
+      "step": 4335
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924412209013129e-05,
+      "loss": 1.0982,
+      "step": 4336
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9243692311768032e-05,
+      "loss": 1.0063,
+      "step": 4337
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924326241605933e-05,
+      "loss": 1.1344,
+      "step": 4338
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242832403010642e-05,
+      "loss": 1.1255,
+      "step": 4339
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9242402272627426e-05,
+      "loss": 1.092,
+      "step": 4340
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924197202491515e-05,
+      "loss": 1.1533,
+      "step": 4341
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9241541659879263e-05,
+      "loss": 1.0189,
+      "step": 4342
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.924111117752524e-05,
+      "loss": 1.0595,
+      "step": 4343
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240680577858534e-05,
+      "loss": 1.0796,
+      "step": 4344
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9240249860884623e-05,
+      "loss": 1.1185,
+      "step": 4345
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9239819026608968e-05,
+      "loss": 1.0139,
+      "step": 4346
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923938807503704e-05,
+      "loss": 1.1224,
+      "step": 4347
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923895700617431e-05,
+      "loss": 1.1117,
+      "step": 4348
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923852582002625e-05,
+      "loss": 1.124,
+      "step": 4349
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9238094516598335e-05,
+      "loss": 1.0746,
+      "step": 4350
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237663095896038e-05,
+      "loss": 1.1025,
+      "step": 4351
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9237231557924838e-05,
+      "loss": 1.1088,
+      "step": 4352
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9236799902690214e-05,
+      "loss": 1.1655,
+      "step": 4353
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923636813019764e-05,
+      "loss": 1.2231,
+      "step": 4354
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235936240452605e-05,
+      "loss": 1.0564,
+      "step": 4355
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235504233460583e-05,
+      "loss": 1.1643,
+      "step": 4356
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9235072109227065e-05,
+      "loss": 1.1333,
+      "step": 4357
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9234639867757537e-05,
+      "loss": 1.0066,
+      "step": 4358
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923420750905748e-05,
+      "loss": 1.0835,
+      "step": 4359
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9233775033132385e-05,
+      "loss": 1.1013,
+      "step": 4360
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923334243998775e-05,
+      "loss": 0.9967,
+      "step": 4361
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9232909729629056e-05,
+      "loss": 1.1501,
+      "step": 4362
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92324769020618e-05,
+      "loss": 1.2414,
+      "step": 4363
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923204395729148e-05,
+      "loss": 1.0348,
+      "step": 4364
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9231610895323584e-05,
+      "loss": 0.9675,
+      "step": 4365
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923117771616362e-05,
+      "loss": 1.19,
+      "step": 4366
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9230744419817077e-05,
+      "loss": 1.1162,
+      "step": 4367
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.923031100628946e-05,
+      "loss": 1.0859,
+      "step": 4368
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229877475586273e-05,
+      "loss": 0.9789,
+      "step": 4369
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9229443827713017e-05,
+      "loss": 1.1277,
+      "step": 4370
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.92290100626752e-05,
+      "loss": 1.1223,
+      "step": 4371
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922857618047832e-05,
+      "loss": 0.9984,
+      "step": 4372
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9228142181127897e-05,
+      "loss": 1.0643,
+      "step": 4373
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9227708064629435e-05,
+      "loss": 1.0367,
+      "step": 4374
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922727383098844e-05,
+      "loss": 1.0902,
+      "step": 4375
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9226839480210432e-05,
+      "loss": 1.2559,
+      "step": 4376
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922640501230092e-05,
+      "loss": 1.2127,
+      "step": 4377
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225970427265424e-05,
+      "loss": 1.1125,
+      "step": 4378
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225535725109456e-05,
+      "loss": 1.1946,
+      "step": 4379
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9225100905838538e-05,
+      "loss": 1.1031,
+      "step": 4380
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224665969458187e-05,
+      "loss": 1.1977,
+      "step": 4381
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9224230915973928e-05,
+      "loss": 1.1119,
+      "step": 4382
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922379574539128e-05,
+      "loss": 0.951,
+      "step": 4383
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9223360457715768e-05,
+      "loss": 1.0873,
+      "step": 4384
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222925052952922e-05,
+      "loss": 1.0046,
+      "step": 4385
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222489531108263e-05,
+      "loss": 1.1159,
+      "step": 4386
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9222053892187323e-05,
+      "loss": 1.1228,
+      "step": 4387
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9221618136195635e-05,
+      "loss": 1.1035,
+      "step": 4388
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922118226313873e-05,
+      "loss": 1.2398,
+      "step": 4389
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9220746273022134e-05,
+      "loss": 1.2051,
+      "step": 4390
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.922031016585139e-05,
+      "loss": 1.2381,
+      "step": 4391
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921987394163203e-05,
+      "loss": 1.1959,
+      "step": 4392
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9219437600369596e-05,
+      "loss": 1.0655,
+      "step": 4393
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921900114206962e-05,
+      "loss": 1.1869,
+      "step": 4394
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9218564566737647e-05,
+      "loss": 1.0961,
+      "step": 4395
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921812787437922e-05,
+      "loss": 1.0864,
+      "step": 4396
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217691064999885e-05,
+      "loss": 1.1383,
+      "step": 4397
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9217254138605182e-05,
+      "loss": 1.1696,
+      "step": 4398
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921681709520066e-05,
+      "loss": 1.0598,
+      "step": 4399
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9216379934791863e-05,
+      "loss": 1.0482,
+      "step": 4400
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921594265738435e-05,
+      "loss": 1.1932,
+      "step": 4401
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215505262983662e-05,
+      "loss": 1.112,
+      "step": 4402
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9215067751595355e-05,
+      "loss": 1.0601,
+      "step": 4403
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214630123224984e-05,
+      "loss": 1.1048,
+      "step": 4404
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9214192377878106e-05,
+      "loss": 1.0423,
+      "step": 4405
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213754515560278e-05,
+      "loss": 1.1011,
+      "step": 4406
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9213316536277052e-05,
+      "loss": 1.0351,
+      "step": 4407
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212878440034e-05,
+      "loss": 1.0327,
+      "step": 4408
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212440226836672e-05,
+      "loss": 1.1718,
+      "step": 4409
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9212001896690634e-05,
+      "loss": 0.9347,
+      "step": 4410
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211563449601456e-05,
+      "loss": 1.0465,
+      "step": 4411
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9211124885574697e-05,
+      "loss": 1.1089,
+      "step": 4412
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921068620461593e-05,
+      "loss": 0.9823,
+      "step": 4413
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.921024740673072e-05,
+      "loss": 1.142,
+      "step": 4414
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920980849192464e-05,
+      "loss": 1.111,
+      "step": 4415
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9209369460203254e-05,
+      "loss": 1.0749,
+      "step": 4416
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208930311572146e-05,
+      "loss": 0.9977,
+      "step": 4417
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208491046036884e-05,
+      "loss": 1.1394,
+      "step": 4418
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9208051663603052e-05,
+      "loss": 1.2122,
+      "step": 4419
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9207612164276218e-05,
+      "loss": 1.0334,
+      "step": 4420
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920717254806197e-05,
+      "loss": 1.1197,
+      "step": 4421
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206732814965884e-05,
+      "loss": 0.9896,
+      "step": 4422
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9206292964993538e-05,
+      "loss": 1.1942,
+      "step": 4423
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205852998150524e-05,
+      "loss": 1.2515,
+      "step": 4424
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9205412914442425e-05,
+      "loss": 1.0695,
+      "step": 4425
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204972713874824e-05,
+      "loss": 1.1447,
+      "step": 4426
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204532396453313e-05,
+      "loss": 1.1215,
+      "step": 4427
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9204091962183485e-05,
+      "loss": 1.1862,
+      "step": 4428
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203651411070923e-05,
+      "loss": 1.0971,
+      "step": 4429
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9203210743121222e-05,
+      "loss": 1.1609,
+      "step": 4430
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920276995833998e-05,
+      "loss": 1.0588,
+      "step": 4431
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920232905673279e-05,
+      "loss": 1.2518,
+      "step": 4432
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201888038305246e-05,
+      "loss": 0.9966,
+      "step": 4433
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920144690306295e-05,
+      "loss": 1.105,
+      "step": 4434
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9201005651011507e-05,
+      "loss": 1.1551,
+      "step": 4435
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.920056428215651e-05,
+      "loss": 1.2026,
+      "step": 4436
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9200122796503566e-05,
+      "loss": 1.0479,
+      "step": 4437
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9199681194058274e-05,
+      "loss": 1.1163,
+      "step": 4438
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919923947482625e-05,
+      "loss": 1.0981,
+      "step": 4439
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9198797638813097e-05,
+      "loss": 1.0838,
+      "step": 4440
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919835568602442e-05,
+      "loss": 0.9739,
+      "step": 4441
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197913616465836e-05,
+      "loss": 0.9845,
+      "step": 4442
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197471430142952e-05,
+      "loss": 1.2485,
+      "step": 4443
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9197029127061385e-05,
+      "loss": 1.1759,
+      "step": 4444
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9196586707226744e-05,
+      "loss": 1.0526,
+      "step": 4445
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919614417064465e-05,
+      "loss": 1.1506,
+      "step": 4446
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195701517320723e-05,
+      "loss": 1.0247,
+      "step": 4447
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9195258747260578e-05,
+      "loss": 1.0605,
+      "step": 4448
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194815860469836e-05,
+      "loss": 1.2243,
+      "step": 4449
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9194372856954123e-05,
+      "loss": 1.0629,
+      "step": 4450
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919392973671906e-05,
+      "loss": 1.0248,
+      "step": 4451
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9193486499770273e-05,
+      "loss": 1.1483,
+      "step": 4452
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.919304314611339e-05,
+      "loss": 1.1663,
+      "step": 4453
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192599675754033e-05,
+      "loss": 1.0869,
+      "step": 4454
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 1.9192156088697836e-05,
+      "loss": 1.0904,
+      "step": 4455
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191712384950434e-05,
+      "loss": 1.0084,
+      "step": 4456
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9191268564517454e-05,
+      "loss": 1.2322,
+      "step": 4457
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190824627404533e-05,
+      "loss": 1.0794,
+      "step": 4458
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9190380573617306e-05,
+      "loss": 1.0781,
+      "step": 4459
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918993640316141e-05,
+      "loss": 1.0939,
+      "step": 4460
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918949211604248e-05,
+      "loss": 1.0528,
+      "step": 4461
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9189047712266164e-05,
+      "loss": 1.1344,
+      "step": 4462
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188603191838098e-05,
+      "loss": 1.114,
+      "step": 4463
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9188158554763926e-05,
+      "loss": 1.0933,
+      "step": 4464
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187713801049295e-05,
+      "loss": 1.0679,
+      "step": 4465
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9187268930699847e-05,
+      "loss": 0.9228,
+      "step": 4466
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918682394372123e-05,
+      "loss": 0.9314,
+      "step": 4467
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9186378840119095e-05,
+      "loss": 1.0645,
+      "step": 4468
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185933619899093e-05,
+      "loss": 1.1154,
+      "step": 4469
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9185488283066873e-05,
+      "loss": 0.8383,
+      "step": 4470
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918504282962809e-05,
+      "loss": 1.055,
+      "step": 4471
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91845972595884e-05,
+      "loss": 1.0081,
+      "step": 4472
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9184151572953456e-05,
+      "loss": 1.0105,
+      "step": 4473
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9183705769728917e-05,
+      "loss": 0.9906,
+      "step": 4474
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.918325984992045e-05,
+      "loss": 1.1037,
+      "step": 4475
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182813813533703e-05,
+      "loss": 1.1276,
+      "step": 4476
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9182367660574342e-05,
+      "loss": 1.1993,
+      "step": 4477
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181921391048038e-05,
+      "loss": 1.0896,
+      "step": 4478
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181475004960453e-05,
+      "loss": 1.1592,
+      "step": 4479
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9181028502317248e-05,
+      "loss": 1.0468,
+      "step": 4480
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180581883124095e-05,
+      "loss": 1.0634,
+      "step": 4481
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9180135147386666e-05,
+      "loss": 0.9526,
+      "step": 4482
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179688295110632e-05,
+      "loss": 1.0355,
+      "step": 4483
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9179241326301662e-05,
+      "loss": 1.0389,
+      "step": 4484
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917879424096543e-05,
+      "loss": 1.1325,
+      "step": 4485
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9178347039107614e-05,
+      "loss": 1.0022,
+      "step": 4486
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177899720733893e-05,
+      "loss": 1.092,
+      "step": 4487
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177452285849943e-05,
+      "loss": 1.2109,
+      "step": 4488
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9177004734461442e-05,
+      "loss": 1.0494,
+      "step": 4489
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9176557066574073e-05,
+      "loss": 0.9709,
+      "step": 4490
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917610928219352e-05,
+      "loss": 1.1048,
+      "step": 4491
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9175661381325468e-05,
+      "loss": 1.1968,
+      "step": 4492
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91752133639756e-05,
+      "loss": 1.0398,
+      "step": 4493
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9174765230149608e-05,
+      "loss": 1.1695,
+      "step": 4494
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917431697985318e-05,
+      "loss": 1.0038,
+      "step": 4495
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173868613092e-05,
+      "loss": 1.0784,
+      "step": 4496
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9173420129871764e-05,
+      "loss": 1.0622,
+      "step": 4497
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917297153019817e-05,
+      "loss": 1.2619,
+      "step": 4498
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172522814076908e-05,
+      "loss": 1.0557,
+      "step": 4499
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9172073981513673e-05,
+      "loss": 1.3407,
+      "step": 4500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171625032514166e-05,
+      "loss": 1.2098,
+      "step": 4501
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9171175967084084e-05,
+      "loss": 1.1016,
+      "step": 4502
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917072678522913e-05,
+      "loss": 1.1781,
+      "step": 4503
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.917027748695501e-05,
+      "loss": 1.0561,
+      "step": 4504
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9169828072267415e-05,
+      "loss": 1.1273,
+      "step": 4505
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916937854117206e-05,
+      "loss": 1.1637,
+      "step": 4506
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916892889367465e-05,
+      "loss": 1.226,
+      "step": 4507
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916847912978089e-05,
+      "loss": 1.1882,
+      "step": 4508
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9168029249496495e-05,
+      "loss": 1.0562,
+      "step": 4509
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167579252827173e-05,
+      "loss": 1.0481,
+      "step": 4510
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9167129139778638e-05,
+      "loss": 1.063,
+      "step": 4511
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91666789103566e-05,
+      "loss": 1.1056,
+      "step": 4512
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9166228564566778e-05,
+      "loss": 1.1697,
+      "step": 4513
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9165778102414888e-05,
+      "loss": 1.0181,
+      "step": 4514
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916532752390665e-05,
+      "loss": 1.0009,
+      "step": 4515
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164876829047783e-05,
+      "loss": 1.1381,
+      "step": 4516
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9164426017844006e-05,
+      "loss": 1.0654,
+      "step": 4517
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163975090301047e-05,
+      "loss": 1.0149,
+      "step": 4518
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163524046424625e-05,
+      "loss": 0.9962,
+      "step": 4519
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9163072886220472e-05,
+      "loss": 1.2789,
+      "step": 4520
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9162621609694304e-05,
+      "loss": 0.9652,
+      "step": 4521
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.916217021685186e-05,
+      "loss": 1.1211,
+      "step": 4522
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161718707698873e-05,
+      "loss": 1.1556,
+      "step": 4523
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9161267082241067e-05,
+      "loss": 1.1878,
+      "step": 4524
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160815340484177e-05,
+      "loss": 1.1294,
+      "step": 4525
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9160363482433937e-05,
+      "loss": 1.0542,
+      "step": 4526
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159911508096084e-05,
+      "loss": 1.0347,
+      "step": 4527
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915945941747636e-05,
+      "loss": 1.1839,
+      "step": 4528
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9159007210580495e-05,
+      "loss": 1.0827,
+      "step": 4529
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9158554887414237e-05,
+      "loss": 1.0428,
+      "step": 4530
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915810244798333e-05,
+      "loss": 1.0734,
+      "step": 4531
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157649892293507e-05,
+      "loss": 1.2181,
+      "step": 4532
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9157197220350524e-05,
+      "loss": 1.0231,
+      "step": 4533
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9156744432160122e-05,
+      "loss": 1.154,
+      "step": 4534
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915629152772805e-05,
+      "loss": 1.0026,
+      "step": 4535
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9155838507060056e-05,
+      "loss": 1.2095,
+      "step": 4536
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915538537016189e-05,
+      "loss": 1.0494,
+      "step": 4537
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9154932117039312e-05,
+      "loss": 1.2291,
+      "step": 4538
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915447874769807e-05,
+      "loss": 1.1214,
+      "step": 4539
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915402526214392e-05,
+      "loss": 1.0082,
+      "step": 4540
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153571660382617e-05,
+      "loss": 0.9943,
+      "step": 4541
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9153117942419924e-05,
+      "loss": 1.2028,
+      "step": 4542
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9152664108261597e-05,
+      "loss": 1.1146,
+      "step": 4543
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91522101579134e-05,
+      "loss": 1.0386,
+      "step": 4544
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.915175609138109e-05,
+      "loss": 1.0572,
+      "step": 4545
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9151301908670437e-05,
+      "loss": 1.1067,
+      "step": 4546
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150847609787203e-05,
+      "loss": 1.0438,
+      "step": 4547
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9150393194737164e-05,
+      "loss": 0.9736,
+      "step": 4548
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149938663526076e-05,
+      "loss": 1.0175,
+      "step": 4549
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149484016159715e-05,
+      "loss": 1.1463,
+      "step": 4550
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9149029252643855e-05,
+      "loss": 1.1213,
+      "step": 4551
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9148574372984264e-05,
+      "loss": 1.077,
+      "step": 4552
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914811937718672e-05,
+      "loss": 1.1887,
+      "step": 4553
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147664265257e-05,
+      "loss": 1.031,
+      "step": 4554
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9147209037200877e-05,
+      "loss": 1.0539,
+      "step": 4555
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9146753693024135e-05,
+      "loss": 1.0666,
+      "step": 4556
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.914629823273255e-05,
+      "loss": 1.0693,
+      "step": 4557
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145842656331906e-05,
+      "loss": 1.109,
+      "step": 4558
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9145386963827986e-05,
+      "loss": 1.1689,
+      "step": 4559
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144931155226575e-05,
+      "loss": 1.1372,
+      "step": 4560
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144475230533462e-05,
+      "loss": 1.0987,
+      "step": 4561
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9144019189754432e-05,
+      "loss": 1.0467,
+      "step": 4562
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143563032895273e-05,
+      "loss": 1.0836,
+      "step": 4563
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9143106759961775e-05,
+      "loss": 1.1096,
+      "step": 4564
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142650370959735e-05,
+      "loss": 1.1275,
+      "step": 4565
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9142193865894943e-05,
+      "loss": 1.0568,
+      "step": 4566
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91417372447732e-05,
+      "loss": 1.0541,
+      "step": 4567
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9141280507600294e-05,
+      "loss": 1.0887,
+      "step": 4568
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9140823654382026e-05,
+      "loss": 0.9667,
+      "step": 4569
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91403666851242e-05,
+      "loss": 1.0681,
+      "step": 4570
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913990959983261e-05,
+      "loss": 1.2088,
+      "step": 4571
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9139452398513065e-05,
+      "loss": 1.1454,
+      "step": 4572
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138995081171368e-05,
+      "loss": 1.0528,
+      "step": 4573
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138537647813323e-05,
+      "loss": 1.1048,
+      "step": 4574
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9138080098444735e-05,
+      "loss": 1.2689,
+      "step": 4575
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9137622433071413e-05,
+      "loss": 1.0546,
+      "step": 4576
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913716465169917e-05,
+      "loss": 1.164,
+      "step": 4577
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9136706754333813e-05,
+      "loss": 0.9743,
+      "step": 4578
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913624874098116e-05,
+      "loss": 1.0956,
+      "step": 4579
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913579061164702e-05,
+      "loss": 1.0314,
+      "step": 4580
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9135332366337213e-05,
+      "loss": 1.1152,
+      "step": 4581
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134874005057557e-05,
+      "loss": 1.1294,
+      "step": 4582
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9134415527813865e-05,
+      "loss": 1.131,
+      "step": 4583
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9133956934611963e-05,
+      "loss": 1.0311,
+      "step": 4584
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913349822545767e-05,
+      "loss": 1.0261,
+      "step": 4585
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.913303940035681e-05,
+      "loss": 1.2219,
+      "step": 4586
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132580459315205e-05,
+      "loss": 1.0817,
+      "step": 4587
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9132121402338685e-05,
+      "loss": 1.1531,
+      "step": 4588
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131662229433075e-05,
+      "loss": 1.1765,
+      "step": 4589
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9131202940604202e-05,
+      "loss": 0.9443,
+      "step": 4590
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130743535857905e-05,
+      "loss": 1.1199,
+      "step": 4591
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9130284015200008e-05,
+      "loss": 1.1153,
+      "step": 4592
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129824378636348e-05,
+      "loss": 1.098,
+      "step": 4593
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9129364626172754e-05,
+      "loss": 1.0665,
+      "step": 4594
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128904757815072e-05,
+      "loss": 0.9413,
+      "step": 4595
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9128444773569134e-05,
+      "loss": 1.0331,
+      "step": 4596
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912798467344078e-05,
+      "loss": 1.0969,
+      "step": 4597
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9127524457435852e-05,
+      "loss": 1.0674,
+      "step": 4598
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912706412556019e-05,
+      "loss": 1.1063,
+      "step": 4599
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912660367781964e-05,
+      "loss": 1.005,
+      "step": 4600
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9126143114220042e-05,
+      "loss": 1.0016,
+      "step": 4601
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125682434767253e-05,
+      "loss": 0.9672,
+      "step": 4602
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9125221639467112e-05,
+      "loss": 1.1079,
+      "step": 4603
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9124760728325475e-05,
+      "loss": 1.1555,
+      "step": 4604
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912429970134819e-05,
+      "loss": 0.9922,
+      "step": 4605
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9123838558541107e-05,
+      "loss": 1.2322,
+      "step": 4606
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912337729991008e-05,
+      "loss": 1.001,
+      "step": 4607
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912291592546097e-05,
+      "loss": 1.1934,
+      "step": 4608
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9122454435199632e-05,
+      "loss": 1.0281,
+      "step": 4609
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121992829131925e-05,
+      "loss": 1.0843,
+      "step": 4610
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121531107263703e-05,
+      "loss": 1.1528,
+      "step": 4611
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9121069269600837e-05,
+      "loss": 1.0428,
+      "step": 4612
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.912060731614918e-05,
+      "loss": 1.0732,
+      "step": 4613
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9120145246914603e-05,
+      "loss": 0.9907,
+      "step": 4614
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119683061902973e-05,
+      "loss": 1.0184,
+      "step": 4615
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9119220761120153e-05,
+      "loss": 1.1338,
+      "step": 4616
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911875834457201e-05,
+      "loss": 1.0614,
+      "step": 4617
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911829581226442e-05,
+      "loss": 1.1775,
+      "step": 4618
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117833164203248e-05,
+      "loss": 1.0406,
+      "step": 4619
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9117370400394374e-05,
+      "loss": 0.9813,
+      "step": 4620
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911690752084367e-05,
+      "loss": 1.2264,
+      "step": 4621
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911644452555701e-05,
+      "loss": 1.126,
+      "step": 4622
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115981414540275e-05,
+      "loss": 1.1488,
+      "step": 4623
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911551818779934e-05,
+      "loss": 0.9789,
+      "step": 4624
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9115054845340088e-05,
+      "loss": 1.2029,
+      "step": 4625
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114591387168403e-05,
+      "loss": 1.1583,
+      "step": 4626
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9114127813290164e-05,
+      "loss": 1.0636,
+      "step": 4627
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.911366412371126e-05,
+      "loss": 0.9717,
+      "step": 4628
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9113200318437576e-05,
+      "loss": 1.0176,
+      "step": 4629
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112736397474997e-05,
+      "loss": 1.0235,
+      "step": 4630
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9112272360829415e-05,
+      "loss": 1.0408,
+      "step": 4631
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111808208506722e-05,
+      "loss": 1.0939,
+      "step": 4632
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9111343940512806e-05,
+      "loss": 1.101,
+      "step": 4633
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110879556853564e-05,
+      "loss": 1.1133,
+      "step": 4634
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9110415057534896e-05,
+      "loss": 0.994,
+      "step": 4635
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910995044256269e-05,
+      "loss": 1.0012,
+      "step": 4636
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109485711942843e-05,
+      "loss": 1.0255,
+      "step": 4637
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9109020865681263e-05,
+      "loss": 1.0218,
+      "step": 4638
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910855590378385e-05,
+      "loss": 1.0801,
+      "step": 4639
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9108090826256496e-05,
+      "loss": 0.9903,
+      "step": 4640
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910762563310512e-05,
+      "loss": 1.0585,
+      "step": 4641
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9107160324335617e-05,
+      "loss": 1.1222,
+      "step": 4642
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106694899953894e-05,
+      "loss": 1.1179,
+      "step": 4643
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9106229359965868e-05,
+      "loss": 1.1798,
+      "step": 4644
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105763704377442e-05,
+      "loss": 1.0412,
+      "step": 4645
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9105297933194524e-05,
+      "loss": 1.1678,
+      "step": 4646
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104832046423035e-05,
+      "loss": 1.1131,
+      "step": 4647
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9104366044068885e-05,
+      "loss": 1.0707,
+      "step": 4648
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910389992613799e-05,
+      "loss": 1.1516,
+      "step": 4649
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9103433692636267e-05,
+      "loss": 1.074,
+      "step": 4650
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102967343569634e-05,
+      "loss": 1.0967,
+      "step": 4651
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9102500878944018e-05,
+      "loss": 1.0515,
+      "step": 4652
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.910203429876533e-05,
+      "loss": 1.0124,
+      "step": 4653
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91015676030395e-05,
+      "loss": 1.0145,
+      "step": 4654
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9101100791772453e-05,
+      "loss": 1.2529,
+      "step": 4655
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9100633864970105e-05,
+      "loss": 1.2019,
+      "step": 4656
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.91001668226384e-05,
+      "loss": 1.1417,
+      "step": 4657
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9099699664783255e-05,
+      "loss": 0.9623,
+      "step": 4658
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90992323914106e-05,
+      "loss": 1.0543,
+      "step": 4659
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098765002526372e-05,
+      "loss": 1.1271,
+      "step": 4660
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9098297498136503e-05,
+      "loss": 1.2113,
+      "step": 4661
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097829878246926e-05,
+      "loss": 1.0839,
+      "step": 4662
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9097362142863583e-05,
+      "loss": 1.1241,
+      "step": 4663
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096894291992406e-05,
+      "loss": 1.0845,
+      "step": 4664
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9096426325639333e-05,
+      "loss": 1.0768,
+      "step": 4665
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095958243810312e-05,
+      "loss": 1.1124,
+      "step": 4666
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095490046511277e-05,
+      "loss": 1.1809,
+      "step": 4667
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9095021733748172e-05,
+      "loss": 1.2229,
+      "step": 4668
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909455330552695e-05,
+      "loss": 1.1538,
+      "step": 4669
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9094084761853553e-05,
+      "loss": 1.1152,
+      "step": 4670
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093616102733926e-05,
+      "loss": 1.155,
+      "step": 4671
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9093147328174022e-05,
+      "loss": 0.9693,
+      "step": 4672
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909267843817979e-05,
+      "loss": 1.1037,
+      "step": 4673
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9092209432757184e-05,
+      "loss": 1.0172,
+      "step": 4674
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9091740311912163e-05,
+      "loss": 1.1273,
+      "step": 4675
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.909127107565067e-05,
+      "loss": 1.1167,
+      "step": 4676
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090801723978667e-05,
+      "loss": 1.0441,
+      "step": 4677
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9090332256902118e-05,
+      "loss": 1.2733,
+      "step": 4678
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089862674426976e-05,
+      "loss": 1.1531,
+      "step": 4679
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9089392976559208e-05,
+      "loss": 1.1536,
+      "step": 4680
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908892316330477e-05,
+      "loss": 1.0916,
+      "step": 4681
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908845323466963e-05,
+      "loss": 1.0915,
+      "step": 4682
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908798319065975e-05,
+      "loss": 1.046,
+      "step": 4683
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087513031281105e-05,
+      "loss": 1.1726,
+      "step": 4684
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9087042756539657e-05,
+      "loss": 1.0391,
+      "step": 4685
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086572366441376e-05,
+      "loss": 1.0244,
+      "step": 4686
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9086101860992236e-05,
+      "loss": 1.0224,
+      "step": 4687
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908563124019821e-05,
+      "loss": 1.2231,
+      "step": 4688
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908516050406527e-05,
+      "loss": 1.2422,
+      "step": 4689
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084689652599394e-05,
+      "loss": 1.0275,
+      "step": 4690
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9084218685806557e-05,
+      "loss": 1.0894,
+      "step": 4691
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9083747603692738e-05,
+      "loss": 1.088,
+      "step": 4692
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908327640626392e-05,
+      "loss": 1.1549,
+      "step": 4693
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908280509352608e-05,
+      "loss": 1.2099,
+      "step": 4694
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.908233366548521e-05,
+      "loss": 1.0447,
+      "step": 4695
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081862122147284e-05,
+      "loss": 1.083,
+      "step": 4696
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9081390463518295e-05,
+      "loss": 1.1537,
+      "step": 4697
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080918689604225e-05,
+      "loss": 1.0022,
+      "step": 4698
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9080446800411068e-05,
+      "loss": 1.2372,
+      "step": 4699
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079974795944815e-05,
+      "loss": 1.1095,
+      "step": 4700
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079502676211455e-05,
+      "loss": 0.9579,
+      "step": 4701
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9079030441216982e-05,
+      "loss": 1.1784,
+      "step": 4702
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907855809096739e-05,
+      "loss": 1.0647,
+      "step": 4703
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907808562546868e-05,
+      "loss": 1.1455,
+      "step": 4704
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907761304472684e-05,
+      "loss": 1.0371,
+      "step": 4705
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907714034874788e-05,
+      "loss": 1.0116,
+      "step": 4706
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076667537537794e-05,
+      "loss": 1.0819,
+      "step": 4707
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9076194611102585e-05,
+      "loss": 1.0604,
+      "step": 4708
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907572156944826e-05,
+      "loss": 1.2009,
+      "step": 4709
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9075248412580822e-05,
+      "loss": 1.0931,
+      "step": 4710
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074775140506277e-05,
+      "loss": 1.0817,
+      "step": 4711
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9074301753230636e-05,
+      "loss": 1.0519,
+      "step": 4712
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90738282507599e-05,
+      "loss": 1.1793,
+      "step": 4713
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907335463310009e-05,
+      "loss": 1.0049,
+      "step": 4714
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072880900257215e-05,
+      "loss": 1.0063,
+      "step": 4715
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9072407052237287e-05,
+      "loss": 1.1416,
+      "step": 4716
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9071933089046324e-05,
+      "loss": 0.9979,
+      "step": 4717
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907145901069034e-05,
+      "loss": 1.0556,
+      "step": 4718
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9070984817175355e-05,
+      "loss": 1.2123,
+      "step": 4719
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907051050850739e-05,
+      "loss": 1.021,
+      "step": 4720
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.907003608469246e-05,
+      "loss": 0.9679,
+      "step": 4721
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069561545736593e-05,
+      "loss": 1.2634,
+      "step": 4722
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9069086891645815e-05,
+      "loss": 1.1156,
+      "step": 4723
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906861212242615e-05,
+      "loss": 1.078,
+      "step": 4724
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9068137238083624e-05,
+      "loss": 1.0982,
+      "step": 4725
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906766223862426e-05,
+      "loss": 1.2203,
+      "step": 4726
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.90671871240541e-05,
+      "loss": 0.9155,
+      "step": 4727
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066711894379163e-05,
+      "loss": 1.1814,
+      "step": 4728
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9066236549605492e-05,
+      "loss": 0.9757,
+      "step": 4729
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9065761089739115e-05,
+      "loss": 1.1388,
+      "step": 4730
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906528551478607e-05,
+      "loss": 1.1158,
+      "step": 4731
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9064809824752398e-05,
+      "loss": 0.9149,
+      "step": 4732
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906433401964413e-05,
+      "loss": 1.1476,
+      "step": 4733
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063858099467307e-05,
+      "loss": 1.0638,
+      "step": 4734
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9063382064227976e-05,
+      "loss": 1.1142,
+      "step": 4735
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906290591393218e-05,
+      "loss": 1.135,
+      "step": 4736
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9062429648585957e-05,
+      "loss": 1.1108,
+      "step": 4737
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906195326819536e-05,
+      "loss": 0.9221,
+      "step": 4738
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061476772766435e-05,
+      "loss": 1.0869,
+      "step": 4739
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9061000162305227e-05,
+      "loss": 1.0708,
+      "step": 4740
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.906052343681779e-05,
+      "loss": 0.9653,
+      "step": 4741
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9060046596310175e-05,
+      "loss": 1.0572,
+      "step": 4742
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 1.9059569640788436e-05,
+      "loss": 1.1874,
+      "step": 4743
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905909257025863e-05,
+      "loss": 1.1082,
+      "step": 4744
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058615384726805e-05,
+      "loss": 1.2598,
+      "step": 4745
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9058138084199026e-05,
+      "loss": 1.1206,
+      "step": 4746
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057660668681347e-05,
+      "loss": 1.0567,
+      "step": 4747
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9057183138179837e-05,
+      "loss": 1.0868,
+      "step": 4748
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905670549270055e-05,
+      "loss": 1.1218,
+      "step": 4749
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9056227732249554e-05,
+      "loss": 1.14,
+      "step": 4750
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905574985683291e-05,
+      "loss": 1.0289,
+      "step": 4751
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9055271866456688e-05,
+      "loss": 0.9674,
+      "step": 4752
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9054793761126954e-05,
+      "loss": 1.1703,
+      "step": 4753
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905431554084978e-05,
+      "loss": 1.1134,
+      "step": 4754
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9053837205631236e-05,
+      "loss": 1.0652,
+      "step": 4755
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905335875547739e-05,
+      "loss": 1.1119,
+      "step": 4756
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905288019039432e-05,
+      "loss": 1.1352,
+      "step": 4757
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90524015103881e-05,
+      "loss": 1.2698,
+      "step": 4758
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.905192271546481e-05,
+      "loss": 0.9918,
+      "step": 4759
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9051443805630522e-05,
+      "loss": 1.0276,
+      "step": 4760
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050964780891325e-05,
+      "loss": 1.12,
+      "step": 4761
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050485641253288e-05,
+      "loss": 1.1197,
+      "step": 4762
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9050006386722496e-05,
+      "loss": 1.1734,
+      "step": 4763
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049527017305044e-05,
+      "loss": 1.1283,
+      "step": 4764
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9049047533007004e-05,
+      "loss": 1.1517,
+      "step": 4765
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904856793383447e-05,
+      "loss": 1.2316,
+      "step": 4766
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904808821979353e-05,
+      "loss": 1.192,
+      "step": 4767
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047608390890268e-05,
+      "loss": 1.1588,
+      "step": 4768
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9047128447130786e-05,
+      "loss": 1.164,
+      "step": 4769
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9046648388521168e-05,
+      "loss": 1.0721,
+      "step": 4770
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904616821506751e-05,
+      "loss": 1.0631,
+      "step": 4771
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904568792677591e-05,
+      "loss": 1.0454,
+      "step": 4772
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9045207523652466e-05,
+      "loss": 1.0478,
+      "step": 4773
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904472700570327e-05,
+      "loss": 1.2369,
+      "step": 4774
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9044246372934426e-05,
+      "loss": 1.0248,
+      "step": 4775
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904376562535204e-05,
+      "loss": 1.2038,
+      "step": 4776
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9043284762962207e-05,
+      "loss": 1.1089,
+      "step": 4777
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042803785771037e-05,
+      "loss": 1.2341,
+      "step": 4778
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9042322693784632e-05,
+      "loss": 1.0568,
+      "step": 4779
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9041841487009098e-05,
+      "loss": 1.0841,
+      "step": 4780
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.904136016545055e-05,
+      "loss": 1.086,
+      "step": 4781
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040878729115094e-05,
+      "loss": 1.0728,
+      "step": 4782
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9040397178008846e-05,
+      "loss": 1.2217,
+      "step": 4783
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039915512137914e-05,
+      "loss": 1.1206,
+      "step": 4784
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9039433731508414e-05,
+      "loss": 1.1981,
+      "step": 4785
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038951836126464e-05,
+      "loss": 1.1346,
+      "step": 4786
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9038469825998178e-05,
+      "loss": 1.1698,
+      "step": 4787
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9037987701129673e-05,
+      "loss": 1.0461,
+      "step": 4788
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903750546152708e-05,
+      "loss": 1.1303,
+      "step": 4789
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903702310719651e-05,
+      "loss": 1.0921,
+      "step": 4790
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903654063814409e-05,
+      "loss": 1.0133,
+      "step": 4791
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9036058054375948e-05,
+      "loss": 1.0407,
+      "step": 4792
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903557535589821e-05,
+      "loss": 0.9646,
+      "step": 4793
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9035092542716996e-05,
+      "loss": 1.0488,
+      "step": 4794
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903460961483844e-05,
+      "loss": 1.1091,
+      "step": 4795
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9034126572268675e-05,
+      "loss": 1.0469,
+      "step": 4796
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9033643415013833e-05,
+      "loss": 1.0851,
+      "step": 4797
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.903316014308004e-05,
+      "loss": 1.1734,
+      "step": 4798
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032676756473442e-05,
+      "loss": 1.1751,
+      "step": 4799
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9032193255200163e-05,
+      "loss": 1.1098,
+      "step": 4800
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031709639266354e-05,
+      "loss": 1.1295,
+      "step": 4801
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9031225908678145e-05,
+      "loss": 1.2492,
+      "step": 4802
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030742063441682e-05,
+      "loss": 1.1828,
+      "step": 4803
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9030258103563105e-05,
+      "loss": 1.0612,
+      "step": 4804
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9029774029048555e-05,
+      "loss": 1.2669,
+      "step": 4805
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902928983990418e-05,
+      "loss": 1.0566,
+      "step": 4806
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902880553613613e-05,
+      "loss": 1.1924,
+      "step": 4807
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902832111775055e-05,
+      "loss": 1.2143,
+      "step": 4808
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9027836584753586e-05,
+      "loss": 1.1328,
+      "step": 4809
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902735193715139e-05,
+      "loss": 1.03,
+      "step": 4810
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902686717495012e-05,
+      "loss": 1.1223,
+      "step": 4811
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9026382298155928e-05,
+      "loss": 1.0698,
+      "step": 4812
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9025897306774966e-05,
+      "loss": 1.1193,
+      "step": 4813
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902541220081339e-05,
+      "loss": 1.2427,
+      "step": 4814
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902492698027736e-05,
+      "loss": 1.0775,
+      "step": 4815
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9024441645173045e-05,
+      "loss": 1.1246,
+      "step": 4816
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902395619550659e-05,
+      "loss": 1.0889,
+      "step": 4817
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902347063128417e-05,
+      "loss": 0.9687,
+      "step": 4818
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902298495251194e-05,
+      "loss": 1.0435,
+      "step": 4819
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902249915919607e-05,
+      "loss": 1.0814,
+      "step": 4820
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9022013251342733e-05,
+      "loss": 1.092,
+      "step": 4821
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021527228958086e-05,
+      "loss": 1.0661,
+      "step": 4822
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9021041092048306e-05,
+      "loss": 1.1706,
+      "step": 4823
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.902055484061956e-05,
+      "loss": 1.1127,
+      "step": 4824
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9020068474678026e-05,
+      "loss": 1.0485,
+      "step": 4825
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9019581994229873e-05,
+      "loss": 1.1292,
+      "step": 4826
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901909539928128e-05,
+      "loss": 1.0833,
+      "step": 4827
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018608689838425e-05,
+      "loss": 1.0314,
+      "step": 4828
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9018121865907485e-05,
+      "loss": 1.1053,
+      "step": 4829
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901763492749464e-05,
+      "loss": 1.0524,
+      "step": 4830
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9017147874606072e-05,
+      "loss": 1.0886,
+      "step": 4831
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901666070724796e-05,
+      "loss": 1.0194,
+      "step": 4832
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9016173425426498e-05,
+      "loss": 1.1084,
+      "step": 4833
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901568602914786e-05,
+      "loss": 1.0875,
+      "step": 4834
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9015198518418243e-05,
+      "loss": 1.2705,
+      "step": 4835
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901471089324383e-05,
+      "loss": 0.9696,
+      "step": 4836
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9014223153630812e-05,
+      "loss": 1.0686,
+      "step": 4837
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9013735299585385e-05,
+      "loss": 1.1617,
+      "step": 4838
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901324733111374e-05,
+      "loss": 1.1798,
+      "step": 4839
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012759248222066e-05,
+      "loss": 1.1304,
+      "step": 4840
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9012271050916565e-05,
+      "loss": 1.1618,
+      "step": 4841
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.901178273920343e-05,
+      "loss": 1.2006,
+      "step": 4842
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9011294313088872e-05,
+      "loss": 0.9854,
+      "step": 4843
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010805772579078e-05,
+      "loss": 1.1351,
+      "step": 4844
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9010317117680252e-05,
+      "loss": 1.0887,
+      "step": 4845
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.90098283483986e-05,
+      "loss": 1.1155,
+      "step": 4846
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9009339464740328e-05,
+      "loss": 1.0937,
+      "step": 4847
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900885046671164e-05,
+      "loss": 1.1514,
+      "step": 4848
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9008361354318746e-05,
+      "loss": 1.096,
+      "step": 4849
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9007872127567853e-05,
+      "loss": 1.0613,
+      "step": 4850
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900738278646517e-05,
+      "loss": 1.1322,
+      "step": 4851
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900689333101691e-05,
+      "loss": 1.1537,
+      "step": 4852
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9006403761229292e-05,
+      "loss": 0.9913,
+      "step": 4853
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900591407710852e-05,
+      "loss": 1.1746,
+      "step": 4854
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900542427866082e-05,
+      "loss": 1.2093,
+      "step": 4855
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900493436589241e-05,
+      "loss": 1.0221,
+      "step": 4856
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9004444338809504e-05,
+      "loss": 1.0675,
+      "step": 4857
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9003954197418322e-05,
+      "loss": 0.9975,
+      "step": 4858
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900346394172509e-05,
+      "loss": 1.113,
+      "step": 4859
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900297357173603e-05,
+      "loss": 1.179,
+      "step": 4860
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9002483087457365e-05,
+      "loss": 1.1284,
+      "step": 4861
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9001992488895325e-05,
+      "loss": 0.9872,
+      "step": 4862
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900150177605614e-05,
+      "loss": 1.0381,
+      "step": 4863
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900101094894603e-05,
+      "loss": 1.1523,
+      "step": 4864
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.9000520007571238e-05,
+      "loss": 1.0873,
+      "step": 4865
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.900002895193799e-05,
+      "loss": 1.2096,
+      "step": 4866
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899953778205252e-05,
+      "loss": 1.0424,
+      "step": 4867
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8999046497921062e-05,
+      "loss": 1.1469,
+      "step": 4868
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998555099549856e-05,
+      "loss": 1.0979,
+      "step": 4869
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8998063586945137e-05,
+      "loss": 1.0774,
+      "step": 4870
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997571960113147e-05,
+      "loss": 1.07,
+      "step": 4871
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8997080219060125e-05,
+      "loss": 1.0531,
+      "step": 4872
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996588363792314e-05,
+      "loss": 1.1701,
+      "step": 4873
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8996096394315963e-05,
+      "loss": 1.1811,
+      "step": 4874
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995604310637308e-05,
+      "loss": 1.0527,
+      "step": 4875
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8995112112762603e-05,
+      "loss": 1.0166,
+      "step": 4876
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994619800698088e-05,
+      "loss": 1.1471,
+      "step": 4877
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8994127374450026e-05,
+      "loss": 1.0532,
+      "step": 4878
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993634834024656e-05,
+      "loss": 0.9896,
+      "step": 4879
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8993142179428237e-05,
+      "loss": 1.0381,
+      "step": 4880
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899264941066702e-05,
+      "loss": 1.0647,
+      "step": 4881
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8992156527747263e-05,
+      "loss": 1.0693,
+      "step": 4882
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8991663530675224e-05,
+      "loss": 1.0963,
+      "step": 4883
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899117041945716e-05,
+      "loss": 1.098,
+      "step": 4884
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8990677194099326e-05,
+      "loss": 1.125,
+      "step": 4885
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.899018385460799e-05,
+      "loss": 1.026,
+      "step": 4886
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898969040098941e-05,
+      "loss": 1.1562,
+      "step": 4887
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8989196833249854e-05,
+      "loss": 1.0439,
+      "step": 4888
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988703151395587e-05,
+      "loss": 1.0008,
+      "step": 4889
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8988209355432877e-05,
+      "loss": 1.0272,
+      "step": 4890
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987715445367988e-05,
+      "loss": 1.1951,
+      "step": 4891
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8987221421207195e-05,
+      "loss": 1.2282,
+      "step": 4892
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986727282956766e-05,
+      "loss": 1.1298,
+      "step": 4893
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8986233030622977e-05,
+      "loss": 0.9874,
+      "step": 4894
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8985738664212098e-05,
+      "loss": 1.1253,
+      "step": 4895
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898524418373041e-05,
+      "loss": 0.9321,
+      "step": 4896
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984749589184187e-05,
+      "loss": 1.0793,
+      "step": 4897
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8984254880579705e-05,
+      "loss": 1.1016,
+      "step": 4898
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983760057923254e-05,
+      "loss": 1.1127,
+      "step": 4899
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8983265121221107e-05,
+      "loss": 0.9935,
+      "step": 4900
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8982770070479547e-05,
+      "loss": 1.1417,
+      "step": 4901
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898227490570486e-05,
+      "loss": 1.0736,
+      "step": 4902
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898177962690334e-05,
+      "loss": 1.079,
+      "step": 4903
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.898128423408126e-05,
+      "loss": 1.2383,
+      "step": 4904
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980788727244918e-05,
+      "loss": 1.1415,
+      "step": 4905
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8980293106400602e-05,
+      "loss": 0.9815,
+      "step": 4906
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979797371554603e-05,
+      "loss": 1.176,
+      "step": 4907
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8979301522713216e-05,
+      "loss": 1.0665,
+      "step": 4908
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978805559882736e-05,
+      "loss": 1.1683,
+      "step": 4909
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8978309483069455e-05,
+      "loss": 1.1257,
+      "step": 4910
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977813292279678e-05,
+      "loss": 1.0717,
+      "step": 4911
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8977316987519697e-05,
+      "loss": 1.1522,
+      "step": 4912
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8976820568795813e-05,
+      "loss": 1.0756,
+      "step": 4913
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897632403611433e-05,
+      "loss": 1.1314,
+      "step": 4914
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897582738948155e-05,
+      "loss": 1.0014,
+      "step": 4915
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8975330628903782e-05,
+      "loss": 1.0443,
+      "step": 4916
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974833754387326e-05,
+      "loss": 1.0469,
+      "step": 4917
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8974336765938495e-05,
+      "loss": 1.1489,
+      "step": 4918
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973839663563593e-05,
+      "loss": 1.0555,
+      "step": 4919
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8973342447268936e-05,
+      "loss": 1.1384,
+      "step": 4920
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972845117060833e-05,
+      "loss": 1.0164,
+      "step": 4921
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8972347672945596e-05,
+      "loss": 1.0103,
+      "step": 4922
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.897185011492954e-05,
+      "loss": 0.9023,
+      "step": 4923
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8971352443018985e-05,
+      "loss": 1.09,
+      "step": 4924
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970854657220246e-05,
+      "loss": 1.0827,
+      "step": 4925
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8970356757539644e-05,
+      "loss": 1.1484,
+      "step": 4926
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969858743983496e-05,
+      "loss": 1.1388,
+      "step": 4927
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8969360616558127e-05,
+      "loss": 1.1081,
+      "step": 4928
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896886237526986e-05,
+      "loss": 1.1371,
+      "step": 4929
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8968364020125023e-05,
+      "loss": 1.1493,
+      "step": 4930
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967865551129935e-05,
+      "loss": 1.0412,
+      "step": 4931
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8967366968290936e-05,
+      "loss": 1.1075,
+      "step": 4932
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896686827161434e-05,
+      "loss": 1.1836,
+      "step": 4933
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8966369461106485e-05,
+      "loss": 1.2985,
+      "step": 4934
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896587053677371e-05,
+      "loss": 1.0892,
+      "step": 4935
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8965371498622337e-05,
+      "loss": 1.2451,
+      "step": 4936
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964872346658712e-05,
+      "loss": 1.2185,
+      "step": 4937
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8964373080889163e-05,
+      "loss": 1.1459,
+      "step": 4938
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896387370132003e-05,
+      "loss": 1.1433,
+      "step": 4939
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8963374207957655e-05,
+      "loss": 1.1467,
+      "step": 4940
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8962874600808377e-05,
+      "loss": 1.0445,
+      "step": 4941
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896237487987854e-05,
+      "loss": 1.0489,
+      "step": 4942
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8961875045174483e-05,
+      "loss": 1.0747,
+      "step": 4943
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896137509670256e-05,
+      "loss": 1.0896,
+      "step": 4944
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.896087503446911e-05,
+      "loss": 1.1436,
+      "step": 4945
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8960374858480484e-05,
+      "loss": 1.0225,
+      "step": 4946
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8959874568743027e-05,
+      "loss": 1.0548,
+      "step": 4947
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89593741652631e-05,
+      "loss": 1.3141,
+      "step": 4948
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958873648047048e-05,
+      "loss": 1.1153,
+      "step": 4949
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8958373017101227e-05,
+      "loss": 1.0517,
+      "step": 4950
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895787227243199e-05,
+      "loss": 1.0876,
+      "step": 4951
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8957371414045696e-05,
+      "loss": 1.1541,
+      "step": 4952
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8956870441948703e-05,
+      "loss": 0.9722,
+      "step": 4953
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895636935614737e-05,
+      "loss": 1.0348,
+      "step": 4954
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895586815664806e-05,
+      "loss": 1.0356,
+      "step": 4955
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8955366843457132e-05,
+      "loss": 1.186,
+      "step": 4956
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8954865416580956e-05,
+      "loss": 1.0356,
+      "step": 4957
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895436387602589e-05,
+      "loss": 0.9266,
+      "step": 4958
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953862221798306e-05,
+      "loss": 0.9804,
+      "step": 4959
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8953360453904568e-05,
+      "loss": 1.1855,
+      "step": 4960
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952858572351053e-05,
+      "loss": 1.0081,
+      "step": 4961
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8952356577144125e-05,
+      "loss": 1.1489,
+      "step": 4962
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895185446829016e-05,
+      "loss": 1.0108,
+      "step": 4963
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8951352245795532e-05,
+      "loss": 1.0934,
+      "step": 4964
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8950849909666614e-05,
+      "loss": 1.0569,
+      "step": 4965
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.895034745990979e-05,
+      "loss": 1.0811,
+      "step": 4966
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949844896531425e-05,
+      "loss": 1.0886,
+      "step": 4967
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8949342219537916e-05,
+      "loss": 1.0682,
+      "step": 4968
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8948839428935632e-05,
+      "loss": 1.0179,
+      "step": 4969
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894833652473096e-05,
+      "loss": 1.1192,
+      "step": 4970
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894783350693028e-05,
+      "loss": 1.0197,
+      "step": 4971
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8947330375539988e-05,
+      "loss": 1.1539,
+      "step": 4972
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894682713056646e-05,
+      "loss": 1.0912,
+      "step": 4973
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894632377201609e-05,
+      "loss": 1.0601,
+      "step": 4974
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8945820299895266e-05,
+      "loss": 1.0121,
+      "step": 4975
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894531671421038e-05,
+      "loss": 0.997,
+      "step": 4976
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944813014967825e-05,
+      "loss": 1.1452,
+      "step": 4977
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8944309202173997e-05,
+      "loss": 1.068,
+      "step": 4978
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894380527583529e-05,
+      "loss": 0.9922,
+      "step": 4979
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89433012359581e-05,
+      "loss": 1.1201,
+      "step": 4980
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8942797082548825e-05,
+      "loss": 1.1993,
+      "step": 4981
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.894229281561387e-05,
+      "loss": 1.0527,
+      "step": 4982
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941788435159632e-05,
+      "loss": 1.1628,
+      "step": 4983
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8941283941192516e-05,
+      "loss": 0.936,
+      "step": 4984
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940779333718925e-05,
+      "loss": 1.0589,
+      "step": 4985
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8940274612745265e-05,
+      "loss": 0.9686,
+      "step": 4986
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8939769778277944e-05,
+      "loss": 1.2151,
+      "step": 4987
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893926483032337e-05,
+      "loss": 1.2083,
+      "step": 4988
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938759768887954e-05,
+      "loss": 1.2901,
+      "step": 4989
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8938254593978106e-05,
+      "loss": 1.0949,
+      "step": 4990
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937749305600245e-05,
+      "loss": 1.0364,
+      "step": 4991
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8937243903760776e-05,
+      "loss": 1.236,
+      "step": 4992
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893673838846612e-05,
+      "loss": 1.0763,
+      "step": 4993
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8936232759722696e-05,
+      "loss": 1.0252,
+      "step": 4994
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893572701753692e-05,
+      "loss": 1.2325,
+      "step": 4995
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8935221161915212e-05,
+      "loss": 1.0487,
+      "step": 4996
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8934715192863992e-05,
+      "loss": 1.0817,
+      "step": 4997
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893420911038969e-05,
+      "loss": 1.0887,
+      "step": 4998
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933702914498725e-05,
+      "loss": 1.0371,
+      "step": 4999
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8933196605197525e-05,
+      "loss": 1.1371,
+      "step": 5000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932690182492515e-05,
+      "loss": 1.1464,
+      "step": 5001
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8932183646390126e-05,
+      "loss": 1.0342,
+      "step": 5002
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8931676996896787e-05,
+      "loss": 0.9732,
+      "step": 5003
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893117023401893e-05,
+      "loss": 1.1143,
+      "step": 5004
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.893066335776299e-05,
+      "loss": 0.9953,
+      "step": 5005
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89301563681354e-05,
+      "loss": 1.1459,
+      "step": 5006
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929649265142597e-05,
+      "loss": 1.0009,
+      "step": 5007
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8929142048791015e-05,
+      "loss": 1.1402,
+      "step": 5008
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89286347190871e-05,
+      "loss": 1.1089,
+      "step": 5009
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8928127276037283e-05,
+      "loss": 1.1108,
+      "step": 5010
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927619719648015e-05,
+      "loss": 1.0253,
+      "step": 5011
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8927112049925735e-05,
+      "loss": 1.0085,
+      "step": 5012
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8926604266876887e-05,
+      "loss": 1.1224,
+      "step": 5013
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892609637050792e-05,
+      "loss": 1.1012,
+      "step": 5014
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925588360825276e-05,
+      "loss": 1.0568,
+      "step": 5015
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8925080237835408e-05,
+      "loss": 1.1435,
+      "step": 5016
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924572001544766e-05,
+      "loss": 1.1132,
+      "step": 5017
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8924063651959805e-05,
+      "loss": 0.9599,
+      "step": 5018
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923555189086973e-05,
+      "loss": 1.0866,
+      "step": 5019
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8923046612932726e-05,
+      "loss": 1.0749,
+      "step": 5020
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8922537923503524e-05,
+      "loss": 1.0272,
+      "step": 5021
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.892202912080582e-05,
+      "loss": 1.1259,
+      "step": 5022
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921520204846075e-05,
+      "loss": 0.9973,
+      "step": 5023
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8921011175630747e-05,
+      "loss": 1.1884,
+      "step": 5024
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8920502033166305e-05,
+      "loss": 1.0087,
+      "step": 5025
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919992777459202e-05,
+      "loss": 1.1236,
+      "step": 5026
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8919483408515914e-05,
+      "loss": 1.18,
+      "step": 5027
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.89189739263429e-05,
+      "loss": 0.9869,
+      "step": 5028
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.891846433094663e-05,
+      "loss": 0.9534,
+      "step": 5029
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 1.8917954622333573e-05,
+      "loss": 1.038,
+      "step": 5030
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8917444800510198e-05,
+      "loss": 1.0096,
+      "step": 5031
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891693486548298e-05,
+      "loss": 1.1128,
+      "step": 5032
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8916424817258387e-05,
+      "loss": 1.1022,
+      "step": 5033
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.89159146558429e-05,
+      "loss": 1.1501,
+      "step": 5034
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8915404381242993e-05,
+      "loss": 1.0992,
+      "step": 5035
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914893993465142e-05,
+      "loss": 0.8938,
+      "step": 5036
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8914383492515828e-05,
+      "loss": 1.004,
+      "step": 5037
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913872878401537e-05,
+      "loss": 1.0915,
+      "step": 5038
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8913362151128737e-05,
+      "loss": 1.0543,
+      "step": 5039
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8912851310703925e-05,
+      "loss": 1.1442,
+      "step": 5040
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891234035713358e-05,
+      "loss": 1.0549,
+      "step": 5041
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891182929042419e-05,
+      "loss": 1.0998,
+      "step": 5042
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.891131811058224e-05,
+      "loss": 1.0899,
+      "step": 5043
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910806817614228e-05,
+      "loss": 1.244,
+      "step": 5044
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8910295411526628e-05,
+      "loss": 1.0494,
+      "step": 5045
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890978389232595e-05,
+      "loss": 1.1183,
+      "step": 5046
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8909272260018676e-05,
+      "loss": 1.1572,
+      "step": 5047
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908760514611308e-05,
+      "loss": 1.1426,
+      "step": 5048
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8908248656110336e-05,
+      "loss": 1.1002,
+      "step": 5049
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907736684522263e-05,
+      "loss": 1.1388,
+      "step": 5050
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8907224599853585e-05,
+      "loss": 1.074,
+      "step": 5051
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906712402110805e-05,
+      "loss": 1.1441,
+      "step": 5052
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8906200091300423e-05,
+      "loss": 1.0913,
+      "step": 5053
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905687667428947e-05,
+      "loss": 1.0191,
+      "step": 5054
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8905175130502876e-05,
+      "loss": 1.1439,
+      "step": 5055
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904662480528718e-05,
+      "loss": 1.1252,
+      "step": 5056
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8904149717512985e-05,
+      "loss": 1.1234,
+      "step": 5057
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903636841462183e-05,
+      "loss": 1.1309,
+      "step": 5058
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8903123852382826e-05,
+      "loss": 1.0986,
+      "step": 5059
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902610750281417e-05,
+      "loss": 1.1866,
+      "step": 5060
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8902097535164484e-05,
+      "loss": 0.9924,
+      "step": 5061
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890158420703853e-05,
+      "loss": 1.1682,
+      "step": 5062
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.890107076591008e-05,
+      "loss": 1.217,
+      "step": 5063
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900557211785642e-05,
+      "loss": 1.0095,
+      "step": 5064
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8900043544671746e-05,
+      "loss": 1.1063,
+      "step": 5065
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899529764574903e-05,
+      "loss": 1.138,
+      "step": 5066
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8899015871501646e-05,
+      "loss": 1.2444,
+      "step": 5067
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8898501865458487e-05,
+      "loss": 1.038,
+      "step": 5068
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889798774645196e-05,
+      "loss": 1.0126,
+      "step": 5069
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8897473514488594e-05,
+      "loss": 1.0729,
+      "step": 5070
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896959169574905e-05,
+      "loss": 1.2531,
+      "step": 5071
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8896444711717433e-05,
+      "loss": 1.1359,
+      "step": 5072
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895930140922703e-05,
+      "loss": 1.1566,
+      "step": 5073
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8895415457197248e-05,
+      "loss": 0.97,
+      "step": 5074
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894900660547603e-05,
+      "loss": 1.0623,
+      "step": 5075
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8894385750980308e-05,
+      "loss": 1.1677,
+      "step": 5076
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889387072850189e-05,
+      "loss": 1.0812,
+      "step": 5077
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8893355593118895e-05,
+      "loss": 1.1118,
+      "step": 5078
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8892840344837856e-05,
+      "loss": 0.8905,
+      "step": 5079
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889232498366532e-05,
+      "loss": 1.0337,
+      "step": 5080
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8891809509607824e-05,
+      "loss": 1.0722,
+      "step": 5081
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.889129392267192e-05,
+      "loss": 1.03,
+      "step": 5082
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890778222864142e-05,
+      "loss": 1.1213,
+      "step": 5083
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8890262410191043e-05,
+      "loss": 1.1163,
+      "step": 5084
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888974648465917e-05,
+      "loss": 1.0785,
+      "step": 5085
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8889230446275074e-05,
+      "loss": 1.1504,
+      "step": 5086
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888714295045302e-05,
+      "loss": 0.9715,
+      "step": 5087
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8888198030976413e-05,
+      "loss": 1.0469,
+      "step": 5088
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8887681654074952e-05,
+      "loss": 1.0159,
+      "step": 5089
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888716516434748e-05,
+      "loss": 1.1886,
+      "step": 5090
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886648561800556e-05,
+      "loss": 1.0719,
+      "step": 5091
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8886131846440732e-05,
+      "loss": 1.0756,
+      "step": 5092
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885615018274572e-05,
+      "loss": 1.0595,
+      "step": 5093
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8885098077308634e-05,
+      "loss": 1.1933,
+      "step": 5094
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888458102354948e-05,
+      "loss": 1.1769,
+      "step": 5095
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888406385700368e-05,
+      "loss": 0.9817,
+      "step": 5096
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888354657767779e-05,
+      "loss": 1.1088,
+      "step": 5097
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8883029185578384e-05,
+      "loss": 1.0292,
+      "step": 5098
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888251168071203e-05,
+      "loss": 1.0937,
+      "step": 5099
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888199406308529e-05,
+      "loss": 1.1804,
+      "step": 5100
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.888147633270474e-05,
+      "loss": 1.0968,
+      "step": 5101
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880958489576952e-05,
+      "loss": 1.0625,
+      "step": 5102
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8880440533708503e-05,
+      "loss": 1.1359,
+      "step": 5103
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8879922465105966e-05,
+      "loss": 1.2418,
+      "step": 5104
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8879404283775914e-05,
+      "loss": 1.1773,
+      "step": 5105
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878885989724933e-05,
+      "loss": 1.0927,
+      "step": 5106
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8878367582959593e-05,
+      "loss": 1.0797,
+      "step": 5107
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8877849063486483e-05,
+      "loss": 1.0765,
+      "step": 5108
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887733043131218e-05,
+      "loss": 1.1178,
+      "step": 5109
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887681168644327e-05,
+      "loss": 1.0984,
+      "step": 5110
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887629282888634e-05,
+      "loss": 1.0904,
+      "step": 5111
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875773858647974e-05,
+      "loss": 1.1067,
+      "step": 5112
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8875254775734765e-05,
+      "loss": 1.2244,
+      "step": 5113
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8874735580153292e-05,
+      "loss": 1.0736,
+      "step": 5114
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887421627191016e-05,
+      "loss": 1.1745,
+      "step": 5115
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887369685101195e-05,
+      "loss": 1.0493,
+      "step": 5116
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887317731746526e-05,
+      "loss": 0.9042,
+      "step": 5117
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.887265767127669e-05,
+      "loss": 1.0928,
+      "step": 5118
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8872137912452825e-05,
+      "loss": 1.0247,
+      "step": 5119
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871618041000274e-05,
+      "loss": 0.9878,
+      "step": 5120
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8871098056925633e-05,
+      "loss": 1.1317,
+      "step": 5121
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870577960235505e-05,
+      "loss": 0.9469,
+      "step": 5122
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8870057750936487e-05,
+      "loss": 1.0623,
+      "step": 5123
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886953742903519e-05,
+      "loss": 0.9926,
+      "step": 5124
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886901699453821e-05,
+      "loss": 1.0834,
+      "step": 5125
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8868496447452164e-05,
+      "loss": 1.0618,
+      "step": 5126
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867975787783655e-05,
+      "loss": 1.0451,
+      "step": 5127
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8867455015539292e-05,
+      "loss": 1.1437,
+      "step": 5128
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8866934130725685e-05,
+      "loss": 1.1702,
+      "step": 5129
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886641313334945e-05,
+      "loss": 1.1875,
+      "step": 5130
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88658920234172e-05,
+      "loss": 1.1419,
+      "step": 5131
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886537080093555e-05,
+      "loss": 1.0016,
+      "step": 5132
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864849465911116e-05,
+      "loss": 1.0425,
+      "step": 5133
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8864328018350516e-05,
+      "loss": 1.091,
+      "step": 5134
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863806458260373e-05,
+      "loss": 1.0573,
+      "step": 5135
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8863284785647303e-05,
+      "loss": 0.9914,
+      "step": 5136
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886276300051793e-05,
+      "loss": 1.0807,
+      "step": 5137
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8862241102878876e-05,
+      "loss": 1.1232,
+      "step": 5138
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861719092736773e-05,
+      "loss": 1.171,
+      "step": 5139
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8861196970098246e-05,
+      "loss": 1.2305,
+      "step": 5140
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886067473496992e-05,
+      "loss": 1.0751,
+      "step": 5141
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.886015238735842e-05,
+      "loss": 1.0411,
+      "step": 5142
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859629927270387e-05,
+      "loss": 1.1359,
+      "step": 5143
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8859107354712444e-05,
+      "loss": 0.9746,
+      "step": 5144
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8858584669691233e-05,
+      "loss": 1.0686,
+      "step": 5145
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885806187221339e-05,
+      "loss": 1.093,
+      "step": 5146
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885753896228554e-05,
+      "loss": 1.1149,
+      "step": 5147
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8857015939914335e-05,
+      "loss": 1.1125,
+      "step": 5148
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8856492805106405e-05,
+      "loss": 1.1211,
+      "step": 5149
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855969557868397e-05,
+      "loss": 1.0499,
+      "step": 5150
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8855446198206948e-05,
+      "loss": 1.1391,
+      "step": 5151
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885492272612871e-05,
+      "loss": 1.1531,
+      "step": 5152
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8854399141640315e-05,
+      "loss": 1.0843,
+      "step": 5153
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853875444748423e-05,
+      "loss": 1.1974,
+      "step": 5154
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8853351635459676e-05,
+      "loss": 1.088,
+      "step": 5155
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8852827713780727e-05,
+      "loss": 1.1446,
+      "step": 5156
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885230367971822e-05,
+      "loss": 1.0701,
+      "step": 5157
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.885177953327882e-05,
+      "loss": 1.1347,
+      "step": 5158
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8851255274469165e-05,
+      "loss": 1.1836,
+      "step": 5159
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850730903295916e-05,
+      "loss": 1.1012,
+      "step": 5160
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8850206419765738e-05,
+      "loss": 1.1756,
+      "step": 5161
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8849681823885283e-05,
+      "loss": 1.1225,
+      "step": 5162
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884915711566121e-05,
+      "loss": 1.28,
+      "step": 5163
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884863229510018e-05,
+      "loss": 1.186,
+      "step": 5164
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884810736220885e-05,
+      "loss": 1.1184,
+      "step": 5165
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8847582316993898e-05,
+      "loss": 1.1106,
+      "step": 5166
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884705715946198e-05,
+      "loss": 1.0323,
+      "step": 5167
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884653188961976e-05,
+      "loss": 1.1662,
+      "step": 5168
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884600650747391e-05,
+      "loss": 1.1441,
+      "step": 5169
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8845481013031103e-05,
+      "loss": 1.1418,
+      "step": 5170
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844955406298005e-05,
+      "loss": 1.162,
+      "step": 5171
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8844429687281288e-05,
+      "loss": 1.1468,
+      "step": 5172
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884390385598763e-05,
+      "loss": 0.9778,
+      "step": 5173
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8843377912423698e-05,
+      "loss": 1.1195,
+      "step": 5174
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884285185659618e-05,
+      "loss": 1.1267,
+      "step": 5175
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.884232568851175e-05,
+      "loss": 1.1137,
+      "step": 5176
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841799408177083e-05,
+      "loss": 1.1038,
+      "step": 5177
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8841273015598865e-05,
+      "loss": 1.1425,
+      "step": 5178
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840746510783774e-05,
+      "loss": 1.1319,
+      "step": 5179
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8840219893738498e-05,
+      "loss": 0.9749,
+      "step": 5180
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839693164469718e-05,
+      "loss": 1.2337,
+      "step": 5181
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8839166322984126e-05,
+      "loss": 1.0788,
+      "step": 5182
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8838639369288404e-05,
+      "loss": 0.9568,
+      "step": 5183
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883811230338925e-05,
+      "loss": 1.0067,
+      "step": 5184
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837585125293344e-05,
+      "loss": 1.1105,
+      "step": 5185
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8837057835007385e-05,
+      "loss": 1.1724,
+      "step": 5186
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836530432538066e-05,
+      "loss": 1.0786,
+      "step": 5187
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8836002917892084e-05,
+      "loss": 1.0474,
+      "step": 5188
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883547529107613e-05,
+      "loss": 1.0461,
+      "step": 5189
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883494755209691e-05,
+      "loss": 1.1079,
+      "step": 5190
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8834419700961115e-05,
+      "loss": 1.0371,
+      "step": 5191
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833891737675452e-05,
+      "loss": 0.8834,
+      "step": 5192
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8833363662246622e-05,
+      "loss": 1.0355,
+      "step": 5193
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8832835474681325e-05,
+      "loss": 1.1377,
+      "step": 5194
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.883230717498627e-05,
+      "loss": 0.992,
+      "step": 5195
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831778763168166e-05,
+      "loss": 1.1471,
+      "step": 5196
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8831250239233713e-05,
+      "loss": 1.1017,
+      "step": 5197
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830721603189624e-05,
+      "loss": 1.0557,
+      "step": 5198
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8830192855042616e-05,
+      "loss": 1.1434,
+      "step": 5199
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829663994799396e-05,
+      "loss": 1.0613,
+      "step": 5200
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8829135022466673e-05,
+      "loss": 1.1228,
+      "step": 5201
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882860593805117e-05,
+      "loss": 1.0614,
+      "step": 5202
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88280767415596e-05,
+      "loss": 1.0102,
+      "step": 5203
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827547432998683e-05,
+      "loss": 1.1027,
+      "step": 5204
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8827018012375135e-05,
+      "loss": 1.2234,
+      "step": 5205
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882648847969568e-05,
+      "loss": 1.086,
+      "step": 5206
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825958834967038e-05,
+      "loss": 1.1462,
+      "step": 5207
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8825429078195934e-05,
+      "loss": 1.0411,
+      "step": 5208
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8824899209389093e-05,
+      "loss": 1.0603,
+      "step": 5209
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882436922855324e-05,
+      "loss": 0.9854,
+      "step": 5210
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823839135695103e-05,
+      "loss": 1.0289,
+      "step": 5211
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8823308930821414e-05,
+      "loss": 1.0893,
+      "step": 5212
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822778613938905e-05,
+      "loss": 1.1173,
+      "step": 5213
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8822248185054302e-05,
+      "loss": 1.0873,
+      "step": 5214
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8821717644174343e-05,
+      "loss": 1.0495,
+      "step": 5215
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882118699130576e-05,
+      "loss": 1.0537,
+      "step": 5216
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8820656226455292e-05,
+      "loss": 0.9816,
+      "step": 5217
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.882012534962968e-05,
+      "loss": 1.114,
+      "step": 5218
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819594360835655e-05,
+      "loss": 1.0147,
+      "step": 5219
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8819063260079965e-05,
+      "loss": 1.0735,
+      "step": 5220
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818532047369352e-05,
+      "loss": 1.0526,
+      "step": 5221
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8818000722710552e-05,
+      "loss": 1.0636,
+      "step": 5222
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881746928611032e-05,
+      "loss": 1.0986,
+      "step": 5223
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816937737575393e-05,
+      "loss": 1.0501,
+      "step": 5224
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8816406077112526e-05,
+      "loss": 1.0737,
+      "step": 5225
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881587430472846e-05,
+      "loss": 1.1014,
+      "step": 5226
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8815342420429958e-05,
+      "loss": 1.2964,
+      "step": 5227
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8814810424223765e-05,
+      "loss": 1.0568,
+      "step": 5228
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881427831611663e-05,
+      "loss": 1.0813,
+      "step": 5229
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813746096115314e-05,
+      "loss": 1.1686,
+      "step": 5230
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8813213764226573e-05,
+      "loss": 1.0029,
+      "step": 5231
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812681320457167e-05,
+      "loss": 1.0333,
+      "step": 5232
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8812148764813848e-05,
+      "loss": 1.1414,
+      "step": 5233
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881161609730338e-05,
+      "loss": 1.046,
+      "step": 5234
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881108331793253e-05,
+      "loss": 1.0947,
+      "step": 5235
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8810550426708052e-05,
+      "loss": 1.1361,
+      "step": 5236
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.881001742363672e-05,
+      "loss": 1.0925,
+      "step": 5237
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8809484308725294e-05,
+      "loss": 1.118,
+      "step": 5238
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8808951081980543e-05,
+      "loss": 1.077,
+      "step": 5239
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880841774340924e-05,
+      "loss": 1.1664,
+      "step": 5240
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880788429301815e-05,
+      "loss": 1.2186,
+      "step": 5241
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8807350730814052e-05,
+      "loss": 1.0758,
+      "step": 5242
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806817056803707e-05,
+      "loss": 1.1207,
+      "step": 5243
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8806283270993907e-05,
+      "loss": 1.0395,
+      "step": 5244
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8805749373391414e-05,
+      "loss": 1.1538,
+      "step": 5245
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880521536400301e-05,
+      "loss": 1.0559,
+      "step": 5246
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8804681242835477e-05,
+      "loss": 1.1239,
+      "step": 5247
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880414700989559e-05,
+      "loss": 1.1676,
+      "step": 5248
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8803612665190136e-05,
+      "loss": 1.1834,
+      "step": 5249
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.88030782087259e-05,
+      "loss": 1.1443,
+      "step": 5250
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802543640509655e-05,
+      "loss": 1.1092,
+      "step": 5251
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8802008960548202e-05,
+      "loss": 1.1183,
+      "step": 5252
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.880147416884832e-05,
+      "loss": 1.1594,
+      "step": 5253
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800939265416796e-05,
+      "loss": 1.1899,
+      "step": 5254
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8800404250260433e-05,
+      "loss": 1.1254,
+      "step": 5255
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8799869123386007e-05,
+      "loss": 1.0704,
+      "step": 5256
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879933388480032e-05,
+      "loss": 1.2336,
+      "step": 5257
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8798798534510165e-05,
+      "loss": 0.9978,
+      "step": 5258
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879826307252234e-05,
+      "loss": 1.1819,
+      "step": 5259
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797727498843637e-05,
+      "loss": 1.0529,
+      "step": 5260
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8797191813480863e-05,
+      "loss": 1.0606,
+      "step": 5261
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879665601644081e-05,
+      "loss": 1.0783,
+      "step": 5262
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8796120107730284e-05,
+      "loss": 1.0022,
+      "step": 5263
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879558408735609e-05,
+      "loss": 1.1685,
+      "step": 5264
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879504795532503e-05,
+      "loss": 1.0376,
+      "step": 5265
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8794511711643907e-05,
+      "loss": 1.0925,
+      "step": 5266
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8793975356319535e-05,
+      "loss": 1.1306,
+      "step": 5267
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879343888935872e-05,
+      "loss": 1.1919,
+      "step": 5268
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879290231076827e-05,
+      "loss": 1.0581,
+      "step": 5269
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8792365620555003e-05,
+      "loss": 0.9597,
+      "step": 5270
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879182881872572e-05,
+      "loss": 1.1501,
+      "step": 5271
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879129190528725e-05,
+      "loss": 0.9856,
+      "step": 5272
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8790754880246397e-05,
+      "loss": 1.0322,
+      "step": 5273
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.879021774360999e-05,
+      "loss": 1.1017,
+      "step": 5274
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789680495384837e-05,
+      "loss": 1.1154,
+      "step": 5275
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8789143135577764e-05,
+      "loss": 1.0925,
+      "step": 5276
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878860566419559e-05,
+      "loss": 1.2294,
+      "step": 5277
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8788068081245137e-05,
+      "loss": 1.1333,
+      "step": 5278
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8787530386733232e-05,
+      "loss": 1.0508,
+      "step": 5279
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786992580666703e-05,
+      "loss": 1.0702,
+      "step": 5280
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8786454663052375e-05,
+      "loss": 1.1149,
+      "step": 5281
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785916633897072e-05,
+      "loss": 1.1315,
+      "step": 5282
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8785378493207632e-05,
+      "loss": 1.0419,
+      "step": 5283
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878484024099088e-05,
+      "loss": 0.9499,
+      "step": 5284
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8784301877253655e-05,
+      "loss": 1.0189,
+      "step": 5285
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8783763402002786e-05,
+      "loss": 1.1447,
+      "step": 5286
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.878322481524511e-05,
+      "loss": 1.0145,
+      "step": 5287
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782686116987468e-05,
+      "loss": 1.132,
+      "step": 5288
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8782147307236695e-05,
+      "loss": 1.2155,
+      "step": 5289
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781608385999628e-05,
+      "loss": 1.1882,
+      "step": 5290
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8781069353283118e-05,
+      "loss": 0.9369,
+      "step": 5291
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8780530209093996e-05,
+      "loss": 1.0211,
+      "step": 5292
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779990953439116e-05,
+      "loss": 1.0026,
+      "step": 5293
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8779451586325318e-05,
+      "loss": 1.0963,
+      "step": 5294
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877891210775945e-05,
+      "loss": 1.1411,
+      "step": 5295
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8778372517748364e-05,
+      "loss": 1.2094,
+      "step": 5296
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8777832816298905e-05,
+      "loss": 1.0654,
+      "step": 5297
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877729300341793e-05,
+      "loss": 1.0614,
+      "step": 5298
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776753079112285e-05,
+      "loss": 1.0099,
+      "step": 5299
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8776213043388826e-05,
+      "loss": 1.1202,
+      "step": 5300
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8775672896254412e-05,
+      "loss": 1.0531,
+      "step": 5301
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.87751326377159e-05,
+      "loss": 1.0416,
+      "step": 5302
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774592267780142e-05,
+      "loss": 1.1381,
+      "step": 5303
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8774051786454008e-05,
+      "loss": 1.0852,
+      "step": 5304
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.877351119374435e-05,
+      "loss": 1.0652,
+      "step": 5305
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772970489658033e-05,
+      "loss": 1.1054,
+      "step": 5306
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8772429674201925e-05,
+      "loss": 0.999,
+      "step": 5307
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771888747382885e-05,
+      "loss": 1.1112,
+      "step": 5308
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8771347709207788e-05,
+      "loss": 1.1814,
+      "step": 5309
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770806559683494e-05,
+      "loss": 1.0915,
+      "step": 5310
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8770265298816876e-05,
+      "loss": 1.1239,
+      "step": 5311
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8769723926614807e-05,
+      "loss": 1.0039,
+      "step": 5312
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.876918244308416e-05,
+      "loss": 1.0642,
+      "step": 5313
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768640848231805e-05,
+      "loss": 1.1374,
+      "step": 5314
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8768099142064618e-05,
+      "loss": 1.1238,
+      "step": 5315
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767557324589482e-05,
+      "loss": 1.091,
+      "step": 5316
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8767015395813268e-05,
+      "loss": 1.0744,
+      "step": 5317
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 1.8766473355742854e-05,
+      "loss": 1.0838,
+      "step": 5318
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8765931204385128e-05,
+      "loss": 0.977,
+      "step": 5319
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876538894174697e-05,
+      "loss": 1.0946,
+      "step": 5320
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8764846567835265e-05,
+      "loss": 1.0943,
+      "step": 5321
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876430408265689e-05,
+      "loss": 1.0549,
+      "step": 5322
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763761486218743e-05,
+      "loss": 1.1046,
+      "step": 5323
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8763218778527707e-05,
+      "loss": 1.2221,
+      "step": 5324
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876267595959067e-05,
+      "loss": 1.1492,
+      "step": 5325
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876213302941453e-05,
+      "loss": 1.1272,
+      "step": 5326
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.876158998800617e-05,
+      "loss": 1.2335,
+      "step": 5327
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8761046835372483e-05,
+      "loss": 1.0938,
+      "step": 5328
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8760503571520373e-05,
+      "loss": 1.1189,
+      "step": 5329
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759960196456733e-05,
+      "loss": 1.1593,
+      "step": 5330
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8759416710188458e-05,
+      "loss": 1.1384,
+      "step": 5331
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875887311272245e-05,
+      "loss": 1.0026,
+      "step": 5332
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875832940406561e-05,
+      "loss": 1.159,
+      "step": 5333
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757785584224842e-05,
+      "loss": 1.0149,
+      "step": 5334
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8757241653207044e-05,
+      "loss": 1.0599,
+      "step": 5335
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8756697611019127e-05,
+      "loss": 0.8939,
+      "step": 5336
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875615345766799e-05,
+      "loss": 1.2272,
+      "step": 5337
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875560919316055e-05,
+      "loss": 1.1137,
+      "step": 5338
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875506481750371e-05,
+      "loss": 1.1548,
+      "step": 5339
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8754520330704383e-05,
+      "loss": 1.0821,
+      "step": 5340
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753975732769482e-05,
+      "loss": 1.1015,
+      "step": 5341
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8753431023705915e-05,
+      "loss": 1.041,
+      "step": 5342
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87528862035206e-05,
+      "loss": 1.1871,
+      "step": 5343
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8752341272220457e-05,
+      "loss": 1.0829,
+      "step": 5344
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8751796229812403e-05,
+      "loss": 1.1978,
+      "step": 5345
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875125107630335e-05,
+      "loss": 1.1998,
+      "step": 5346
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875070581170023e-05,
+      "loss": 1.1378,
+      "step": 5347
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.875016043600995e-05,
+      "loss": 0.993,
+      "step": 5348
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874961494923945e-05,
+      "loss": 1.0045,
+      "step": 5349
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8749069351395638e-05,
+      "loss": 1.0307,
+      "step": 5350
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8748523642485456e-05,
+      "loss": 1.0706,
+      "step": 5351
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874797782251582e-05,
+      "loss": 1.1865,
+      "step": 5352
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8747431891493667e-05,
+      "loss": 1.068,
+      "step": 5353
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874688584942592e-05,
+      "loss": 1.0091,
+      "step": 5354
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8746339696319515e-05,
+      "loss": 1.0858,
+      "step": 5355
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745793432181387e-05,
+      "loss": 1.0832,
+      "step": 5356
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8745247057018467e-05,
+      "loss": 0.9459,
+      "step": 5357
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874470057083769e-05,
+      "loss": 1.1252,
+      "step": 5358
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8744153973646e-05,
+      "loss": 1.1178,
+      "step": 5359
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874360726545033e-05,
+      "loss": 1.0747,
+      "step": 5360
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8743060446257622e-05,
+      "loss": 1.0962,
+      "step": 5361
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8742513516074818e-05,
+      "loss": 1.05,
+      "step": 5362
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874196647490886e-05,
+      "loss": 0.9822,
+      "step": 5363
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8741419322766694e-05,
+      "loss": 1.1871,
+      "step": 5364
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8740872059655266e-05,
+      "loss": 1.0748,
+      "step": 5365
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.874032468558152e-05,
+      "loss": 1.1248,
+      "step": 5366
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8739777200552408e-05,
+      "loss": 1.0355,
+      "step": 5367
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873922960457488e-05,
+      "loss": 1.0332,
+      "step": 5368
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8738681897655887e-05,
+      "loss": 1.0168,
+      "step": 5369
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873813407980238e-05,
+      "loss": 0.9962,
+      "step": 5370
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737586151021316e-05,
+      "loss": 1.0889,
+      "step": 5371
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8737038111319653e-05,
+      "loss": 0.8712,
+      "step": 5372
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873648996070434e-05,
+      "loss": 1.0604,
+      "step": 5373
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8735941699182344e-05,
+      "loss": 1.0909,
+      "step": 5374
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.873539332676062e-05,
+      "loss": 1.047,
+      "step": 5375
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734844843446135e-05,
+      "loss": 1.1123,
+      "step": 5376
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8734296249245845e-05,
+      "loss": 1.0469,
+      "step": 5377
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733747544166717e-05,
+      "loss": 1.1355,
+      "step": 5378
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8733198728215716e-05,
+      "loss": 1.0939,
+      "step": 5379
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732649801399808e-05,
+      "loss": 1.0868,
+      "step": 5380
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8732100763725966e-05,
+      "loss": 0.9901,
+      "step": 5381
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731551615201156e-05,
+      "loss": 1.204,
+      "step": 5382
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8731002355832354e-05,
+      "loss": 1.0326,
+      "step": 5383
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8730452985626523e-05,
+      "loss": 1.0514,
+      "step": 5384
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729903504590642e-05,
+      "loss": 1.0833,
+      "step": 5385
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8729353912731694e-05,
+      "loss": 1.2173,
+      "step": 5386
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8728804210056646e-05,
+      "loss": 1.1838,
+      "step": 5387
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872825439657248e-05,
+      "loss": 1.0513,
+      "step": 5388
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727704472286176e-05,
+      "loss": 1.0642,
+      "step": 5389
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8727154437204712e-05,
+      "loss": 1.1294,
+      "step": 5390
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726604291335074e-05,
+      "loss": 1.0511,
+      "step": 5391
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8726054034684245e-05,
+      "loss": 1.0426,
+      "step": 5392
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872550366725921e-05,
+      "loss": 1.1278,
+      "step": 5393
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724953189066955e-05,
+      "loss": 1.1426,
+      "step": 5394
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8724402600114473e-05,
+      "loss": 1.0257,
+      "step": 5395
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8723851900408743e-05,
+      "loss": 1.0248,
+      "step": 5396
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.872330108995677e-05,
+      "loss": 1.0383,
+      "step": 5397
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722750168765534e-05,
+      "loss": 1.1217,
+      "step": 5398
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8722199136842034e-05,
+      "loss": 1.0532,
+      "step": 5399
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721647994193262e-05,
+      "loss": 1.1699,
+      "step": 5400
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8721096740826222e-05,
+      "loss": 1.024,
+      "step": 5401
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8720545376747906e-05,
+      "loss": 1.1365,
+      "step": 5402
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719993901965315e-05,
+      "loss": 1.0462,
+      "step": 5403
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8719442316485445e-05,
+      "loss": 1.0249,
+      "step": 5404
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8718890620315306e-05,
+      "loss": 1.2426,
+      "step": 5405
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87183388134619e-05,
+      "loss": 1.1359,
+      "step": 5406
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717786895932228e-05,
+      "loss": 1.1307,
+      "step": 5407
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8717234867733298e-05,
+      "loss": 1.0158,
+      "step": 5408
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871668272887212e-05,
+      "loss": 1.1366,
+      "step": 5409
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87161304793557e-05,
+      "loss": 1.1301,
+      "step": 5410
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871557811919105e-05,
+      "loss": 1.097,
+      "step": 5411
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8715025648385183e-05,
+      "loss": 1.2202,
+      "step": 5412
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871447306694511e-05,
+      "loss": 1.0325,
+      "step": 5413
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871392037487785e-05,
+      "loss": 1.0918,
+      "step": 5414
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8713367572190416e-05,
+      "loss": 1.1183,
+      "step": 5415
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871281465888983e-05,
+      "loss": 0.9875,
+      "step": 5416
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.87122616349831e-05,
+      "loss": 1.0023,
+      "step": 5417
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8711708500477258e-05,
+      "loss": 1.0264,
+      "step": 5418
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.871115525537932e-05,
+      "loss": 1.0365,
+      "step": 5419
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710601899696312e-05,
+      "loss": 1.0909,
+      "step": 5420
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8710048433435257e-05,
+      "loss": 1.1659,
+      "step": 5421
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8709494856603184e-05,
+      "loss": 0.9804,
+      "step": 5422
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708941169207112e-05,
+      "loss": 1.2052,
+      "step": 5423
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8708387371254083e-05,
+      "loss": 1.0347,
+      "step": 5424
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870783346275112e-05,
+      "loss": 1.1507,
+      "step": 5425
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870727944370525e-05,
+      "loss": 1.0001,
+      "step": 5426
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870672531412351e-05,
+      "loss": 1.1307,
+      "step": 5427
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870617107401294e-05,
+      "loss": 1.1295,
+      "step": 5428
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870561672338057e-05,
+      "loss": 1.1574,
+      "step": 5429
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870506226223344e-05,
+      "loss": 1.0815,
+      "step": 5430
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8704507690578585e-05,
+      "loss": 1.1466,
+      "step": 5431
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8703953008423045e-05,
+      "loss": 1.1526,
+      "step": 5432
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.870339821577387e-05,
+      "loss": 1.0216,
+      "step": 5433
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702843312638093e-05,
+      "loss": 1.1234,
+      "step": 5434
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8702288299022765e-05,
+      "loss": 1.1431,
+      "step": 5435
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701733174934924e-05,
+      "loss": 0.9787,
+      "step": 5436
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8701177940381626e-05,
+      "loss": 1.0993,
+      "step": 5437
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700622595369913e-05,
+      "loss": 1.0677,
+      "step": 5438
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8700067139906838e-05,
+      "loss": 1.0182,
+      "step": 5439
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8699511573999452e-05,
+      "loss": 1.0267,
+      "step": 5440
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8698955897654806e-05,
+      "loss": 1.0226,
+      "step": 5441
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869840011087996e-05,
+      "loss": 1.1402,
+      "step": 5442
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869784421368196e-05,
+      "loss": 1.1838,
+      "step": 5443
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8697288206067868e-05,
+      "loss": 1.1038,
+      "step": 5444
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696732088044746e-05,
+      "loss": 1.1485,
+      "step": 5445
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8696175859619647e-05,
+      "loss": 1.1961,
+      "step": 5446
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869561952079964e-05,
+      "loss": 0.9986,
+      "step": 5447
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8695063071591776e-05,
+      "loss": 1.1328,
+      "step": 5448
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869450651200313e-05,
+      "loss": 1.0564,
+      "step": 5449
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869394984204076e-05,
+      "loss": 1.1226,
+      "step": 5450
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8693393061711734e-05,
+      "loss": 0.9436,
+      "step": 5451
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692836171023128e-05,
+      "loss": 0.9442,
+      "step": 5452
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8692279169982002e-05,
+      "loss": 1.0671,
+      "step": 5453
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691722058595428e-05,
+      "loss": 1.0948,
+      "step": 5454
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8691164836870483e-05,
+      "loss": 0.9655,
+      "step": 5455
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.869060750481424e-05,
+      "loss": 0.9649,
+      "step": 5456
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8690050062433768e-05,
+      "loss": 1.1666,
+      "step": 5457
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868949250973615e-05,
+      "loss": 1.0548,
+      "step": 5458
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688934846728462e-05,
+      "loss": 1.081,
+      "step": 5459
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8688377073417783e-05,
+      "loss": 1.0615,
+      "step": 5460
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8687819189811196e-05,
+      "loss": 1.0502,
+      "step": 5461
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868726119591578e-05,
+      "loss": 1.0433,
+      "step": 5462
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686703091738618e-05,
+      "loss": 1.0955,
+      "step": 5463
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8686144877286797e-05,
+      "loss": 1.1773,
+      "step": 5464
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685586552567403e-05,
+      "loss": 1.1876,
+      "step": 5465
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8685028117587528e-05,
+      "loss": 1.1671,
+      "step": 5466
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868446957235425e-05,
+      "loss": 1.1071,
+      "step": 5467
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868391091687467e-05,
+      "loss": 1.1131,
+      "step": 5468
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8683352151155872e-05,
+      "loss": 0.9482,
+      "step": 5469
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868279327520496e-05,
+      "loss": 1.1142,
+      "step": 5470
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868223428902902e-05,
+      "loss": 0.9964,
+      "step": 5471
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868167519263515e-05,
+      "loss": 1.0942,
+      "step": 5472
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8681115986030448e-05,
+      "loss": 1.1377,
+      "step": 5473
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.868055666922201e-05,
+      "loss": 1.135,
+      "step": 5474
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679997242216944e-05,
+      "loss": 0.9833,
+      "step": 5475
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8679437705022345e-05,
+      "loss": 1.0804,
+      "step": 5476
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678878057645317e-05,
+      "loss": 1.1732,
+      "step": 5477
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8678318300092968e-05,
+      "loss": 0.9917,
+      "step": 5478
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86777584323724e-05,
+      "loss": 1.1363,
+      "step": 5479
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8677198454490723e-05,
+      "loss": 1.0782,
+      "step": 5480
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676638366455043e-05,
+      "loss": 1.0639,
+      "step": 5481
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8676078168272472e-05,
+      "loss": 1.0319,
+      "step": 5482
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8675517859950123e-05,
+      "loss": 1.1057,
+      "step": 5483
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674957441495106e-05,
+      "loss": 1.1611,
+      "step": 5484
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8674396912914537e-05,
+      "loss": 1.1598,
+      "step": 5485
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673836274215528e-05,
+      "loss": 1.0676,
+      "step": 5486
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8673275525405204e-05,
+      "loss": 0.9901,
+      "step": 5487
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8672714666490676e-05,
+      "loss": 1.1971,
+      "step": 5488
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.867215369747907e-05,
+      "loss": 1.1637,
+      "step": 5489
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671592618377498e-05,
+      "loss": 1.0347,
+      "step": 5490
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8671031429193095e-05,
+      "loss": 1.0478,
+      "step": 5491
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8670470129932975e-05,
+      "loss": 1.0969,
+      "step": 5492
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8669908720604268e-05,
+      "loss": 0.968,
+      "step": 5493
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8669347201214103e-05,
+      "loss": 1.1494,
+      "step": 5494
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86687855717696e-05,
+      "loss": 0.9763,
+      "step": 5495
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86682238322779e-05,
+      "loss": 1.0111,
+      "step": 5496
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8667661982746125e-05,
+      "loss": 1.0591,
+      "step": 5497
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866710002318141e-05,
+      "loss": 1.0663,
+      "step": 5498
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8666537953590894e-05,
+      "loss": 1.0757,
+      "step": 5499
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665975773981704e-05,
+      "loss": 1.1529,
+      "step": 5500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8665413484360983e-05,
+      "loss": 0.9366,
+      "step": 5501
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8664851084735866e-05,
+      "loss": 1.0648,
+      "step": 5502
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866428857511349e-05,
+      "loss": 0.9897,
+      "step": 5503
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663725955501006e-05,
+      "loss": 1.0023,
+      "step": 5504
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8663163225905544e-05,
+      "loss": 0.988,
+      "step": 5505
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8662600386334255e-05,
+      "loss": 1.106,
+      "step": 5506
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.866203743679428e-05,
+      "loss": 1.1266,
+      "step": 5507
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8661474377292767e-05,
+      "loss": 1.0176,
+      "step": 5508
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660911207836867e-05,
+      "loss": 1.0587,
+      "step": 5509
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8660347928433725e-05,
+      "loss": 0.9917,
+      "step": 5510
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659784539090494e-05,
+      "loss": 1.0765,
+      "step": 5511
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8659221039814323e-05,
+      "loss": 1.0346,
+      "step": 5512
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865865743061237e-05,
+      "loss": 0.9223,
+      "step": 5513
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8658093711491786e-05,
+      "loss": 1.0907,
+      "step": 5514
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865752988245973e-05,
+      "loss": 1.0644,
+      "step": 5515
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8656965943523358e-05,
+      "loss": 1.154,
+      "step": 5516
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865640189468983e-05,
+      "loss": 1.0512,
+      "step": 5517
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86558377359663e-05,
+      "loss": 1.045,
+      "step": 5518
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8655273467359946e-05,
+      "loss": 1.0149,
+      "step": 5519
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865470908887791e-05,
+      "loss": 1.1414,
+      "step": 5520
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8654144600527374e-05,
+      "loss": 1.0359,
+      "step": 5521
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8653580002315496e-05,
+      "loss": 0.9815,
+      "step": 5522
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865301529424944e-05,
+      "loss": 1.2002,
+      "step": 5523
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8652450476336386e-05,
+      "loss": 1.0028,
+      "step": 5524
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8651885548583496e-05,
+      "loss": 1.0669,
+      "step": 5525
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.865132051099794e-05,
+      "loss": 0.9059,
+      "step": 5526
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650755363586895e-05,
+      "loss": 1.1347,
+      "step": 5527
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8650190106357535e-05,
+      "loss": 1.1706,
+      "step": 5528
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649624739317035e-05,
+      "loss": 1.1034,
+      "step": 5529
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8649059262472575e-05,
+      "loss": 1.2876,
+      "step": 5530
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8648493675831326e-05,
+      "loss": 1.1471,
+      "step": 5531
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864792797940048e-05,
+      "loss": 1.0559,
+      "step": 5532
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8647362173187202e-05,
+      "loss": 1.1896,
+      "step": 5533
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864679625719869e-05,
+      "loss": 1.0042,
+      "step": 5534
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864623023144212e-05,
+      "loss": 1.0592,
+      "step": 5535
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864566409592468e-05,
+      "loss": 1.1233,
+      "step": 5536
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8645097850653556e-05,
+      "loss": 1.0682,
+      "step": 5537
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8644531495635934e-05,
+      "loss": 1.0642,
+      "step": 5538
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864396503087901e-05,
+      "loss": 0.9947,
+      "step": 5539
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864339845638997e-05,
+      "loss": 1.1837,
+      "step": 5540
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642831772176006e-05,
+      "loss": 1.1246,
+      "step": 5541
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8642264978244315e-05,
+      "loss": 1.1159,
+      "step": 5542
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8641698074602094e-05,
+      "loss": 0.9947,
+      "step": 5543
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.864113106125653e-05,
+      "loss": 1.0383,
+      "step": 5544
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8640563938214832e-05,
+      "loss": 1.0184,
+      "step": 5545
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8639996705484194e-05,
+      "loss": 1.1274,
+      "step": 5546
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863942936307182e-05,
+      "loss": 0.9602,
+      "step": 5547
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863886191098491e-05,
+      "loss": 1.0347,
+      "step": 5548
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8638294349230662e-05,
+      "loss": 1.0914,
+      "step": 5549
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637726677816294e-05,
+      "loss": 1.0276,
+      "step": 5550
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8637158896748997e-05,
+      "loss": 1.0614,
+      "step": 5551
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636591006035993e-05,
+      "loss": 1.1351,
+      "step": 5552
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8636023005684484e-05,
+      "loss": 1.1527,
+      "step": 5553
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863545489570168e-05,
+      "loss": 1.125,
+      "step": 5554
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634886676094795e-05,
+      "loss": 1.1099,
+      "step": 5555
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8634318346871042e-05,
+      "loss": 1.0939,
+      "step": 5556
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633749908037634e-05,
+      "loss": 1.0465,
+      "step": 5557
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8633181359601794e-05,
+      "loss": 1.0272,
+      "step": 5558
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863261270157073e-05,
+      "loss": 1.0236,
+      "step": 5559
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8632043933951665e-05,
+      "loss": 1.0749,
+      "step": 5560
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.863147505675182e-05,
+      "loss": 1.1493,
+      "step": 5561
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630906069978418e-05,
+      "loss": 1.1441,
+      "step": 5562
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8630336973638676e-05,
+      "loss": 1.0072,
+      "step": 5563
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8629767767739827e-05,
+      "loss": 1.1283,
+      "step": 5564
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862919845228909e-05,
+      "loss": 1.1899,
+      "step": 5565
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8628629027293698e-05,
+      "loss": 1.0927,
+      "step": 5566
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862805949276087e-05,
+      "loss": 1.0885,
+      "step": 5567
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862748984869785e-05,
+      "loss": 1.1288,
+      "step": 5568
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626920095111857e-05,
+      "loss": 1.0817,
+      "step": 5569
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8626350232010132e-05,
+      "loss": 1.1332,
+      "step": 5570
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625780259399906e-05,
+      "loss": 1.0027,
+      "step": 5571
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8625210177288414e-05,
+      "loss": 1.0984,
+      "step": 5572
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862463998568289e-05,
+      "loss": 1.1825,
+      "step": 5573
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862406968459058e-05,
+      "loss": 1.1037,
+      "step": 5574
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862349927401872e-05,
+      "loss": 1.1064,
+      "step": 5575
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.862292875397455e-05,
+      "loss": 1.1796,
+      "step": 5576
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8622358124465313e-05,
+      "loss": 1.0865,
+      "step": 5577
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621787385498257e-05,
+      "loss": 1.0642,
+      "step": 5578
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8621216537080618e-05,
+      "loss": 1.0338,
+      "step": 5579
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8620645579219648e-05,
+      "loss": 1.0219,
+      "step": 5580
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86200745119226e-05,
+      "loss": 1.0489,
+      "step": 5581
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861950333519672e-05,
+      "loss": 1.1705,
+      "step": 5582
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8618932049049255e-05,
+      "loss": 1.2129,
+      "step": 5583
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861836065348746e-05,
+      "loss": 1.0766,
+      "step": 5584
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861778914851859e-05,
+      "loss": 1.0676,
+      "step": 5585
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8617217534149898e-05,
+      "loss": 1.1535,
+      "step": 5586
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616645810388646e-05,
+      "loss": 1.0361,
+      "step": 5587
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8616073977242083e-05,
+      "loss": 1.134,
+      "step": 5588
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861550203471747e-05,
+      "loss": 1.2246,
+      "step": 5589
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614929982822073e-05,
+      "loss": 1.0863,
+      "step": 5590
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8614357821563153e-05,
+      "loss": 1.1612,
+      "step": 5591
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861378555094797e-05,
+      "loss": 1.0481,
+      "step": 5592
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861321317098379e-05,
+      "loss": 1.0321,
+      "step": 5593
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612640681677877e-05,
+      "loss": 0.9905,
+      "step": 5594
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8612068083037503e-05,
+      "loss": 1.0683,
+      "step": 5595
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.861149537506994e-05,
+      "loss": 1.1033,
+      "step": 5596
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610922557782446e-05,
+      "loss": 0.9827,
+      "step": 5597
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8610349631182302e-05,
+      "loss": 1.1496,
+      "step": 5598
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860977659527678e-05,
+      "loss": 1.0994,
+      "step": 5599
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8609203450073155e-05,
+      "loss": 1.0745,
+      "step": 5600
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.86086301955787e-05,
+      "loss": 1.0255,
+      "step": 5601
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860805683180069e-05,
+      "loss": 1.0518,
+      "step": 5602
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.8607483358746413e-05,
+      "loss": 1.0436,
+      "step": 5603
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860690977642314e-05,
+      "loss": 0.9799,
+      "step": 5604
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 1.860633608483816e-05,
+      "loss": 1.0392,
+      "step": 5605
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8605762283998746e-05,
+      "loss": 1.123,
+      "step": 5606
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860518837391219e-05,
+      "loss": 1.1714,
+      "step": 5607
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8604614354585776e-05,
+      "loss": 1.0522,
+      "step": 5608
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860404022602679e-05,
+      "loss": 0.9664,
+      "step": 5609
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860346598824252e-05,
+      "loss": 0.9837,
+      "step": 5610
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602891641240258e-05,
+      "loss": 0.9807,
+      "step": 5611
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8602317185027293e-05,
+      "loss": 1.2666,
+      "step": 5612
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860174261961092e-05,
+      "loss": 1.0933,
+      "step": 5613
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8601167944998427e-05,
+      "loss": 1.0629,
+      "step": 5614
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8600593161197116e-05,
+      "loss": 1.0769,
+      "step": 5615
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.860001826821428e-05,
+      "loss": 1.121,
+      "step": 5616
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859944326605722e-05,
+      "loss": 1.0891,
+      "step": 5617
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8598868154733233e-05,
+      "loss": 1.1144,
+      "step": 5618
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859829293424962e-05,
+      "loss": 1.0805,
+      "step": 5619
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597717604613683e-05,
+      "loss": 0.9504,
+      "step": 5620
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8597142165832727e-05,
+      "loss": 1.1071,
+      "step": 5621
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8596566617914056e-05,
+      "loss": 1.0232,
+      "step": 5622
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595990960864975e-05,
+      "loss": 1.0482,
+      "step": 5623
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8595415194692796e-05,
+      "loss": 1.1192,
+      "step": 5624
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594839319404822e-05,
+      "loss": 1.1393,
+      "step": 5625
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8594263335008373e-05,
+      "loss": 0.9944,
+      "step": 5626
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859368724151075e-05,
+      "loss": 1.1538,
+      "step": 5627
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859311103891927e-05,
+      "loss": 1.2299,
+      "step": 5628
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859253472724125e-05,
+      "loss": 1.1206,
+      "step": 5629
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.859195830648401e-05,
+      "loss": 1.1255,
+      "step": 5630
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8591381776654858e-05,
+      "loss": 1.031,
+      "step": 5631
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590805137761122e-05,
+      "loss": 0.9983,
+      "step": 5632
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8590228389810112e-05,
+      "loss": 1.0888,
+      "step": 5633
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589651532809157e-05,
+      "loss": 1.1056,
+      "step": 5634
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8589074566765578e-05,
+      "loss": 1.1625,
+      "step": 5635
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85884974916867e-05,
+      "loss": 1.0709,
+      "step": 5636
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858792030757985e-05,
+      "loss": 1.08,
+      "step": 5637
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858734301445235e-05,
+      "loss": 1.1392,
+      "step": 5638
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586765612311533e-05,
+      "loss": 1.0298,
+      "step": 5639
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8586188101164732e-05,
+      "loss": 1.1119,
+      "step": 5640
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585610481019268e-05,
+      "loss": 1.2145,
+      "step": 5641
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8585032751882484e-05,
+      "loss": 1.0824,
+      "step": 5642
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8584454913761707e-05,
+      "loss": 1.188,
+      "step": 5643
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583876966664277e-05,
+      "loss": 1.1276,
+      "step": 5644
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8583298910597532e-05,
+      "loss": 1.1401,
+      "step": 5645
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8582720745568804e-05,
+      "loss": 1.0531,
+      "step": 5646
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858214247158544e-05,
+      "loss": 1.2947,
+      "step": 5647
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8581564088654775e-05,
+      "loss": 1.0619,
+      "step": 5648
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.858098559678415e-05,
+      "loss": 1.1029,
+      "step": 5649
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8580406995980915e-05,
+      "loss": 1.0367,
+      "step": 5650
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8579828286252414e-05,
+      "loss": 1.0699,
+      "step": 5651
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857924946760599e-05,
+      "loss": 1.1501,
+      "step": 5652
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857867054004899e-05,
+      "loss": 1.1695,
+      "step": 5653
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857809150358877e-05,
+      "loss": 1.1423,
+      "step": 5654
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8577512358232675e-05,
+      "loss": 1.0554,
+      "step": 5655
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857693310398806e-05,
+      "loss": 1.0586,
+      "step": 5656
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8576353740862275e-05,
+      "loss": 1.0969,
+      "step": 5657
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575774268862682e-05,
+      "loss": 1.0412,
+      "step": 5658
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8575194687996628e-05,
+      "loss": 1.121,
+      "step": 5659
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574614998271478e-05,
+      "loss": 1.1762,
+      "step": 5660
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8574035199694585e-05,
+      "loss": 1.0089,
+      "step": 5661
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857345529227331e-05,
+      "loss": 1.1519,
+      "step": 5662
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857287527601502e-05,
+      "loss": 1.1318,
+      "step": 5663
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8572295150927076e-05,
+      "loss": 1.0025,
+      "step": 5664
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571714917016838e-05,
+      "loss": 0.9509,
+      "step": 5665
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8571134574291674e-05,
+      "loss": 1.0484,
+      "step": 5666
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.857055412275896e-05,
+      "loss": 1.1516,
+      "step": 5667
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856997356242605e-05,
+      "loss": 1.0557,
+      "step": 5668
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8569392893300327e-05,
+      "loss": 1.1034,
+      "step": 5669
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856881211538915e-05,
+      "loss": 1.1026,
+      "step": 5670
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8568231228699903e-05,
+      "loss": 0.9785,
+      "step": 5671
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8567650233239953e-05,
+      "loss": 1.0625,
+      "step": 5672
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856706912901668e-05,
+      "loss": 0.9497,
+      "step": 5673
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8566487916037456e-05,
+      "loss": 1.032,
+      "step": 5674
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565906594309667e-05,
+      "loss": 1.1155,
+      "step": 5675
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8565325163840683e-05,
+      "loss": 1.0793,
+      "step": 5676
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564743624637894e-05,
+      "loss": 1.0749,
+      "step": 5677
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8564161976708675e-05,
+      "loss": 0.9764,
+      "step": 5678
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856358022006041e-05,
+      "loss": 0.9626,
+      "step": 5679
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8562998354700498e-05,
+      "loss": 1.0476,
+      "step": 5680
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856241638063631e-05,
+      "loss": 1.1791,
+      "step": 5681
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8561834297875235e-05,
+      "loss": 1.0053,
+      "step": 5682
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856125210642467e-05,
+      "loss": 0.9956,
+      "step": 5683
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8560669806292e-05,
+      "loss": 1.2192,
+      "step": 5684
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.856008739748462e-05,
+      "loss": 0.9836,
+      "step": 5685
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8559504880009924e-05,
+      "loss": 1.1353,
+      "step": 5686
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558922253875303e-05,
+      "loss": 1.068,
+      "step": 5687
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8558339519088155e-05,
+      "loss": 1.0069,
+      "step": 5688
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855775667565588e-05,
+      "loss": 1.1065,
+      "step": 5689
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8557173723585876e-05,
+      "loss": 1.0938,
+      "step": 5690
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8556590662885542e-05,
+      "loss": 1.1147,
+      "step": 5691
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855600749356228e-05,
+      "loss": 1.1402,
+      "step": 5692
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8555424215623493e-05,
+      "loss": 0.9326,
+      "step": 5693
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554840829076587e-05,
+      "loss": 1.0961,
+      "step": 5694
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8554257333928965e-05,
+      "loss": 1.1983,
+      "step": 5695
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855367373018804e-05,
+      "loss": 1.1261,
+      "step": 5696
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8553090017861214e-05,
+      "loss": 1.0476,
+      "step": 5697
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85525061969559e-05,
+      "loss": 1.1841,
+      "step": 5698
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.855192226747951e-05,
+      "loss": 1.2416,
+      "step": 5699
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8551338229439454e-05,
+      "loss": 1.0406,
+      "step": 5700
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550754082843152e-05,
+      "loss": 1.0449,
+      "step": 5701
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8550169827698013e-05,
+      "loss": 1.1229,
+      "step": 5702
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549585464011457e-05,
+      "loss": 1.1655,
+      "step": 5703
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8549000991790903e-05,
+      "loss": 1.1033,
+      "step": 5704
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854841641104377e-05,
+      "loss": 1.0043,
+      "step": 5705
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547831721777476e-05,
+      "loss": 1.0348,
+      "step": 5706
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8547246923999446e-05,
+      "loss": 1.0185,
+      "step": 5707
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546662017717106e-05,
+      "loss": 1.1871,
+      "step": 5708
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8546077002937882e-05,
+      "loss": 1.0701,
+      "step": 5709
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854549187966919e-05,
+      "loss": 1.1566,
+      "step": 5710
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544906647918472e-05,
+      "loss": 1.0922,
+      "step": 5711
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8544321307693146e-05,
+      "loss": 1.0334,
+      "step": 5712
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543735859000655e-05,
+      "loss": 1.0538,
+      "step": 5713
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8543150301848416e-05,
+      "loss": 1.1238,
+      "step": 5714
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.854256463624387e-05,
+      "loss": 1.0403,
+      "step": 5715
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541978862194457e-05,
+      "loss": 0.9875,
+      "step": 5716
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8541392979707607e-05,
+      "loss": 1.139,
+      "step": 5717
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540806988790755e-05,
+      "loss": 1.2004,
+      "step": 5718
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8540220889451345e-05,
+      "loss": 1.1189,
+      "step": 5719
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539634681696816e-05,
+      "loss": 1.0149,
+      "step": 5720
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8539048365534613e-05,
+      "loss": 0.9877,
+      "step": 5721
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8538461940972173e-05,
+      "loss": 1.0303,
+      "step": 5722
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8537875408016942e-05,
+      "loss": 1.1086,
+      "step": 5723
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853728876667637e-05,
+      "loss": 1.1482,
+      "step": 5724
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536702016957898e-05,
+      "loss": 0.9826,
+      "step": 5725
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8536115158868983e-05,
+      "loss": 1.123,
+      "step": 5726
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853552819241706e-05,
+      "loss": 1.1167,
+      "step": 5727
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.85349411176096e-05,
+      "loss": 0.9039,
+      "step": 5728
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853435393445404e-05,
+      "loss": 1.1993,
+      "step": 5729
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8533766642957844e-05,
+      "loss": 1.0673,
+      "step": 5730
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853317924312846e-05,
+      "loss": 1.0816,
+      "step": 5731
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853259173497335e-05,
+      "loss": 1.2033,
+      "step": 5732
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853200411849997e-05,
+      "loss": 1.0646,
+      "step": 5733
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8531416393715778e-05,
+      "loss": 1.0068,
+      "step": 5734
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.853082856062824e-05,
+      "loss": 1.0541,
+      "step": 5735
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8530240619244812e-05,
+      "loss": 1.0556,
+      "step": 5736
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529652569572962e-05,
+      "loss": 0.9842,
+      "step": 5737
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8529064411620157e-05,
+      "loss": 1.1032,
+      "step": 5738
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8528476145393858e-05,
+      "loss": 1.1107,
+      "step": 5739
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8527887770901535e-05,
+      "loss": 1.1269,
+      "step": 5740
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852729928815066e-05,
+      "loss": 1.1393,
+      "step": 5741
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526710697148696e-05,
+      "loss": 1.0119,
+      "step": 5742
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8526121997903125e-05,
+      "loss": 1.082,
+      "step": 5743
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8525533190421414e-05,
+      "loss": 0.99,
+      "step": 5744
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852494427471104e-05,
+      "loss": 1.1336,
+      "step": 5745
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8524355250779478e-05,
+      "loss": 1.1715,
+      "step": 5746
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523766118634202e-05,
+      "loss": 1.0824,
+      "step": 5747
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8523176878282697e-05,
+      "loss": 1.0442,
+      "step": 5748
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8522587529732442e-05,
+      "loss": 0.987,
+      "step": 5749
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521998072990917e-05,
+      "loss": 1.0974,
+      "step": 5750
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8521408508065602e-05,
+      "loss": 1.0772,
+      "step": 5751
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8520818834963988e-05,
+      "loss": 1.1095,
+      "step": 5752
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.852022905369356e-05,
+      "loss": 1.0805,
+      "step": 5753
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519639164261794e-05,
+      "loss": 1.055,
+      "step": 5754
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8519049166676196e-05,
+      "loss": 1.0227,
+      "step": 5755
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851845906094424e-05,
+      "loss": 1.0877,
+      "step": 5756
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517868847073422e-05,
+      "loss": 1.0727,
+      "step": 5757
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8517278525071244e-05,
+      "loss": 1.0955,
+      "step": 5758
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851668809494519e-05,
+      "loss": 1.1184,
+      "step": 5759
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851609755670275e-05,
+      "loss": 1.2779,
+      "step": 5760
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8515506910351437e-05,
+      "loss": 1.2397,
+      "step": 5761
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514916155898736e-05,
+      "loss": 1.1299,
+      "step": 5762
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8514325293352153e-05,
+      "loss": 1.0088,
+      "step": 5763
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513734322719185e-05,
+      "loss": 1.0383,
+      "step": 5764
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8513143244007334e-05,
+      "loss": 1.0692,
+      "step": 5765
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8512552057224108e-05,
+      "loss": 1.1034,
+      "step": 5766
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.851196076237701e-05,
+      "loss": 1.0421,
+      "step": 5767
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8511369359473545e-05,
+      "loss": 1.0409,
+      "step": 5768
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510777848521217e-05,
+      "loss": 1.1178,
+      "step": 5769
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8510186229527544e-05,
+      "loss": 1.258,
+      "step": 5770
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850959450250003e-05,
+      "loss": 1.2029,
+      "step": 5771
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8509002667446185e-05,
+      "loss": 1.0299,
+      "step": 5772
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8508410724373528e-05,
+      "loss": 1.0209,
+      "step": 5773
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507818673289568e-05,
+      "loss": 1.098,
+      "step": 5774
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8507226514201826e-05,
+      "loss": 1.1301,
+      "step": 5775
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850663424711782e-05,
+      "loss": 1.1104,
+      "step": 5776
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.850604187204506e-05,
+      "loss": 1.0369,
+      "step": 5777
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8505449388991073e-05,
+      "loss": 1.118,
+      "step": 5778
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504856797963383e-05,
+      "loss": 1.0902,
+      "step": 5779
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8504264098969505e-05,
+      "loss": 1.1417,
+      "step": 5780
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503671292016968e-05,
+      "loss": 1.0569,
+      "step": 5781
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8503078377113297e-05,
+      "loss": 1.1028,
+      "step": 5782
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8502485354266014e-05,
+      "loss": 1.1653,
+      "step": 5783
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501892223482655e-05,
+      "loss": 0.9979,
+      "step": 5784
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8501298984770747e-05,
+      "loss": 1.0731,
+      "step": 5785
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500705638137818e-05,
+      "loss": 1.056,
+      "step": 5786
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8500112183591402e-05,
+      "loss": 1.153,
+      "step": 5787
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8499518621139036e-05,
+      "loss": 1.1595,
+      "step": 5788
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498924950788248e-05,
+      "loss": 1.1299,
+      "step": 5789
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8498331172546578e-05,
+      "loss": 1.027,
+      "step": 5790
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8497737286421567e-05,
+      "loss": 1.1308,
+      "step": 5791
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849714329242075e-05,
+      "loss": 1.0913,
+      "step": 5792
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8496549190551675e-05,
+      "loss": 1.0738,
+      "step": 5793
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849595498082187e-05,
+      "loss": 1.0582,
+      "step": 5794
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849536066323889e-05,
+      "loss": 1.0169,
+      "step": 5795
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494766237810275e-05,
+      "loss": 1.1942,
+      "step": 5796
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8494171704543568e-05,
+      "loss": 1.194,
+      "step": 5797
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8493577063446328e-05,
+      "loss": 1.1093,
+      "step": 5798
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849298231452609e-05,
+      "loss": 0.9079,
+      "step": 5799
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8492387457790412e-05,
+      "loss": 1.1915,
+      "step": 5800
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849179249324684e-05,
+      "loss": 1.0062,
+      "step": 5801
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8491197420902935e-05,
+      "loss": 1.1789,
+      "step": 5802
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8490602240766246e-05,
+      "loss": 1.1368,
+      "step": 5803
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.849000695284433e-05,
+      "loss": 1.1454,
+      "step": 5804
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848941155714474e-05,
+      "loss": 1.0814,
+      "step": 5805
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8488816053675043e-05,
+      "loss": 1.0316,
+      "step": 5806
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848822044244279e-05,
+      "loss": 1.0252,
+      "step": 5807
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8487624723455544e-05,
+      "loss": 1.0293,
+      "step": 5808
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848702889672087e-05,
+      "loss": 1.0502,
+      "step": 5809
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848643296224633e-05,
+      "loss": 1.1178,
+      "step": 5810
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8485836920039493e-05,
+      "loss": 1.0537,
+      "step": 5811
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848524077010792e-05,
+      "loss": 1.0924,
+      "step": 5812
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848464451245918e-05,
+      "loss": 1.137,
+      "step": 5813
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8484048147100847e-05,
+      "loss": 1.1643,
+      "step": 5814
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8483451674040487e-05,
+      "loss": 1.1238,
+      "step": 5815
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482855093285672e-05,
+      "loss": 1.1489,
+      "step": 5816
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8482258404843977e-05,
+      "loss": 1.0056,
+      "step": 5817
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481661608722978e-05,
+      "loss": 1.0798,
+      "step": 5818
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8481064704930248e-05,
+      "loss": 1.0653,
+      "step": 5819
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.848046769347337e-05,
+      "loss": 1.0955,
+      "step": 5820
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479870574359913e-05,
+      "loss": 1.1618,
+      "step": 5821
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8479273347597463e-05,
+      "loss": 1.069,
+      "step": 5822
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8478676013193606e-05,
+      "loss": 0.9393,
+      "step": 5823
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847807857115592e-05,
+      "loss": 1.0303,
+      "step": 5824
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847748102149199e-05,
+      "loss": 1.0164,
+      "step": 5825
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476883364209398e-05,
+      "loss": 1.1512,
+      "step": 5826
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8476285599315736e-05,
+      "loss": 0.967,
+      "step": 5827
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475687726818594e-05,
+      "loss": 1.1127,
+      "step": 5828
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8475089746725556e-05,
+      "loss": 1.1198,
+      "step": 5829
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8474491659044216e-05,
+      "loss": 1.0653,
+      "step": 5830
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473893463782167e-05,
+      "loss": 1.0783,
+      "step": 5831
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8473295160947005e-05,
+      "loss": 1.1358,
+      "step": 5832
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847269675054632e-05,
+      "loss": 1.228,
+      "step": 5833
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847209823258771e-05,
+      "loss": 1.1185,
+      "step": 5834
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8471499607078777e-05,
+      "loss": 1.0747,
+      "step": 5835
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8470900874027114e-05,
+      "loss": 1.1524,
+      "step": 5836
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.847030203344033e-05,
+      "loss": 1.0014,
+      "step": 5837
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8469703085326022e-05,
+      "loss": 1.1566,
+      "step": 5838
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846910402969179e-05,
+      "loss": 1.1664,
+      "step": 5839
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8468504866545248e-05,
+      "loss": 1.1437,
+      "step": 5840
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846790559589399e-05,
+      "loss": 1.0183,
+      "step": 5841
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8467306217745637e-05,
+      "loss": 1.1191,
+      "step": 5842
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466706732107785e-05,
+      "loss": 1.0419,
+      "step": 5843
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8466107138988056e-05,
+      "loss": 1.1938,
+      "step": 5844
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8465507438394052e-05,
+      "loss": 1.0646,
+      "step": 5845
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.846490763033339e-05,
+      "loss": 1.1154,
+      "step": 5846
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8464307714813687e-05,
+      "loss": 1.2009,
+      "step": 5847
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463707691842556e-05,
+      "loss": 1.0174,
+      "step": 5848
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8463107561427612e-05,
+      "loss": 1.0869,
+      "step": 5849
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8462507323576476e-05,
+      "loss": 1.1361,
+      "step": 5850
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8461906978296772e-05,
+      "loss": 0.9721,
+      "step": 5851
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8461306525596113e-05,
+      "loss": 1.0705,
+      "step": 5852
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460705965482124e-05,
+      "loss": 1.0878,
+      "step": 5853
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8460105297962433e-05,
+      "loss": 1.0477,
+      "step": 5854
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8459504523044657e-05,
+      "loss": 1.3171,
+      "step": 5855
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458903640736435e-05,
+      "loss": 0.9252,
+      "step": 5856
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8458302651045384e-05,
+      "loss": 1.0983,
+      "step": 5857
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457701553979138e-05,
+      "loss": 1.0936,
+      "step": 5858
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8457100349545327e-05,
+      "loss": 1.0733,
+      "step": 5859
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8456499037751583e-05,
+      "loss": 1.0138,
+      "step": 5860
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845589761860554e-05,
+      "loss": 1.0663,
+      "step": 5861
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845529609211483e-05,
+      "loss": 0.989,
+      "step": 5862
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454694458287096e-05,
+      "loss": 1.1347,
+      "step": 5863
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8454092717129967e-05,
+      "loss": 1.0115,
+      "step": 5864
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8453490868651093e-05,
+      "loss": 1.0738,
+      "step": 5865
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84528889128581e-05,
+      "loss": 0.9972,
+      "step": 5866
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845228684975864e-05,
+      "loss": 1.0089,
+      "step": 5867
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845168467936035e-05,
+      "loss": 1.1653,
+      "step": 5868
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8451082401670878e-05,
+      "loss": 1.0206,
+      "step": 5869
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.845048001669787e-05,
+      "loss": 0.9235,
+      "step": 5870
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449877524448975e-05,
+      "loss": 1.2589,
+      "step": 5871
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8449274924931832e-05,
+      "loss": 1.0276,
+      "step": 5872
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84486722181541e-05,
+      "loss": 0.9622,
+      "step": 5873
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8448069404123426e-05,
+      "loss": 0.9753,
+      "step": 5874
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844746648284747e-05,
+      "loss": 1.0159,
+      "step": 5875
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446863454333873e-05,
+      "loss": 1.1119,
+      "step": 5876
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8446260318590298e-05,
+      "loss": 1.0635,
+      "step": 5877
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.84456570756244e-05,
+      "loss": 1.0988,
+      "step": 5878
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8445053725443843e-05,
+      "loss": 1.0693,
+      "step": 5879
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8444450268056274e-05,
+      "loss": 1.051,
+      "step": 5880
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443846703469367e-05,
+      "loss": 1.1392,
+      "step": 5881
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8443243031690773e-05,
+      "loss": 1.0168,
+      "step": 5882
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844263925272816e-05,
+      "loss": 1.0688,
+      "step": 5883
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844203536658919e-05,
+      "loss": 1.0923,
+      "step": 5884
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844143137328154e-05,
+      "loss": 1.0911,
+      "step": 5885
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.844082727281286e-05,
+      "loss": 1.1217,
+      "step": 5886
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8440223065190833e-05,
+      "loss": 1.1131,
+      "step": 5887
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843961875042313e-05,
+      "loss": 1.193,
+      "step": 5888
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8439014328517408e-05,
+      "loss": 1.0831,
+      "step": 5889
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8438409799481352e-05,
+      "loss": 0.9184,
+      "step": 5890
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.843780516332263e-05,
+      "loss": 1.1481,
+      "step": 5891
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8437200420048928e-05,
+      "loss": 1.1096,
+      "step": 5892
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 1.8436595569667908e-05,
+      "loss": 1.0857,
+      "step": 5893
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843599061218726e-05,
+      "loss": 1.1272,
+      "step": 5894
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843538554761466e-05,
+      "loss": 1.0934,
+      "step": 5895
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8434780375957786e-05,
+      "loss": 1.1374,
+      "step": 5896
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843417509722433e-05,
+      "loss": 1.1472,
+      "step": 5897
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8433569711421964e-05,
+      "loss": 1.0551,
+      "step": 5898
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843296421855838e-05,
+      "loss": 1.1737,
+      "step": 5899
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8432358618641262e-05,
+      "loss": 1.1801,
+      "step": 5900
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84317529116783e-05,
+      "loss": 0.8178,
+      "step": 5901
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.843114709767718e-05,
+      "loss": 1.188,
+      "step": 5902
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8430541176645595e-05,
+      "loss": 0.9753,
+      "step": 5903
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842993514859124e-05,
+      "loss": 1.0882,
+      "step": 5904
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84293290135218e-05,
+      "loss": 1.0066,
+      "step": 5905
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428722771444976e-05,
+      "loss": 1.0305,
+      "step": 5906
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8428116422368464e-05,
+      "loss": 1.0692,
+      "step": 5907
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8427509966299957e-05,
+      "loss": 1.1973,
+      "step": 5908
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842690340324716e-05,
+      "loss": 1.0924,
+      "step": 5909
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8426296733217766e-05,
+      "loss": 1.1397,
+      "step": 5910
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425689956219487e-05,
+      "loss": 0.9681,
+      "step": 5911
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8425083072260012e-05,
+      "loss": 1.1852,
+      "step": 5912
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842447608134706e-05,
+      "loss": 1.0885,
+      "step": 5913
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423868983488323e-05,
+      "loss": 1.0896,
+      "step": 5914
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8423261778691514e-05,
+      "loss": 0.9868,
+      "step": 5915
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842265446696434e-05,
+      "loss": 1.0104,
+      "step": 5916
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8422047048314515e-05,
+      "loss": 1.0754,
+      "step": 5917
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.842143952274974e-05,
+      "loss": 0.9554,
+      "step": 5918
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420831890277743e-05,
+      "loss": 0.9871,
+      "step": 5919
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8420224150906224e-05,
+      "loss": 1.145,
+      "step": 5920
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419616304642904e-05,
+      "loss": 1.2025,
+      "step": 5921
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8419008351495494e-05,
+      "loss": 1.068,
+      "step": 5922
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8418400291471718e-05,
+      "loss": 1.0728,
+      "step": 5923
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8417792124579295e-05,
+      "loss": 0.9816,
+      "step": 5924
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841718385082594e-05,
+      "loss": 1.1506,
+      "step": 5925
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841657547021938e-05,
+      "loss": 1.1579,
+      "step": 5926
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415966982767336e-05,
+      "loss": 1.1105,
+      "step": 5927
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8415358388477534e-05,
+      "loss": 1.1003,
+      "step": 5928
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414749687357698e-05,
+      "loss": 0.9422,
+      "step": 5929
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8414140879415555e-05,
+      "loss": 1.0587,
+      "step": 5930
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8413531964658834e-05,
+      "loss": 1.0995,
+      "step": 5931
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412922943095267e-05,
+      "loss": 1.1496,
+      "step": 5932
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8412313814732585e-05,
+      "loss": 1.121,
+      "step": 5933
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841170457957852e-05,
+      "loss": 1.071,
+      "step": 5934
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8411095237640803e-05,
+      "loss": 1.0833,
+      "step": 5935
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.841048578892717e-05,
+      "loss": 1.1033,
+      "step": 5936
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409876233445363e-05,
+      "loss": 1.0158,
+      "step": 5937
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8409266571203116e-05,
+      "loss": 1.0137,
+      "step": 5938
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408656802208172e-05,
+      "loss": 0.9214,
+      "step": 5939
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8408046926468263e-05,
+      "loss": 1.1604,
+      "step": 5940
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8407436943991143e-05,
+      "loss": 1.0049,
+      "step": 5941
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406826854784548e-05,
+      "loss": 0.9803,
+      "step": 5942
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8406216658856223e-05,
+      "loss": 1.1388,
+      "step": 5943
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8405606356213918e-05,
+      "loss": 1.1159,
+      "step": 5944
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404995946865372e-05,
+      "loss": 1.1625,
+      "step": 5945
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8404385430818347e-05,
+      "loss": 1.2041,
+      "step": 5946
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8403774808080586e-05,
+      "loss": 1.1138,
+      "step": 5947
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840316407865984e-05,
+      "loss": 1.1856,
+      "step": 5948
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8402553242563863e-05,
+      "loss": 1.0851,
+      "step": 5949
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.840194229980041e-05,
+      "loss": 1.0335,
+      "step": 5950
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8401331250377238e-05,
+      "loss": 1.1005,
+      "step": 5951
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.84007200943021e-05,
+      "loss": 1.032,
+      "step": 5952
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8400108831582753e-05,
+      "loss": 0.9449,
+      "step": 5953
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839949746222697e-05,
+      "loss": 1.0985,
+      "step": 5954
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398885986242497e-05,
+      "loss": 1.0902,
+      "step": 5955
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8398274403637104e-05,
+      "loss": 0.9921,
+      "step": 5956
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839766271441855e-05,
+      "loss": 1.0664,
+      "step": 5957
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8397050918594604e-05,
+      "loss": 1.1191,
+      "step": 5958
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8396439016173033e-05,
+      "loss": 1.0754,
+      "step": 5959
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83958270071616e-05,
+      "loss": 1.1909,
+      "step": 5960
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8395214891568084e-05,
+      "loss": 1.1099,
+      "step": 5961
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8394602669400242e-05,
+      "loss": 1.0665,
+      "step": 5962
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839399034066586e-05,
+      "loss": 1.1782,
+      "step": 5963
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83933779053727e-05,
+      "loss": 1.2026,
+      "step": 5964
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392765363528545e-05,
+      "loss": 0.9175,
+      "step": 5965
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8392152715141164e-05,
+      "loss": 1.2698,
+      "step": 5966
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839153996021834e-05,
+      "loss": 1.1068,
+      "step": 5967
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.839092709876785e-05,
+      "loss": 1.167,
+      "step": 5968
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8390314130797473e-05,
+      "loss": 0.9972,
+      "step": 5969
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838970105631499e-05,
+      "loss": 1.0538,
+      "step": 5970
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8389087875328182e-05,
+      "loss": 1.085,
+      "step": 5971
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8388474587844837e-05,
+      "loss": 1.0728,
+      "step": 5972
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8387861193872744e-05,
+      "loss": 1.1378,
+      "step": 5973
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838724769341968e-05,
+      "loss": 1.1038,
+      "step": 5974
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838663408649344e-05,
+      "loss": 1.199,
+      "step": 5975
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8386020373101812e-05,
+      "loss": 1.0982,
+      "step": 5976
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8385406553252585e-05,
+      "loss": 1.1444,
+      "step": 5977
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8384792626953552e-05,
+      "loss": 0.9822,
+      "step": 5978
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838417859421251e-05,
+      "loss": 1.1459,
+      "step": 5979
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838356445503725e-05,
+      "loss": 1.0917,
+      "step": 5980
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838295020943557e-05,
+      "loss": 1.0096,
+      "step": 5981
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8382335857415268e-05,
+      "loss": 1.0838,
+      "step": 5982
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838172139898414e-05,
+      "loss": 1.1688,
+      "step": 5983
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8381106834149992e-05,
+      "loss": 1.1224,
+      "step": 5984
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.838049216292062e-05,
+      "loss": 1.0836,
+      "step": 5985
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837987738530383e-05,
+      "loss": 1.0398,
+      "step": 5986
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8379262501307425e-05,
+      "loss": 1.0122,
+      "step": 5987
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837864751093921e-05,
+      "loss": 0.9945,
+      "step": 5988
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8378032414206995e-05,
+      "loss": 1.0511,
+      "step": 5989
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8377417211118586e-05,
+      "loss": 1.1745,
+      "step": 5990
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8376801901681798e-05,
+      "loss": 1.0838,
+      "step": 5991
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837618648590443e-05,
+      "loss": 1.0961,
+      "step": 5992
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837557096379431e-05,
+      "loss": 1.0652,
+      "step": 5993
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374955335359242e-05,
+      "loss": 0.9161,
+      "step": 5994
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8374339600607042e-05,
+      "loss": 1.1825,
+      "step": 5995
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373723759545526e-05,
+      "loss": 1.0749,
+      "step": 5996
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8373107812182516e-05,
+      "loss": 1.0364,
+      "step": 5997
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8372491758525834e-05,
+      "loss": 1.0451,
+      "step": 5998
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8371875598583287e-05,
+      "loss": 1.0825,
+      "step": 5999
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.837125933236271e-05,
+      "loss": 1.129,
+      "step": 6000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370642959871923e-05,
+      "loss": 1.0325,
+      "step": 6001
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8370026481118746e-05,
+      "loss": 1.0532,
+      "step": 6002
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836940989611101e-05,
+      "loss": 1.1999,
+      "step": 6003
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368793204856545e-05,
+      "loss": 0.9067,
+      "step": 6004
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8368176407363168e-05,
+      "loss": 1.132,
+      "step": 6005
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8367559503638717e-05,
+      "loss": 1.2743,
+      "step": 6006
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836694249369103e-05,
+      "loss": 0.9978,
+      "step": 6007
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8366325377527927e-05,
+      "loss": 1.1478,
+      "step": 6008
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836570815515725e-05,
+      "loss": 1.1694,
+      "step": 6009
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836509082658683e-05,
+      "loss": 1.0922,
+      "step": 6010
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8364473391824505e-05,
+      "loss": 1.1127,
+      "step": 6011
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8363855850878117e-05,
+      "loss": 1.2067,
+      "step": 6012
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83632382037555e-05,
+      "loss": 1.0685,
+      "step": 6013
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362620450464502e-05,
+      "loss": 1.0344,
+      "step": 6014
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8362002591012957e-05,
+      "loss": 1.1525,
+      "step": 6015
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836138462540871e-05,
+      "loss": 0.9799,
+      "step": 6016
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.836076655365961e-05,
+      "loss": 1.0248,
+      "step": 6017
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8360148375773504e-05,
+      "loss": 1.0797,
+      "step": 6018
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835953009175823e-05,
+      "loss": 1.0243,
+      "step": 6019
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358911701621647e-05,
+      "loss": 1.0214,
+      "step": 6020
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8358293205371603e-05,
+      "loss": 0.9133,
+      "step": 6021
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357674603015946e-05,
+      "loss": 1.0355,
+      "step": 6022
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8357055894562532e-05,
+      "loss": 1.0797,
+      "step": 6023
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8356437080019213e-05,
+      "loss": 1.1076,
+      "step": 6024
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835581815939385e-05,
+      "loss": 1.0814,
+      "step": 6025
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835519913269429e-05,
+      "loss": 1.1036,
+      "step": 6026
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8354579999928403e-05,
+      "loss": 1.0708,
+      "step": 6027
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.835396076110404e-05,
+      "loss": 0.9883,
+      "step": 6028
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8353341416229068e-05,
+      "loss": 1.0272,
+      "step": 6029
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352721965311346e-05,
+      "loss": 1.0362,
+      "step": 6030
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8352102408358737e-05,
+      "loss": 1.1047,
+      "step": 6031
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8351482745379106e-05,
+      "loss": 0.9212,
+      "step": 6032
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8350862976380323e-05,
+      "loss": 1.2793,
+      "step": 6033
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8350243101370254e-05,
+      "loss": 1.0643,
+      "step": 6034
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349623120356766e-05,
+      "loss": 1.035,
+      "step": 6035
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8349003033347733e-05,
+      "loss": 1.0104,
+      "step": 6036
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8348382840351022e-05,
+      "loss": 0.9676,
+      "step": 6037
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347762541374508e-05,
+      "loss": 1.1339,
+      "step": 6038
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8347142136426068e-05,
+      "loss": 1.0824,
+      "step": 6039
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8346521625513576e-05,
+      "loss": 1.0673,
+      "step": 6040
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834590100864491e-05,
+      "loss": 0.9427,
+      "step": 6041
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834528028582795e-05,
+      "loss": 1.0416,
+      "step": 6042
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834465945707057e-05,
+      "loss": 1.1265,
+      "step": 6043
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8344038522380655e-05,
+      "loss": 1.0241,
+      "step": 6044
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834341748176609e-05,
+      "loss": 0.8794,
+      "step": 6045
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8342796335234755e-05,
+      "loss": 1.0563,
+      "step": 6046
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.834217508279454e-05,
+      "loss": 0.9576,
+      "step": 6047
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8341553724453326e-05,
+      "loss": 0.9597,
+      "step": 6048
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340932260219e-05,
+      "loss": 1.125,
+      "step": 6049
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8340310690099455e-05,
+      "loss": 1.0466,
+      "step": 6050
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339689014102587e-05,
+      "loss": 1.0628,
+      "step": 6051
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8339067232236276e-05,
+      "loss": 1.0698,
+      "step": 6052
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8338445344508425e-05,
+      "loss": 1.0856,
+      "step": 6053
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833782335092692e-05,
+      "loss": 1.0756,
+      "step": 6054
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8337201251499665e-05,
+      "loss": 1.1343,
+      "step": 6055
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8336579046234554e-05,
+      "loss": 1.095,
+      "step": 6056
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8335956735139487e-05,
+      "loss": 1.0341,
+      "step": 6057
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833533431822236e-05,
+      "loss": 1.076,
+      "step": 6058
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833471179549108e-05,
+      "loss": 1.1851,
+      "step": 6059
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833408916695355e-05,
+      "loss": 1.1324,
+      "step": 6060
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8333466432617667e-05,
+      "loss": 1.0617,
+      "step": 6061
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833284359249134e-05,
+      "loss": 1.0411,
+      "step": 6062
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8332220646582477e-05,
+      "loss": 1.0145,
+      "step": 6063
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8331597594898985e-05,
+      "loss": 1.0765,
+      "step": 6064
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8330974437448773e-05,
+      "loss": 1.0315,
+      "step": 6065
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.833035117423976e-05,
+      "loss": 0.997,
+      "step": 6066
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832972780527984e-05,
+      "loss": 1.2537,
+      "step": 6067
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8329104330576944e-05,
+      "loss": 1.079,
+      "step": 6068
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8328480750138976e-05,
+      "loss": 1.0652,
+      "step": 6069
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327857063973854e-05,
+      "loss": 1.0683,
+      "step": 6070
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8327233272089503e-05,
+      "loss": 0.9696,
+      "step": 6071
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8326609374493835e-05,
+      "loss": 1.0803,
+      "step": 6072
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832598537119477e-05,
+      "loss": 1.0751,
+      "step": 6073
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832536126220023e-05,
+      "loss": 1.1879,
+      "step": 6074
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832473704751814e-05,
+      "loss": 1.0572,
+      "step": 6075
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8324112727156426e-05,
+      "loss": 1.1595,
+      "step": 6076
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8323488301123007e-05,
+      "loss": 1.1418,
+      "step": 6077
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322863769425812e-05,
+      "loss": 0.9675,
+      "step": 6078
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8322239132072774e-05,
+      "loss": 1.0534,
+      "step": 6079
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.832161438907182e-05,
+      "loss": 1.0023,
+      "step": 6080
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320989540430876e-05,
+      "loss": 1.1055,
+      "step": 6081
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8320364586157882e-05,
+      "loss": 1.0515,
+      "step": 6082
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8319739526260764e-05,
+      "loss": 1.0508,
+      "step": 6083
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8319114360747467e-05,
+      "loss": 1.0434,
+      "step": 6084
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8318489089625916e-05,
+      "loss": 1.1343,
+      "step": 6085
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317863712904055e-05,
+      "loss": 0.9607,
+      "step": 6086
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8317238230589818e-05,
+      "loss": 1.0713,
+      "step": 6087
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831661264269115e-05,
+      "loss": 1.1384,
+      "step": 6088
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8315986949215997e-05,
+      "loss": 1.0527,
+      "step": 6089
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831536115017229e-05,
+      "loss": 1.0467,
+      "step": 6090
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314735245567986e-05,
+      "loss": 1.0154,
+      "step": 6091
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8314109235411018e-05,
+      "loss": 1.0029,
+      "step": 6092
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8313483119709344e-05,
+      "loss": 1.0007,
+      "step": 6093
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312856898470905e-05,
+      "loss": 0.9529,
+      "step": 6094
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8312230571703654e-05,
+      "loss": 1.0057,
+      "step": 6095
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831160413941554e-05,
+      "loss": 1.1612,
+      "step": 6096
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831097760161452e-05,
+      "loss": 1.0775,
+      "step": 6097
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.831035095830854e-05,
+      "loss": 1.279,
+      "step": 6098
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8309724209505564e-05,
+      "loss": 1.2404,
+      "step": 6099
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830909735521354e-05,
+      "loss": 1.0514,
+      "step": 6100
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8308470395440432e-05,
+      "loss": 1.0279,
+      "step": 6101
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8307843330194197e-05,
+      "loss": 1.0304,
+      "step": 6102
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8307216159482792e-05,
+      "loss": 0.9657,
+      "step": 6103
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8306588883314183e-05,
+      "loss": 0.9552,
+      "step": 6104
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830596150169633e-05,
+      "loss": 1.1444,
+      "step": 6105
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8305334014637202e-05,
+      "loss": 1.2095,
+      "step": 6106
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830470642214476e-05,
+      "loss": 0.9924,
+      "step": 6107
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8304078724226973e-05,
+      "loss": 1.1127,
+      "step": 6108
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8303450920891806e-05,
+      "loss": 1.0792,
+      "step": 6109
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.830282301214724e-05,
+      "loss": 0.982,
+      "step": 6110
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8302194998001228e-05,
+      "loss": 1.0451,
+      "step": 6111
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8301566878461757e-05,
+      "loss": 1.0931,
+      "step": 6112
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.83009386535368e-05,
+      "loss": 1.0232,
+      "step": 6113
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8300310323234324e-05,
+      "loss": 1.1753,
+      "step": 6114
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829968188756231e-05,
+      "loss": 1.1079,
+      "step": 6115
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8299053346528738e-05,
+      "loss": 1.1027,
+      "step": 6116
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829842470014158e-05,
+      "loss": 1.1201,
+      "step": 6117
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297795948408824e-05,
+      "loss": 1.241,
+      "step": 6118
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8297167091338447e-05,
+      "loss": 1.1076,
+      "step": 6119
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8296538128938436e-05,
+      "loss": 0.9772,
+      "step": 6120
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295909061216775e-05,
+      "loss": 1.0392,
+      "step": 6121
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8295279888181445e-05,
+      "loss": 1.049,
+      "step": 6122
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294650609840435e-05,
+      "loss": 1.0561,
+      "step": 6123
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8294021226201736e-05,
+      "loss": 1.0476,
+      "step": 6124
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829339173727334e-05,
+      "loss": 0.9884,
+      "step": 6125
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292762143063233e-05,
+      "loss": 1.1254,
+      "step": 6126
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8292132443579406e-05,
+      "loss": 1.1734,
+      "step": 6127
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.829150263882986e-05,
+      "loss": 1.1245,
+      "step": 6128
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290872728822586e-05,
+      "loss": 1.045,
+      "step": 6129
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8290242713565578e-05,
+      "loss": 1.1167,
+      "step": 6130
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828961259306684e-05,
+      "loss": 1.1408,
+      "step": 6131
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288982367334362e-05,
+      "loss": 0.96,
+      "step": 6132
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8288352036376156e-05,
+      "loss": 1.1671,
+      "step": 6133
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287721600200213e-05,
+      "loss": 1.0429,
+      "step": 6134
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8287091058814544e-05,
+      "loss": 1.0938,
+      "step": 6135
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8286460412227153e-05,
+      "loss": 1.0203,
+      "step": 6136
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828582966044604e-05,
+      "loss": 0.9942,
+      "step": 6137
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8285198803479218e-05,
+      "loss": 1.1011,
+      "step": 6138
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828456784133469e-05,
+      "loss": 1.1484,
+      "step": 6139
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828393677402047e-05,
+      "loss": 1.0315,
+      "step": 6140
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828330560154457e-05,
+      "loss": 1.1146,
+      "step": 6141
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282674323915e-05,
+      "loss": 1.1575,
+      "step": 6142
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8282042941139775e-05,
+      "loss": 1.0798,
+      "step": 6143
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828141145322691e-05,
+      "loss": 1.1036,
+      "step": 6144
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.828077986018442e-05,
+      "loss": 1.1389,
+      "step": 6145
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8280148162020323e-05,
+      "loss": 1.0446,
+      "step": 6146
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827951635874264e-05,
+      "loss": 1.0971,
+      "step": 6147
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827888445035939e-05,
+      "loss": 0.9416,
+      "step": 6148
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82782524368786e-05,
+      "loss": 1.097,
+      "step": 6149
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8277620318308287e-05,
+      "loss": 1.1674,
+      "step": 6150
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276988094656472e-05,
+      "loss": 0.9746,
+      "step": 6151
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8276355765931196e-05,
+      "loss": 0.9496,
+      "step": 6152
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8275723332140468e-05,
+      "loss": 1.1679,
+      "step": 6153
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827509079329233e-05,
+      "loss": 1.015,
+      "step": 6154
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8274458149394804e-05,
+      "loss": 1.0683,
+      "step": 6155
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273825400455923e-05,
+      "loss": 1.0772,
+      "step": 6156
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8273192546483724e-05,
+      "loss": 1.1162,
+      "step": 6157
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8272559587486236e-05,
+      "loss": 1.0129,
+      "step": 6158
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8271926523471496e-05,
+      "loss": 1.2106,
+      "step": 6159
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.827129335444754e-05,
+      "loss": 0.9378,
+      "step": 6160
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8270660080422405e-05,
+      "loss": 0.9977,
+      "step": 6161
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8270026701404132e-05,
+      "loss": 1.0874,
+      "step": 6162
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826939321740076e-05,
+      "loss": 0.9138,
+      "step": 6163
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826875962842033e-05,
+      "loss": 1.0513,
+      "step": 6164
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8268125934470892e-05,
+      "loss": 0.9654,
+      "step": 6165
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826749213556048e-05,
+      "loss": 1.0867,
+      "step": 6166
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8266858231697146e-05,
+      "loss": 1.1433,
+      "step": 6167
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826622422288894e-05,
+      "loss": 1.0189,
+      "step": 6168
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8265590109143905e-05,
+      "loss": 1.1158,
+      "step": 6169
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826495589047009e-05,
+      "loss": 1.0326,
+      "step": 6170
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826432156687555e-05,
+      "loss": 1.0185,
+      "step": 6171
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263687138368338e-05,
+      "loss": 0.9762,
+      "step": 6172
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8263052604956504e-05,
+      "loss": 1.1964,
+      "step": 6173
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826241796664811e-05,
+      "loss": 1.0784,
+      "step": 6174
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8261783223451203e-05,
+      "loss": 1.1231,
+      "step": 6175
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.826114837537385e-05,
+      "loss": 0.9715,
+      "step": 6176
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.82605134224241e-05,
+      "loss": 0.99,
+      "step": 6177
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8259878364610026e-05,
+      "loss": 1.1135,
+      "step": 6178
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.825924320193968e-05,
+      "loss": 1.0681,
+      "step": 6179
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 1.8258607934421127e-05,
+      "loss": 1.0493,
+      "step": 6180
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257972562062435e-05,
+      "loss": 1.0754,
+      "step": 6181
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8257337084871668e-05,
+      "loss": 1.1711,
+      "step": 6182
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825670150285689e-05,
+      "loss": 1.0892,
+      "step": 6183
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825606581602618e-05,
+      "loss": 1.1039,
+      "step": 6184
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8255430024387593e-05,
+      "loss": 1.0967,
+      "step": 6185
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8254794127949213e-05,
+      "loss": 0.9573,
+      "step": 6186
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82541581267191e-05,
+      "loss": 1.0087,
+      "step": 6187
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825352202070534e-05,
+      "loss": 1.1424,
+      "step": 6188
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8252885809916e-05,
+      "loss": 1.0333,
+      "step": 6189
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825224949435916e-05,
+      "loss": 1.0981,
+      "step": 6190
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8251613074042896e-05,
+      "loss": 1.1994,
+      "step": 6191
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8250976548975294e-05,
+      "loss": 1.0695,
+      "step": 6192
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.825033991916442e-05,
+      "loss": 0.9731,
+      "step": 6193
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824970318461837e-05,
+      "loss": 1.1485,
+      "step": 6194
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824906634534522e-05,
+      "loss": 1.224,
+      "step": 6195
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824842940135306e-05,
+      "loss": 1.0542,
+      "step": 6196
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247792352649967e-05,
+      "loss": 1.1501,
+      "step": 6197
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8247155199244034e-05,
+      "loss": 1.0562,
+      "step": 6198
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824651794114335e-05,
+      "loss": 1.1401,
+      "step": 6199
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245880578356003e-05,
+      "loss": 1.3679,
+      "step": 6200
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8245243110890084e-05,
+      "loss": 1.0805,
+      "step": 6201
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8244605538753685e-05,
+      "loss": 1.1418,
+      "step": 6202
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82439678619549e-05,
+      "loss": 1.3987,
+      "step": 6203
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8243330080501826e-05,
+      "loss": 1.0328,
+      "step": 6204
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242692194402554e-05,
+      "loss": 1.0847,
+      "step": 6205
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8242054203665193e-05,
+      "loss": 1.1537,
+      "step": 6206
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.824141610829783e-05,
+      "loss": 1.0194,
+      "step": 6207
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8240777908308568e-05,
+      "loss": 1.0478,
+      "step": 6208
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8240139603705516e-05,
+      "loss": 1.003,
+      "step": 6209
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823950119449677e-05,
+      "loss": 0.9298,
+      "step": 6210
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8238862680690434e-05,
+      "loss": 1.1198,
+      "step": 6211
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823822406229462e-05,
+      "loss": 1.0177,
+      "step": 6212
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8237585339317426e-05,
+      "loss": 1.0133,
+      "step": 6213
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236946511766974e-05,
+      "loss": 1.0013,
+      "step": 6214
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8236307579651357e-05,
+      "loss": 1.1228,
+      "step": 6215
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82356685429787e-05,
+      "loss": 1.1039,
+      "step": 6216
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8235029401757102e-05,
+      "loss": 1.1294,
+      "step": 6217
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823439015599469e-05,
+      "loss": 1.1799,
+      "step": 6218
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233750805699574e-05,
+      "loss": 1.0813,
+      "step": 6219
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8233111350879865e-05,
+      "loss": 0.9551,
+      "step": 6220
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823247179154369e-05,
+      "loss": 0.9733,
+      "step": 6221
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.823183212769916e-05,
+      "loss": 1.1207,
+      "step": 6222
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.82311923593544e-05,
+      "loss": 1.1569,
+      "step": 6223
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8230552486517527e-05,
+      "loss": 1.1392,
+      "step": 6224
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822991250919667e-05,
+      "loss": 1.0541,
+      "step": 6225
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822927242739995e-05,
+      "loss": 1.0485,
+      "step": 6226
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8228632241135492e-05,
+      "loss": 1.0384,
+      "step": 6227
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227991950411426e-05,
+      "loss": 1.078,
+      "step": 6228
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8227351555235877e-05,
+      "loss": 1.0045,
+      "step": 6229
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226711055616976e-05,
+      "loss": 1.2397,
+      "step": 6230
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8226070451562852e-05,
+      "loss": 1.2313,
+      "step": 6231
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8225429743081638e-05,
+      "loss": 1.2256,
+      "step": 6232
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224788930181475e-05,
+      "loss": 1.1718,
+      "step": 6233
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8224148012870485e-05,
+      "loss": 1.191,
+      "step": 6234
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8223506991156812e-05,
+      "loss": 1.0283,
+      "step": 6235
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8222865865048595e-05,
+      "loss": 1.1493,
+      "step": 6236
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.822222463455397e-05,
+      "loss": 1.0565,
+      "step": 6237
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8221583299681072e-05,
+      "loss": 1.2229,
+      "step": 6238
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220941860438053e-05,
+      "loss": 1.0607,
+      "step": 6239
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8220300316833046e-05,
+      "loss": 1.0867,
+      "step": 6240
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219658668874203e-05,
+      "loss": 0.942,
+      "step": 6241
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8219016916569666e-05,
+      "loss": 1.0051,
+      "step": 6242
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821837505992758e-05,
+      "loss": 1.1338,
+      "step": 6243
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217733098956098e-05,
+      "loss": 1.0176,
+      "step": 6244
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8217091033663367e-05,
+      "loss": 1.0093,
+      "step": 6245
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8216448864057536e-05,
+      "loss": 0.9661,
+      "step": 6246
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821580659014676e-05,
+      "loss": 1.0519,
+      "step": 6247
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8215164211939186e-05,
+      "loss": 1.1817,
+      "step": 6248
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821452172944298e-05,
+      "loss": 1.0977,
+      "step": 6249
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821387914266629e-05,
+      "loss": 1.1931,
+      "step": 6250
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821323645161727e-05,
+      "loss": 1.2326,
+      "step": 6251
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8212593656304092e-05,
+      "loss": 1.0538,
+      "step": 6252
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211950756734906e-05,
+      "loss": 1.0767,
+      "step": 6253
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8211307752917872e-05,
+      "loss": 1.0378,
+      "step": 6254
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.821066464486116e-05,
+      "loss": 1.1555,
+      "step": 6255
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8210021432572927e-05,
+      "loss": 1.1423,
+      "step": 6256
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8209378116061348e-05,
+      "loss": 1.1146,
+      "step": 6257
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8208734695334578e-05,
+      "loss": 1.0715,
+      "step": 6258
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820809117040079e-05,
+      "loss": 1.0247,
+      "step": 6259
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8207447541268154e-05,
+      "loss": 1.0607,
+      "step": 6260
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820680380794484e-05,
+      "loss": 1.0793,
+      "step": 6261
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8206159970439024e-05,
+      "loss": 1.0582,
+      "step": 6262
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820551602875887e-05,
+      "loss": 1.1321,
+      "step": 6263
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8204871982912563e-05,
+      "loss": 1.1692,
+      "step": 6264
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.820422783290827e-05,
+      "loss": 1.0644,
+      "step": 6265
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8203583578754177e-05,
+      "loss": 1.1199,
+      "step": 6266
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202939220458456e-05,
+      "loss": 1.1144,
+      "step": 6267
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8202294758029287e-05,
+      "loss": 1.1606,
+      "step": 6268
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201650191474853e-05,
+      "loss": 1.0892,
+      "step": 6269
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8201005520803342e-05,
+      "loss": 1.2437,
+      "step": 6270
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8200360746022923e-05,
+      "loss": 1.0739,
+      "step": 6271
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81997158671418e-05,
+      "loss": 1.013,
+      "step": 6272
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8199070884168147e-05,
+      "loss": 1.0194,
+      "step": 6273
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8198425797110154e-05,
+      "loss": 1.0279,
+      "step": 6274
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197780605976006e-05,
+      "loss": 1.0982,
+      "step": 6275
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8197135310773906e-05,
+      "loss": 1.0356,
+      "step": 6276
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8196489911512037e-05,
+      "loss": 0.9465,
+      "step": 6277
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8195844408198594e-05,
+      "loss": 1.1029,
+      "step": 6278
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819519880084177e-05,
+      "loss": 1.0211,
+      "step": 6279
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819455308944976e-05,
+      "loss": 1.0446,
+      "step": 6280
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8193907274030765e-05,
+      "loss": 1.1251,
+      "step": 6281
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.819326135459298e-05,
+      "loss": 1.1827,
+      "step": 6282
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8192615331144607e-05,
+      "loss": 1.1048,
+      "step": 6283
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8191969203693846e-05,
+      "loss": 1.1338,
+      "step": 6284
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81913229722489e-05,
+      "loss": 1.1408,
+      "step": 6285
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190676636817973e-05,
+      "loss": 1.0293,
+      "step": 6286
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8190030197409264e-05,
+      "loss": 1.1536,
+      "step": 6287
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8189383654030987e-05,
+      "loss": 1.1067,
+      "step": 6288
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818873700669135e-05,
+      "loss": 1.0829,
+      "step": 6289
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8188090255398556e-05,
+      "loss": 0.9132,
+      "step": 6290
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8187443400160816e-05,
+      "loss": 0.977,
+      "step": 6291
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8186796440986348e-05,
+      "loss": 1.0546,
+      "step": 6292
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818614937788336e-05,
+      "loss": 1.1769,
+      "step": 6293
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8185502210860063e-05,
+      "loss": 0.9638,
+      "step": 6294
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184854939924682e-05,
+      "loss": 1.0936,
+      "step": 6295
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8184207565085424e-05,
+      "loss": 1.0029,
+      "step": 6296
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8183560086350517e-05,
+      "loss": 1.1222,
+      "step": 6297
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818291250372817e-05,
+      "loss": 0.993,
+      "step": 6298
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8182264817226612e-05,
+      "loss": 0.969,
+      "step": 6299
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818161702685406e-05,
+      "loss": 1.097,
+      "step": 6300
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.818096913261874e-05,
+      "loss": 0.9588,
+      "step": 6301
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8180321134528876e-05,
+      "loss": 1.077,
+      "step": 6302
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179673032592697e-05,
+      "loss": 1.1411,
+      "step": 6303
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8179024826818427e-05,
+      "loss": 1.0293,
+      "step": 6304
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8178376517214296e-05,
+      "loss": 1.1422,
+      "step": 6305
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8177728103788533e-05,
+      "loss": 1.0367,
+      "step": 6306
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817707958654937e-05,
+      "loss": 1.2258,
+      "step": 6307
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817643096550504e-05,
+      "loss": 1.0561,
+      "step": 6308
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817578224066378e-05,
+      "loss": 1.0607,
+      "step": 6309
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8175133412033816e-05,
+      "loss": 0.9871,
+      "step": 6310
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8174484479623397e-05,
+      "loss": 1.1138,
+      "step": 6311
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817383544344075e-05,
+      "loss": 1.055,
+      "step": 6312
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817318630349412e-05,
+      "loss": 1.0576,
+      "step": 6313
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8172537059791748e-05,
+      "loss": 1.0733,
+      "step": 6314
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8171887712341873e-05,
+      "loss": 0.922,
+      "step": 6315
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.817123826115274e-05,
+      "loss": 1.0821,
+      "step": 6316
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8170588706232597e-05,
+      "loss": 1.1219,
+      "step": 6317
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8169939047589682e-05,
+      "loss": 0.9779,
+      "step": 6318
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816928928523225e-05,
+      "loss": 1.2338,
+      "step": 6319
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8168639419168543e-05,
+      "loss": 1.0514,
+      "step": 6320
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167989449406814e-05,
+      "loss": 1.067,
+      "step": 6321
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8167339375955318e-05,
+      "loss": 1.1588,
+      "step": 6322
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81666891988223e-05,
+      "loss": 1.0798,
+      "step": 6323
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816603891801602e-05,
+      "loss": 1.0198,
+      "step": 6324
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8165388533544723e-05,
+      "loss": 1.1146,
+      "step": 6325
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816473804541668e-05,
+      "loss": 1.1963,
+      "step": 6326
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816408745364014e-05,
+      "loss": 1.0959,
+      "step": 6327
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8163436758223357e-05,
+      "loss": 1.0632,
+      "step": 6328
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8162785959174604e-05,
+      "loss": 0.9922,
+      "step": 6329
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816213505650214e-05,
+      "loss": 1.0859,
+      "step": 6330
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816148405021422e-05,
+      "loss": 0.9457,
+      "step": 6331
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816083294031911e-05,
+      "loss": 0.9187,
+      "step": 6332
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.816018172682508e-05,
+      "loss": 1.0832,
+      "step": 6333
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8159530409740397e-05,
+      "loss": 1.1191,
+      "step": 6334
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158878989073328e-05,
+      "loss": 1.0826,
+      "step": 6335
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8158227464832138e-05,
+      "loss": 0.911,
+      "step": 6336
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8157575837025106e-05,
+      "loss": 0.9435,
+      "step": 6337
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8156924105660497e-05,
+      "loss": 1.0438,
+      "step": 6338
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815627227074659e-05,
+      "loss": 1.1227,
+      "step": 6339
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8155620332291656e-05,
+      "loss": 0.9457,
+      "step": 6340
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154968290303975e-05,
+      "loss": 1.1826,
+      "step": 6341
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8154316144791823e-05,
+      "loss": 0.938,
+      "step": 6342
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153663895763475e-05,
+      "loss": 1.1845,
+      "step": 6343
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8153011543227214e-05,
+      "loss": 1.1591,
+      "step": 6344
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815235908719132e-05,
+      "loss": 1.1959,
+      "step": 6345
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8151706527664078e-05,
+      "loss": 1.0643,
+      "step": 6346
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.815105386465377e-05,
+      "loss": 1.0416,
+      "step": 6347
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8150401098168684e-05,
+      "loss": 1.1332,
+      "step": 6348
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8149748228217105e-05,
+      "loss": 1.174,
+      "step": 6349
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814909525480732e-05,
+      "loss": 0.9796,
+      "step": 6350
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814844217794762e-05,
+      "loss": 0.9968,
+      "step": 6351
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147788997646295e-05,
+      "loss": 1.1695,
+      "step": 6352
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8147135713911637e-05,
+      "loss": 1.1252,
+      "step": 6353
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8146482326751935e-05,
+      "loss": 1.0608,
+      "step": 6354
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814582883617549e-05,
+      "loss": 1.0374,
+      "step": 6355
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8145175242190598e-05,
+      "loss": 1.0345,
+      "step": 6356
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814452154480555e-05,
+      "loss": 1.087,
+      "step": 6357
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8143867744028646e-05,
+      "loss": 1.0163,
+      "step": 6358
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814321383986819e-05,
+      "loss": 1.1201,
+      "step": 6359
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814255983233248e-05,
+      "loss": 0.9801,
+      "step": 6360
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.814190572142982e-05,
+      "loss": 1.0048,
+      "step": 6361
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8141251507168514e-05,
+      "loss": 1.0472,
+      "step": 6362
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8140597189556862e-05,
+      "loss": 1.0245,
+      "step": 6363
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813994276860318e-05,
+      "loss": 1.0948,
+      "step": 6364
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8139288244315762e-05,
+      "loss": 1.0193,
+      "step": 6365
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813863361670293e-05,
+      "loss": 0.9942,
+      "step": 6366
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8137978885772986e-05,
+      "loss": 0.8862,
+      "step": 6367
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813732405153425e-05,
+      "loss": 1.1285,
+      "step": 6368
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136669113995026e-05,
+      "loss": 1.178,
+      "step": 6369
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8136014073163632e-05,
+      "loss": 1.1297,
+      "step": 6370
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813535892904838e-05,
+      "loss": 1.0631,
+      "step": 6371
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134703681657593e-05,
+      "loss": 1.0537,
+      "step": 6372
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8134048330999585e-05,
+      "loss": 1.072,
+      "step": 6373
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8133392877082677e-05,
+      "loss": 1.1949,
+      "step": 6374
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.813273731991519e-05,
+      "loss": 1.136,
+      "step": 6375
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8132081659505447e-05,
+      "loss": 1.196,
+      "step": 6376
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8131425895861766e-05,
+      "loss": 1.1885,
+      "step": 6377
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130770028992478e-05,
+      "loss": 1.0356,
+      "step": 6378
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8130114058905906e-05,
+      "loss": 1.0633,
+      "step": 6379
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8129457985610376e-05,
+      "loss": 1.0628,
+      "step": 6380
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812880180911422e-05,
+      "loss": 1.1806,
+      "step": 6381
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8128145529425767e-05,
+      "loss": 1.1017,
+      "step": 6382
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8127489146553348e-05,
+      "loss": 1.1328,
+      "step": 6383
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8126832660505293e-05,
+      "loss": 1.1555,
+      "step": 6384
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812617607128994e-05,
+      "loss": 1.1741,
+      "step": 6385
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8125519378915618e-05,
+      "loss": 0.9989,
+      "step": 6386
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124862583390676e-05,
+      "loss": 1.0867,
+      "step": 6387
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8124205684723436e-05,
+      "loss": 1.0572,
+      "step": 6388
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8123548682922246e-05,
+      "loss": 1.0856,
+      "step": 6389
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122891577995447e-05,
+      "loss": 1.0621,
+      "step": 6390
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8122234369951378e-05,
+      "loss": 1.0855,
+      "step": 6391
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.812157705879838e-05,
+      "loss": 1.0874,
+      "step": 6392
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.81209196445448e-05,
+      "loss": 1.0964,
+      "step": 6393
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8120262127198988e-05,
+      "loss": 1.0667,
+      "step": 6394
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8119604506769285e-05,
+      "loss": 1.1595,
+      "step": 6395
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811894678326404e-05,
+      "loss": 1.1055,
+      "step": 6396
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8118288956691604e-05,
+      "loss": 1.0866,
+      "step": 6397
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8117631027060327e-05,
+      "loss": 0.9644,
+      "step": 6398
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811697299437856e-05,
+      "loss": 1.0197,
+      "step": 6399
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8116314858654662e-05,
+      "loss": 1.1401,
+      "step": 6400
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8115656619896983e-05,
+      "loss": 1.1035,
+      "step": 6401
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114998278113876e-05,
+      "loss": 1.1682,
+      "step": 6402
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8114339833313704e-05,
+      "loss": 1.1017,
+      "step": 6403
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811368128550483e-05,
+      "loss": 1.1234,
+      "step": 6404
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8113022634695602e-05,
+      "loss": 1.0926,
+      "step": 6405
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8112363880894385e-05,
+      "loss": 1.0611,
+      "step": 6406
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8111705024109548e-05,
+      "loss": 1.0405,
+      "step": 6407
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811104606434945e-05,
+      "loss": 1.1162,
+      "step": 6408
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.811038700162246e-05,
+      "loss": 1.0733,
+      "step": 6409
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810972783593694e-05,
+      "loss": 1.1865,
+      "step": 6410
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8109068567301257e-05,
+      "loss": 0.9923,
+      "step": 6411
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8108409195723787e-05,
+      "loss": 1.1296,
+      "step": 6412
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107749721212895e-05,
+      "loss": 1.1394,
+      "step": 6413
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8107090143776952e-05,
+      "loss": 1.1523,
+      "step": 6414
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8106430463424338e-05,
+      "loss": 1.0152,
+      "step": 6415
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105770680163418e-05,
+      "loss": 1.0671,
+      "step": 6416
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8105110794002577e-05,
+      "loss": 1.1691,
+      "step": 6417
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8104450804950185e-05,
+      "loss": 1.0893,
+      "step": 6418
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8103790713014626e-05,
+      "loss": 1.0384,
+      "step": 6419
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810313051820427e-05,
+      "loss": 1.0372,
+      "step": 6420
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8102470220527507e-05,
+      "loss": 1.0444,
+      "step": 6421
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.810180981999272e-05,
+      "loss": 1.0177,
+      "step": 6422
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8101149316608287e-05,
+      "loss": 1.0848,
+      "step": 6423
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8100488710382597e-05,
+      "loss": 1.1124,
+      "step": 6424
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809982800132403e-05,
+      "loss": 1.0123,
+      "step": 6425
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809916718944098e-05,
+      "loss": 1.1374,
+      "step": 6426
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8098506274741834e-05,
+      "loss": 1.0973,
+      "step": 6427
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809784525723498e-05,
+      "loss": 1.0751,
+      "step": 6428
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809718413692881e-05,
+      "loss": 1.1241,
+      "step": 6429
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8096522913831723e-05,
+      "loss": 0.965,
+      "step": 6430
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095861587952105e-05,
+      "loss": 1.1074,
+      "step": 6431
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8095200159298348e-05,
+      "loss": 1.0598,
+      "step": 6432
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809453862787886e-05,
+      "loss": 0.9649,
+      "step": 6433
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8093876993702036e-05,
+      "loss": 1.0287,
+      "step": 6434
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809321525677627e-05,
+      "loss": 1.1511,
+      "step": 6435
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8092553417109965e-05,
+      "loss": 1.2384,
+      "step": 6436
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8091891474711522e-05,
+      "loss": 1.1159,
+      "step": 6437
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8091229429589344e-05,
+      "loss": 0.9567,
+      "step": 6438
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.809056728175184e-05,
+      "loss": 1.1087,
+      "step": 6439
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089905031207412e-05,
+      "loss": 1.1387,
+      "step": 6440
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8089242677964467e-05,
+      "loss": 1.1044,
+      "step": 6441
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8088580222031412e-05,
+      "loss": 1.0263,
+      "step": 6442
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8087917663416662e-05,
+      "loss": 1.0424,
+      "step": 6443
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808725500212862e-05,
+      "loss": 1.0178,
+      "step": 6444
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8086592238175706e-05,
+      "loss": 1.1435,
+      "step": 6445
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085929371566325e-05,
+      "loss": 1.1394,
+      "step": 6446
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8085266402308898e-05,
+      "loss": 1.0982,
+      "step": 6447
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8084603330411847e-05,
+      "loss": 1.0529,
+      "step": 6448
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083940155883577e-05,
+      "loss": 1.0331,
+      "step": 6449
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8083276878732513e-05,
+      "loss": 1.1064,
+      "step": 6450
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8082613498967073e-05,
+      "loss": 0.9134,
+      "step": 6451
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8081950016595683e-05,
+      "loss": 1.0641,
+      "step": 6452
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808128643162676e-05,
+      "loss": 1.0268,
+      "step": 6453
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.808062274406873e-05,
+      "loss": 0.9796,
+      "step": 6454
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807995895393002e-05,
+      "loss": 1.1864,
+      "step": 6455
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8079295061219057e-05,
+      "loss": 1.0194,
+      "step": 6456
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8078631065944264e-05,
+      "loss": 0.9959,
+      "step": 6457
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8077966968114077e-05,
+      "loss": 1.0931,
+      "step": 6458
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807730276773692e-05,
+      "loss": 1.0026,
+      "step": 6459
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807663846482123e-05,
+      "loss": 0.9689,
+      "step": 6460
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075974059375437e-05,
+      "loss": 1.0119,
+      "step": 6461
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8075309551407977e-05,
+      "loss": 1.0814,
+      "step": 6462
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8074644940927283e-05,
+      "loss": 1.0204,
+      "step": 6463
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8073980227941793e-05,
+      "loss": 1.1197,
+      "step": 6464
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.807331541245995e-05,
+      "loss": 1.1097,
+      "step": 6465
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8072650494490185e-05,
+      "loss": 1.146,
+      "step": 6466
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 1.8071985474040943e-05,
+      "loss": 1.0731,
+      "step": 6467
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.807132035112067e-05,
+      "loss": 1.1035,
+      "step": 6468
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.807065512573781e-05,
+      "loss": 1.1553,
+      "step": 6469
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80699897979008e-05,
+      "loss": 1.071,
+      "step": 6470
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806932436761809e-05,
+      "loss": 1.0129,
+      "step": 6471
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8068658834898127e-05,
+      "loss": 0.9848,
+      "step": 6472
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8067993199749364e-05,
+      "loss": 1.067,
+      "step": 6473
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806732746218024e-05,
+      "loss": 1.0903,
+      "step": 6474
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806666162219922e-05,
+      "loss": 1.1217,
+      "step": 6475
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.806599567981475e-05,
+      "loss": 1.1307,
+      "step": 6476
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8065329635035285e-05,
+      "loss": 1.0429,
+      "step": 6477
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8064663487869275e-05,
+      "loss": 1.1047,
+      "step": 6478
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063997238325185e-05,
+      "loss": 1.0015,
+      "step": 6479
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8063330886411464e-05,
+      "loss": 1.2029,
+      "step": 6480
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8062664432136583e-05,
+      "loss": 1.0203,
+      "step": 6481
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061997875508988e-05,
+      "loss": 1.1095,
+      "step": 6482
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8061331216537153e-05,
+      "loss": 1.1317,
+      "step": 6483
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8060664455229536e-05,
+      "loss": 0.9444,
+      "step": 6484
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80599975915946e-05,
+      "loss": 0.9467,
+      "step": 6485
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805933062564081e-05,
+      "loss": 1.2121,
+      "step": 6486
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8058663557376633e-05,
+      "loss": 1.0558,
+      "step": 6487
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057996386810543e-05,
+      "loss": 1.1952,
+      "step": 6488
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8057329113951007e-05,
+      "loss": 1.0216,
+      "step": 6489
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805666173880649e-05,
+      "loss": 1.0969,
+      "step": 6490
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805599426138547e-05,
+      "loss": 1.1095,
+      "step": 6491
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805532668169642e-05,
+      "loss": 1.0941,
+      "step": 6492
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805465899974781e-05,
+      "loss": 1.1421,
+      "step": 6493
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8053991215548122e-05,
+      "loss": 1.0979,
+      "step": 6494
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805332332910583e-05,
+      "loss": 1.129,
+      "step": 6495
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8052655340429414e-05,
+      "loss": 1.1891,
+      "step": 6496
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805198724952735e-05,
+      "loss": 1.066,
+      "step": 6497
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805131905640813e-05,
+      "loss": 1.1106,
+      "step": 6498
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.805065076108022e-05,
+      "loss": 1.0153,
+      "step": 6499
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8049982363552117e-05,
+      "loss": 0.983,
+      "step": 6500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80493138638323e-05,
+      "loss": 1.0319,
+      "step": 6501
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8048645261929257e-05,
+      "loss": 1.1334,
+      "step": 6502
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804797655785148e-05,
+      "loss": 1.0466,
+      "step": 6503
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8047307751607447e-05,
+      "loss": 1.054,
+      "step": 6504
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804663884320566e-05,
+      "loss": 1.0172,
+      "step": 6505
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045969832654605e-05,
+      "loss": 1.0724,
+      "step": 6506
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8045300719962774e-05,
+      "loss": 1.0451,
+      "step": 6507
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804463150513866e-05,
+      "loss": 1.1057,
+      "step": 6508
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8043962188190763e-05,
+      "loss": 1.0234,
+      "step": 6509
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.804329276912758e-05,
+      "loss": 1.0901,
+      "step": 6510
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80426232479576e-05,
+      "loss": 1.0104,
+      "step": 6511
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8041953624689335e-05,
+      "loss": 1.1427,
+      "step": 6512
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8041283899331282e-05,
+      "loss": 1.1202,
+      "step": 6513
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8040614071891935e-05,
+      "loss": 1.0837,
+      "step": 6514
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039944142379805e-05,
+      "loss": 1.067,
+      "step": 6515
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8039274110803396e-05,
+      "loss": 0.9314,
+      "step": 6516
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803860397717121e-05,
+      "loss": 0.995,
+      "step": 6517
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037933741491757e-05,
+      "loss": 1.2288,
+      "step": 6518
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8037263403773545e-05,
+      "loss": 1.1061,
+      "step": 6519
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803659296402508e-05,
+      "loss": 1.1076,
+      "step": 6520
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035922422254884e-05,
+      "loss": 1.0716,
+      "step": 6521
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8035251778471457e-05,
+      "loss": 1.0351,
+      "step": 6522
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8034581032683315e-05,
+      "loss": 1.0473,
+      "step": 6523
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803391018489898e-05,
+      "loss": 1.0549,
+      "step": 6524
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.803323923512696e-05,
+      "loss": 1.0829,
+      "step": 6525
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8032568183375775e-05,
+      "loss": 1.0374,
+      "step": 6526
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031897029653947e-05,
+      "loss": 1.1522,
+      "step": 6527
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8031225773969992e-05,
+      "loss": 0.9958,
+      "step": 6528
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8030554416332435e-05,
+      "loss": 1.1876,
+      "step": 6529
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029882956749795e-05,
+      "loss": 1.0674,
+      "step": 6530
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8029211395230598e-05,
+      "loss": 0.9874,
+      "step": 6531
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802853973178337e-05,
+      "loss": 1.1462,
+      "step": 6532
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8027867966416638e-05,
+      "loss": 1.1063,
+      "step": 6533
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8027196099138923e-05,
+      "loss": 1.1444,
+      "step": 6534
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802652412995876e-05,
+      "loss": 1.127,
+      "step": 6535
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025852058884682e-05,
+      "loss": 1.13,
+      "step": 6536
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8025179885925218e-05,
+      "loss": 1.0449,
+      "step": 6537
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8024507611088897e-05,
+      "loss": 0.9233,
+      "step": 6538
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802383523438426e-05,
+      "loss": 1.1407,
+      "step": 6539
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8023162755819835e-05,
+      "loss": 0.9955,
+      "step": 6540
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8022490175404167e-05,
+      "loss": 0.9623,
+      "step": 6541
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802181749314579e-05,
+      "loss": 1.0338,
+      "step": 6542
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.802114470905324e-05,
+      "loss": 1.0756,
+      "step": 6543
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8020471823135067e-05,
+      "loss": 1.121,
+      "step": 6544
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8019798835399808e-05,
+      "loss": 1.0609,
+      "step": 6545
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8019125745856e-05,
+      "loss": 1.1593,
+      "step": 6546
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8018452554512195e-05,
+      "loss": 1.0593,
+      "step": 6547
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801777926137694e-05,
+      "loss": 0.9905,
+      "step": 6548
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.801710586645878e-05,
+      "loss": 0.9804,
+      "step": 6549
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8016432369766263e-05,
+      "loss": 1.0124,
+      "step": 6550
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015758771307937e-05,
+      "loss": 0.9897,
+      "step": 6551
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8015085071092358e-05,
+      "loss": 1.1411,
+      "step": 6552
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8014411269128075e-05,
+      "loss": 1.0505,
+      "step": 6553
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013737365423642e-05,
+      "loss": 0.9201,
+      "step": 6554
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8013063359987614e-05,
+      "loss": 1.0212,
+      "step": 6555
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8012389252828547e-05,
+      "loss": 1.0142,
+      "step": 6556
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011715043955e-05,
+      "loss": 1.1241,
+      "step": 6557
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8011040733375532e-05,
+      "loss": 0.9842,
+      "step": 6558
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8010366321098698e-05,
+      "loss": 1.0151,
+      "step": 6559
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009691807133064e-05,
+      "loss": 1.1267,
+      "step": 6560
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8009017191487193e-05,
+      "loss": 1.0536,
+      "step": 6561
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8008342474169654e-05,
+      "loss": 1.2198,
+      "step": 6562
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8007667655189e-05,
+      "loss": 1.0816,
+      "step": 6563
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8006992734553806e-05,
+      "loss": 1.0016,
+      "step": 6564
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800631771227264e-05,
+      "loss": 1.1266,
+      "step": 6565
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8005642588354066e-05,
+      "loss": 1.0949,
+      "step": 6566
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800496736280666e-05,
+      "loss": 1.1,
+      "step": 6567
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8004292035638987e-05,
+      "loss": 1.067,
+      "step": 6568
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800361660685963e-05,
+      "loss": 1.0578,
+      "step": 6569
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8002941076477154e-05,
+      "loss": 0.9432,
+      "step": 6570
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.800226544450014e-05,
+      "loss": 1.0661,
+      "step": 6571
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8001589710937162e-05,
+      "loss": 1.1457,
+      "step": 6572
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.80009138757968e-05,
+      "loss": 1.085,
+      "step": 6573
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.8000237939087633e-05,
+      "loss": 1.0508,
+      "step": 6574
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799956190081824e-05,
+      "loss": 1.1261,
+      "step": 6575
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799888576099721e-05,
+      "loss": 1.1289,
+      "step": 6576
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7998209519633115e-05,
+      "loss": 1.0108,
+      "step": 6577
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7997533176734548e-05,
+      "loss": 1.1548,
+      "step": 6578
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7996856732310093e-05,
+      "loss": 1.1335,
+      "step": 6579
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799618018636834e-05,
+      "loss": 1.111,
+      "step": 6580
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799550353891787e-05,
+      "loss": 1.0931,
+      "step": 6581
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799482678996728e-05,
+      "loss": 1.0779,
+      "step": 6582
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799414993952516e-05,
+      "loss": 1.0568,
+      "step": 6583
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7993472987600098e-05,
+      "loss": 1.0083,
+      "step": 6584
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992795934200692e-05,
+      "loss": 1.0074,
+      "step": 6585
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7992118779335532e-05,
+      "loss": 1.0007,
+      "step": 6586
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799144152301322e-05,
+      "loss": 1.1185,
+      "step": 6587
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.799076416524235e-05,
+      "loss": 1.1689,
+      "step": 6588
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7990086706031524e-05,
+      "loss": 1.0637,
+      "step": 6589
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798940914538934e-05,
+      "loss": 1.0374,
+      "step": 6590
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7988731483324397e-05,
+      "loss": 1.0405,
+      "step": 6591
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7988053719845302e-05,
+      "loss": 1.0209,
+      "step": 6592
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7987375854960655e-05,
+      "loss": 1.013,
+      "step": 6593
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7986697888679067e-05,
+      "loss": 1.0367,
+      "step": 6594
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798601982100914e-05,
+      "loss": 1.0376,
+      "step": 6595
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798534165195948e-05,
+      "loss": 1.017,
+      "step": 6596
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7984663381538703e-05,
+      "loss": 1.0053,
+      "step": 6597
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983985009755412e-05,
+      "loss": 1.0429,
+      "step": 6598
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7983306536618222e-05,
+      "loss": 1.0425,
+      "step": 6599
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.798262796213575e-05,
+      "loss": 0.9903,
+      "step": 6600
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7981949286316603e-05,
+      "loss": 1.0993,
+      "step": 6601
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79812705091694e-05,
+      "loss": 1.1548,
+      "step": 6602
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7980591630702756e-05,
+      "loss": 1.0665,
+      "step": 6603
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7979912650925294e-05,
+      "loss": 1.1212,
+      "step": 6604
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797923356984563e-05,
+      "loss": 1.1454,
+      "step": 6605
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797855438747238e-05,
+      "loss": 1.0663,
+      "step": 6606
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797787510381418e-05,
+      "loss": 0.9528,
+      "step": 6607
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7977195718879635e-05,
+      "loss": 1.0549,
+      "step": 6608
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7976516232677383e-05,
+      "loss": 1.1551,
+      "step": 6609
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797583664521604e-05,
+      "loss": 1.0195,
+      "step": 6610
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7975156956504248e-05,
+      "loss": 1.0949,
+      "step": 6611
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.797447716655062e-05,
+      "loss": 1.1386,
+      "step": 6612
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7973797275363793e-05,
+      "loss": 1.026,
+      "step": 6613
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79731172829524e-05,
+      "loss": 1.0232,
+      "step": 6614
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7972437189325066e-05,
+      "loss": 1.197,
+      "step": 6615
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971756994490425e-05,
+      "loss": 1.0426,
+      "step": 6616
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7971076698457123e-05,
+      "loss": 1.1503,
+      "step": 6617
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7970396301233782e-05,
+      "loss": 0.9618,
+      "step": 6618
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969715802829052e-05,
+      "loss": 0.9636,
+      "step": 6619
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7969035203251562e-05,
+      "loss": 1.0678,
+      "step": 6620
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7968354502509956e-05,
+      "loss": 1.0348,
+      "step": 6621
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7967673700612874e-05,
+      "loss": 1.0756,
+      "step": 6622
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796699279756896e-05,
+      "loss": 1.0874,
+      "step": 6623
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796631179338686e-05,
+      "loss": 0.9044,
+      "step": 6624
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7965630688075216e-05,
+      "loss": 1.1042,
+      "step": 6625
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964949481642672e-05,
+      "loss": 1.0713,
+      "step": 6626
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7964268174097877e-05,
+      "loss": 1.2106,
+      "step": 6627
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7963586765449482e-05,
+      "loss": 1.1394,
+      "step": 6628
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796290525570614e-05,
+      "loss": 0.9084,
+      "step": 6629
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7962223644876498e-05,
+      "loss": 1.0637,
+      "step": 6630
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796154193296921e-05,
+      "loss": 1.0716,
+      "step": 6631
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7960860119992927e-05,
+      "loss": 0.9382,
+      "step": 6632
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.796017820595631e-05,
+      "loss": 1.138,
+      "step": 6633
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7959496190868014e-05,
+      "loss": 1.0752,
+      "step": 6634
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795881407473669e-05,
+      "loss": 1.0351,
+      "step": 6635
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7958131857571008e-05,
+      "loss": 1.0813,
+      "step": 6636
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7957449539379626e-05,
+      "loss": 0.9768,
+      "step": 6637
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79567671201712e-05,
+      "loss": 1.0897,
+      "step": 6638
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7956084599954397e-05,
+      "loss": 0.9924,
+      "step": 6639
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795540197873788e-05,
+      "loss": 1.0247,
+      "step": 6640
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954719256530316e-05,
+      "loss": 1.075,
+      "step": 6641
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7954036433340375e-05,
+      "loss": 1.053,
+      "step": 6642
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795335350917672e-05,
+      "loss": 1.1358,
+      "step": 6643
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795267048404802e-05,
+      "loss": 1.1621,
+      "step": 6644
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.795198735796295e-05,
+      "loss": 1.1022,
+      "step": 6645
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7951304130930182e-05,
+      "loss": 1.1553,
+      "step": 6646
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7950620802958383e-05,
+      "loss": 0.9769,
+      "step": 6647
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794993737405623e-05,
+      "loss": 1.0024,
+      "step": 6648
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7949253844232405e-05,
+      "loss": 1.1082,
+      "step": 6649
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7948570213495584e-05,
+      "loss": 1.0608,
+      "step": 6650
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7947886481854438e-05,
+      "loss": 1.087,
+      "step": 6651
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794720264931765e-05,
+      "loss": 1.0898,
+      "step": 6652
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7946518715893905e-05,
+      "loss": 1.0323,
+      "step": 6653
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794583468159188e-05,
+      "loss": 1.1412,
+      "step": 6654
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7945150546420265e-05,
+      "loss": 1.1319,
+      "step": 6655
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794446631038774e-05,
+      "loss": 1.1819,
+      "step": 6656
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794378197350299e-05,
+      "loss": 1.0952,
+      "step": 6657
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7943097535774705e-05,
+      "loss": 1.0992,
+      "step": 6658
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7942412997211573e-05,
+      "loss": 0.9189,
+      "step": 6659
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7941728357822287e-05,
+      "loss": 1.0943,
+      "step": 6660
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.794104361761553e-05,
+      "loss": 1.2383,
+      "step": 6661
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7940358776600003e-05,
+      "loss": 1.1763,
+      "step": 6662
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.79396738347844e-05,
+      "loss": 0.9678,
+      "step": 6663
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793898879217741e-05,
+      "loss": 1.0631,
+      "step": 6664
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793830364878773e-05,
+      "loss": 1.1311,
+      "step": 6665
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7937618404624063e-05,
+      "loss": 1.0722,
+      "step": 6666
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936933059695102e-05,
+      "loss": 1.2006,
+      "step": 6667
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7936247614009553e-05,
+      "loss": 0.8547,
+      "step": 6668
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7935562067576113e-05,
+      "loss": 1.0485,
+      "step": 6669
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934876420403486e-05,
+      "loss": 1.0505,
+      "step": 6670
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7934190672500376e-05,
+      "loss": 1.1155,
+      "step": 6671
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.793350482387549e-05,
+      "loss": 1.0809,
+      "step": 6672
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932818874537535e-05,
+      "loss": 1.0578,
+      "step": 6673
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7932132824495215e-05,
+      "loss": 1.0034,
+      "step": 6674
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7931446673757242e-05,
+      "loss": 1.1423,
+      "step": 6675
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930760422332327e-05,
+      "loss": 1.1722,
+      "step": 6676
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7930074070229174e-05,
+      "loss": 1.1525,
+      "step": 6677
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792938761745651e-05,
+      "loss": 1.0607,
+      "step": 6678
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792870106402304e-05,
+      "loss": 1.1261,
+      "step": 6679
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792801440993748e-05,
+      "loss": 1.0975,
+      "step": 6680
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7927327655208547e-05,
+      "loss": 1.1292,
+      "step": 6681
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7926640799844964e-05,
+      "loss": 1.2032,
+      "step": 6682
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7925953843855444e-05,
+      "loss": 1.0985,
+      "step": 6683
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.792526678724871e-05,
+      "loss": 1.0875,
+      "step": 6684
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7924579630033488e-05,
+      "loss": 1.0189,
+      "step": 6685
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923892372218496e-05,
+      "loss": 1.0375,
+      "step": 6686
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7923205013812456e-05,
+      "loss": 1.0772,
+      "step": 6687
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7922517554824104e-05,
+      "loss": 1.0769,
+      "step": 6688
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921829995262156e-05,
+      "loss": 1.1477,
+      "step": 6689
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7921142335135347e-05,
+      "loss": 1.022,
+      "step": 6690
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7920454574452405e-05,
+      "loss": 1.0683,
+      "step": 6691
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791976671322206e-05,
+      "loss": 1.0695,
+      "step": 6692
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7919078751453044e-05,
+      "loss": 0.9897,
+      "step": 6693
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7918390689154094e-05,
+      "loss": 1.1712,
+      "step": 6694
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791770252633394e-05,
+      "loss": 1.1599,
+      "step": 6695
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791701426300132e-05,
+      "loss": 1.112,
+      "step": 6696
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791632589916497e-05,
+      "loss": 1.0029,
+      "step": 6697
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7915637434833633e-05,
+      "loss": 1.0451,
+      "step": 6698
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914948870016046e-05,
+      "loss": 0.9717,
+      "step": 6699
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7914260204720944e-05,
+      "loss": 1.0638,
+      "step": 6700
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791357143895708e-05,
+      "loss": 1.0453,
+      "step": 6701
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791288257273319e-05,
+      "loss": 0.9568,
+      "step": 6702
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7912193606058025e-05,
+      "loss": 1.1582,
+      "step": 6703
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7911504538940322e-05,
+      "loss": 0.975,
+      "step": 6704
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791081537138884e-05,
+      "loss": 1.0681,
+      "step": 6705
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.791012610341232e-05,
+      "loss": 1.1845,
+      "step": 6706
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7909436735019515e-05,
+      "loss": 1.0913,
+      "step": 6707
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790874726621917e-05,
+      "loss": 1.104,
+      "step": 6708
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790805769702005e-05,
+      "loss": 0.9137,
+      "step": 6709
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7907368027430898e-05,
+      "loss": 1.1292,
+      "step": 6710
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790667825746047e-05,
+      "loss": 1.0159,
+      "step": 6711
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790598838711753e-05,
+      "loss": 0.9774,
+      "step": 6712
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790529841641083e-05,
+      "loss": 1.1792,
+      "step": 6713
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790460834534913e-05,
+      "loss": 1.0627,
+      "step": 6714
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790391817394119e-05,
+      "loss": 1.0602,
+      "step": 6715
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7903227902195774e-05,
+      "loss": 1.0578,
+      "step": 6716
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7902537530121636e-05,
+      "loss": 1.0907,
+      "step": 6717
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.790184705772755e-05,
+      "loss": 0.9755,
+      "step": 6718
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7901156485022277e-05,
+      "loss": 1.0142,
+      "step": 6719
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7900465812014583e-05,
+      "loss": 1.0731,
+      "step": 6720
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7899775038713237e-05,
+      "loss": 1.1282,
+      "step": 6721
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789908416512701e-05,
+      "loss": 1.1591,
+      "step": 6722
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789839319126467e-05,
+      "loss": 1.0853,
+      "step": 6723
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789770211713499e-05,
+      "loss": 0.9661,
+      "step": 6724
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789701094274674e-05,
+      "loss": 1.0773,
+      "step": 6725
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7896319668108698e-05,
+      "loss": 1.033,
+      "step": 6726
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7895628293229635e-05,
+      "loss": 0.8701,
+      "step": 6727
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894936818118335e-05,
+      "loss": 1.0378,
+      "step": 6728
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7894245242783568e-05,
+      "loss": 1.0281,
+      "step": 6729
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7893553567234118e-05,
+      "loss": 1.1138,
+      "step": 6730
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7892861791478766e-05,
+      "loss": 0.9917,
+      "step": 6731
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.789216991552629e-05,
+      "loss": 0.999,
+      "step": 6732
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7891477939385478e-05,
+      "loss": 0.9768,
+      "step": 6733
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890785863065113e-05,
+      "loss": 1.0065,
+      "step": 6734
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7890093686573978e-05,
+      "loss": 1.0616,
+      "step": 6735
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7889401409920857e-05,
+      "loss": 1.041,
+      "step": 6736
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7888709033114548e-05,
+      "loss": 1.1725,
+      "step": 6737
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7888016556163835e-05,
+      "loss": 1.1909,
+      "step": 6738
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.788732397907751e-05,
+      "loss": 1.0299,
+      "step": 6739
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7886631301864362e-05,
+      "loss": 0.9499,
+      "step": 6740
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885938524533184e-05,
+      "loss": 1.0722,
+      "step": 6741
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7885245647092774e-05,
+      "loss": 1.0283,
+      "step": 6742
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7884552669551932e-05,
+      "loss": 1.184,
+      "step": 6743
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883859591919442e-05,
+      "loss": 1.0175,
+      "step": 6744
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7883166414204118e-05,
+      "loss": 1.1316,
+      "step": 6745
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7882473136414744e-05,
+      "loss": 1.0258,
+      "step": 6746
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881779758560134e-05,
+      "loss": 1.1076,
+      "step": 6747
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7881086280649082e-05,
+      "loss": 1.0438,
+      "step": 6748
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7880392702690395e-05,
+      "loss": 1.0273,
+      "step": 6749
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787969902469288e-05,
+      "loss": 1.0517,
+      "step": 6750
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787900524666534e-05,
+      "loss": 1.084,
+      "step": 6751
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787831136861658e-05,
+      "loss": 0.9657,
+      "step": 6752
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.787761739055541e-05,
+      "loss": 1.0739,
+      "step": 6753
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876923312490642e-05,
+      "loss": 1.0802,
+      "step": 6754
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 1.7876229134431084e-05,
+      "loss": 1.0817,
+      "step": 6755
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787553485638555e-05,
+      "loss": 0.9561,
+      "step": 6756
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787484047836286e-05,
+      "loss": 1.154,
+      "step": 6757
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787414600037182e-05,
+      "loss": 1.0202,
+      "step": 6758
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7873451422421243e-05,
+      "loss": 1.0977,
+      "step": 6759
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787275674451996e-05,
+      "loss": 1.0941,
+      "step": 6760
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7872061966676776e-05,
+      "loss": 0.9727,
+      "step": 6761
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.787136708890052e-05,
+      "loss": 1.3093,
+      "step": 6762
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7870672111200008e-05,
+      "loss": 1.1595,
+      "step": 6763
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7869977033584067e-05,
+      "loss": 1.1383,
+      "step": 6764
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786928185606152e-05,
+      "loss": 1.117,
+      "step": 6765
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7868586578641185e-05,
+      "loss": 1.1309,
+      "step": 6766
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78678912013319e-05,
+      "loss": 1.0007,
+      "step": 6767
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786719572414248e-05,
+      "loss": 1.1188,
+      "step": 6768
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7866500147081765e-05,
+      "loss": 1.1394,
+      "step": 6769
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7865804470158575e-05,
+      "loss": 1.0901,
+      "step": 6770
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786510869338175e-05,
+      "loss": 1.1158,
+      "step": 6771
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7864412816760123e-05,
+      "loss": 1.1703,
+      "step": 6772
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7863716840302522e-05,
+      "loss": 1.1318,
+      "step": 6773
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786302076401778e-05,
+      "loss": 1.1516,
+      "step": 6774
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7862324587914743e-05,
+      "loss": 1.2275,
+      "step": 6775
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.786162831200224e-05,
+      "loss": 1.0804,
+      "step": 6776
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860931936289114e-05,
+      "loss": 1.0342,
+      "step": 6777
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7860235460784207e-05,
+      "loss": 1.103,
+      "step": 6778
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7859538885496356e-05,
+      "loss": 0.9498,
+      "step": 6779
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858842210434406e-05,
+      "loss": 1.0914,
+      "step": 6780
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7858145435607202e-05,
+      "loss": 1.1135,
+      "step": 6781
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785744856102359e-05,
+      "loss": 0.9406,
+      "step": 6782
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7856751586692415e-05,
+      "loss": 1.0753,
+      "step": 6783
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785605451262252e-05,
+      "loss": 0.9951,
+      "step": 6784
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785535733882276e-05,
+      "loss": 0.9789,
+      "step": 6785
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785466006530199e-05,
+      "loss": 1.1588,
+      "step": 6786
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785396269206905e-05,
+      "loss": 1.0789,
+      "step": 6787
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7853265219132802e-05,
+      "loss": 1.2028,
+      "step": 6788
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7852567646502097e-05,
+      "loss": 1.0343,
+      "step": 6789
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.785186997418579e-05,
+      "loss": 0.9025,
+      "step": 6790
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7851172202192737e-05,
+      "loss": 1.1335,
+      "step": 6791
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78504743305318e-05,
+      "loss": 1.0494,
+      "step": 6792
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849776359211834e-05,
+      "loss": 1.0802,
+      "step": 6793
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7849078288241698e-05,
+      "loss": 1.1816,
+      "step": 6794
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784838011763026e-05,
+      "loss": 1.1171,
+      "step": 6795
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784768184738638e-05,
+      "loss": 1.1267,
+      "step": 6796
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784698347751892e-05,
+      "loss": 0.9514,
+      "step": 6797
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7846285008036748e-05,
+      "loss": 1.0918,
+      "step": 6798
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784558643894873e-05,
+      "loss": 1.0337,
+      "step": 6799
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7844887770263736e-05,
+      "loss": 1.1236,
+      "step": 6800
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784418900199063e-05,
+      "loss": 1.203,
+      "step": 6801
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784349013413829e-05,
+      "loss": 0.9406,
+      "step": 6802
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7842791166715586e-05,
+      "loss": 1.1313,
+      "step": 6803
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784209209973139e-05,
+      "loss": 1.1007,
+      "step": 6804
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784139293319457e-05,
+      "loss": 1.0927,
+      "step": 6805
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.784069366711401e-05,
+      "loss": 1.0771,
+      "step": 6806
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839994301498588e-05,
+      "loss": 1.0399,
+      "step": 6807
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7839294836357175e-05,
+      "loss": 1.0522,
+      "step": 6808
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7838595271698654e-05,
+      "loss": 1.0743,
+      "step": 6809
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783789560753191e-05,
+      "loss": 1.1156,
+      "step": 6810
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7837195843865815e-05,
+      "loss": 1.1021,
+      "step": 6811
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7836495980709262e-05,
+      "loss": 1.0408,
+      "step": 6812
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783579601807113e-05,
+      "loss": 1.1306,
+      "step": 6813
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7835095955960308e-05,
+      "loss": 1.015,
+      "step": 6814
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783439579438568e-05,
+      "loss": 1.0221,
+      "step": 6815
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7833695533356137e-05,
+      "loss": 1.1175,
+      "step": 6816
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7832995172880566e-05,
+      "loss": 1.1823,
+      "step": 6817
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783229471296786e-05,
+      "loss": 0.9929,
+      "step": 6818
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783159415362691e-05,
+      "loss": 0.9965,
+      "step": 6819
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.783089349486661e-05,
+      "loss": 1.1425,
+      "step": 6820
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7830192736695856e-05,
+      "loss": 1.12,
+      "step": 6821
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782949187912354e-05,
+      "loss": 1.0642,
+      "step": 6822
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7828790922158563e-05,
+      "loss": 1.0135,
+      "step": 6823
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7828089865809817e-05,
+      "loss": 1.0883,
+      "step": 6824
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782738871008621e-05,
+      "loss": 1.0045,
+      "step": 6825
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782668745499664e-05,
+      "loss": 0.9753,
+      "step": 6826
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7825986100550007e-05,
+      "loss": 1.1913,
+      "step": 6827
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.782528464675522e-05,
+      "loss": 0.9872,
+      "step": 6828
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7824583093621175e-05,
+      "loss": 0.9884,
+      "step": 6829
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7823881441156784e-05,
+      "loss": 1.1724,
+      "step": 6830
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7823179689370952e-05,
+      "loss": 1.0665,
+      "step": 6831
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7822477838272588e-05,
+      "loss": 0.9593,
+      "step": 6832
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821775887870602e-05,
+      "loss": 1.0895,
+      "step": 6833
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7821073838173908e-05,
+      "loss": 1.1186,
+      "step": 6834
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7820371689191414e-05,
+      "loss": 1.1235,
+      "step": 6835
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7819669440932033e-05,
+      "loss": 1.132,
+      "step": 6836
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781896709340468e-05,
+      "loss": 0.9608,
+      "step": 6837
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781826464661828e-05,
+      "loss": 1.0884,
+      "step": 6838
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7817562100581737e-05,
+      "loss": 1.0032,
+      "step": 6839
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7816859455303977e-05,
+      "loss": 0.9379,
+      "step": 6840
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781615671079392e-05,
+      "loss": 1.02,
+      "step": 6841
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7815453867060483e-05,
+      "loss": 0.9957,
+      "step": 6842
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7814750924112594e-05,
+      "loss": 1.0047,
+      "step": 6843
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781404788195917e-05,
+      "loss": 1.132,
+      "step": 6844
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7813344740609144e-05,
+      "loss": 1.0338,
+      "step": 6845
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7812641500071435e-05,
+      "loss": 1.0964,
+      "step": 6846
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.781193816035497e-05,
+      "loss": 1.18,
+      "step": 6847
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7811234721468684e-05,
+      "loss": 1.1693,
+      "step": 6848
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7810531183421505e-05,
+      "loss": 0.9754,
+      "step": 6849
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7809827546222363e-05,
+      "loss": 1.0334,
+      "step": 6850
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7809123809880183e-05,
+      "loss": 0.9571,
+      "step": 6851
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7808419974403913e-05,
+      "loss": 1.142,
+      "step": 6852
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807716039802477e-05,
+      "loss": 1.0374,
+      "step": 6853
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7807012006084817e-05,
+      "loss": 0.954,
+      "step": 6854
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780630787325987e-05,
+      "loss": 1.0432,
+      "step": 6855
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7805603641336572e-05,
+      "loss": 0.9741,
+      "step": 6856
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804899310323863e-05,
+      "loss": 1.1162,
+      "step": 6857
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7804194880230684e-05,
+      "loss": 1.0436,
+      "step": 6858
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780349035106598e-05,
+      "loss": 1.055,
+      "step": 6859
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7802785722838694e-05,
+      "loss": 1.0397,
+      "step": 6860
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.780208099555777e-05,
+      "loss": 1.062,
+      "step": 6861
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7801376169232158e-05,
+      "loss": 1.1604,
+      "step": 6862
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.78006712438708e-05,
+      "loss": 1.0813,
+      "step": 6863
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799966219482648e-05,
+      "loss": 0.9045,
+      "step": 6864
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7799261096076647e-05,
+      "loss": 1.0396,
+      "step": 6865
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779855587366176e-05,
+      "loss": 1.0597,
+      "step": 6866
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797850552246926e-05,
+      "loss": 1.0312,
+      "step": 6867
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7797145131841104e-05,
+      "loss": 1.0781,
+      "step": 6868
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7796439612453252e-05,
+      "loss": 1.1605,
+      "step": 6869
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7795733994092324e-05,
+      "loss": 1.232,
+      "step": 6870
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7795028276767282e-05,
+      "loss": 1.0957,
+      "step": 6871
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7794322460487077e-05,
+      "loss": 1.1312,
+      "step": 6872
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7793616545260674e-05,
+      "loss": 1.0642,
+      "step": 6873
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.779291053109703e-05,
+      "loss": 1.0975,
+      "step": 6874
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7792204418005115e-05,
+      "loss": 0.9919,
+      "step": 6875
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7791498205993885e-05,
+      "loss": 1.0717,
+      "step": 6876
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790791895072314e-05,
+      "loss": 1.2008,
+      "step": 6877
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7790085485249356e-05,
+      "loss": 0.9391,
+      "step": 6878
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7789378976533988e-05,
+      "loss": 1.0485,
+      "step": 6879
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778867236893518e-05,
+      "loss": 1.0941,
+      "step": 6880
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7787965662461894e-05,
+      "loss": 1.0176,
+      "step": 6881
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778725885712311e-05,
+      "loss": 1.0057,
+      "step": 6882
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7786551952927796e-05,
+      "loss": 1.1944,
+      "step": 6883
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785844949884924e-05,
+      "loss": 1.0932,
+      "step": 6884
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7785137848003472e-05,
+      "loss": 1.0558,
+      "step": 6885
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7784430647292418e-05,
+      "loss": 1.1478,
+      "step": 6886
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778372334776074e-05,
+      "loss": 1.1468,
+      "step": 6887
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778301594941741e-05,
+      "loss": 1.0589,
+      "step": 6888
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778230845227142e-05,
+      "loss": 1.0883,
+      "step": 6889
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.778160085633174e-05,
+      "loss": 1.1116,
+      "step": 6890
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7780893161607353e-05,
+      "loss": 1.0347,
+      "step": 6891
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7780185368107253e-05,
+      "loss": 0.9215,
+      "step": 6892
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7779477475840418e-05,
+      "loss": 1.0225,
+      "step": 6893
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7778769484815833e-05,
+      "loss": 1.0508,
+      "step": 6894
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777806139504249e-05,
+      "loss": 1.0531,
+      "step": 6895
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7777353206529375e-05,
+      "loss": 1.0505,
+      "step": 6896
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7776644919285483e-05,
+      "loss": 1.0501,
+      "step": 6897
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7775936533319796e-05,
+      "loss": 1.1649,
+      "step": 6898
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777522804864132e-05,
+      "loss": 0.9645,
+      "step": 6899
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7774519465259036e-05,
+      "loss": 1.0001,
+      "step": 6900
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7773810783181947e-05,
+      "loss": 1.1097,
+      "step": 6901
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7773102002419048e-05,
+      "loss": 1.0349,
+      "step": 6902
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777239312297934e-05,
+      "loss": 1.1009,
+      "step": 6903
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777168414487181e-05,
+      "loss": 1.0785,
+      "step": 6904
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.777097506810547e-05,
+      "loss": 1.0813,
+      "step": 6905
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7770265892689317e-05,
+      "loss": 1.117,
+      "step": 6906
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7769556618632356e-05,
+      "loss": 0.987,
+      "step": 6907
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776884724594359e-05,
+      "loss": 1.0221,
+      "step": 6908
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776813777463202e-05,
+      "loss": 1.1318,
+      "step": 6909
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776742820470666e-05,
+      "loss": 1.1902,
+      "step": 6910
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776671853617651e-05,
+      "loss": 1.0022,
+      "step": 6911
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7766008769050585e-05,
+      "loss": 0.979,
+      "step": 6912
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7765298903337896e-05,
+      "loss": 0.9137,
+      "step": 6913
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776458893904745e-05,
+      "loss": 1.1031,
+      "step": 6914
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776387887618826e-05,
+      "loss": 1.099,
+      "step": 6915
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7763168714769337e-05,
+      "loss": 1.1379,
+      "step": 6916
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7762458454799708e-05,
+      "loss": 1.097,
+      "step": 6917
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.776174809628838e-05,
+      "loss": 1.1388,
+      "step": 6918
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7761037639244374e-05,
+      "loss": 1.119,
+      "step": 6919
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7760327083676708e-05,
+      "loss": 0.9254,
+      "step": 6920
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7759616429594398e-05,
+      "loss": 1.0296,
+      "step": 6921
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7758905677006476e-05,
+      "loss": 1.0755,
+      "step": 6922
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775819482592195e-05,
+      "loss": 1.0585,
+      "step": 6923
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7757483876349863e-05,
+      "loss": 1.0485,
+      "step": 6924
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756772828299224e-05,
+      "loss": 1.1389,
+      "step": 6925
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7756061681779063e-05,
+      "loss": 1.1253,
+      "step": 6926
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7755350436798414e-05,
+      "loss": 0.9943,
+      "step": 6927
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7754639093366298e-05,
+      "loss": 1.0203,
+      "step": 6928
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7753927651491753e-05,
+      "loss": 0.9951,
+      "step": 6929
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7753216111183803e-05,
+      "loss": 1.0859,
+      "step": 6930
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775250447245149e-05,
+      "loss": 1.0001,
+      "step": 6931
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.775179273530384e-05,
+      "loss": 0.9648,
+      "step": 6932
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7751080899749888e-05,
+      "loss": 1.1101,
+      "step": 6933
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7750368965798674e-05,
+      "loss": 1.0465,
+      "step": 6934
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774965693345924e-05,
+      "loss": 1.0147,
+      "step": 6935
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748944802740614e-05,
+      "loss": 1.0393,
+      "step": 6936
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7748232573651845e-05,
+      "loss": 1.0966,
+      "step": 6937
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774752024620197e-05,
+      "loss": 0.9669,
+      "step": 6938
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746807820400036e-05,
+      "loss": 1.1823,
+      "step": 6939
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7746095296255085e-05,
+      "loss": 1.065,
+      "step": 6940
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7745382673776158e-05,
+      "loss": 1.0435,
+      "step": 6941
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7744669952972308e-05,
+      "loss": 1.1619,
+      "step": 6942
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774395713385258e-05,
+      "loss": 1.0998,
+      "step": 6943
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7743244216426022e-05,
+      "loss": 1.1725,
+      "step": 6944
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.774253120070169e-05,
+      "loss": 1.1093,
+      "step": 6945
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741818086688624e-05,
+      "loss": 1.1653,
+      "step": 6946
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7741104874395885e-05,
+      "loss": 1.0473,
+      "step": 6947
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7740391563832525e-05,
+      "loss": 1.1255,
+      "step": 6948
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7739678155007604e-05,
+      "loss": 1.1207,
+      "step": 6949
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738964647930168e-05,
+      "loss": 1.0147,
+      "step": 6950
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7738251042609286e-05,
+      "loss": 1.1752,
+      "step": 6951
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7737537339054006e-05,
+      "loss": 1.1114,
+      "step": 6952
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77368235372734e-05,
+      "loss": 1.1059,
+      "step": 6953
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7736109637276523e-05,
+      "loss": 0.8886,
+      "step": 6954
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7735395639072436e-05,
+      "loss": 1.0645,
+      "step": 6955
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7734681542670204e-05,
+      "loss": 1.0212,
+      "step": 6956
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733967348078896e-05,
+      "loss": 1.103,
+      "step": 6957
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7733253055307572e-05,
+      "loss": 1.04,
+      "step": 6958
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773253866436531e-05,
+      "loss": 0.9323,
+      "step": 6959
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7731824175261167e-05,
+      "loss": 0.9091,
+      "step": 6960
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773110958800422e-05,
+      "loss": 1.1005,
+      "step": 6961
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.773039490260354e-05,
+      "loss": 1.0031,
+      "step": 6962
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7729680119068197e-05,
+      "loss": 1.0001,
+      "step": 6963
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772896523740727e-05,
+      "loss": 1.0527,
+      "step": 6964
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7728250257629825e-05,
+      "loss": 1.1533,
+      "step": 6965
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7727535179744954e-05,
+      "loss": 1.0949,
+      "step": 6966
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7726820003761716e-05,
+      "loss": 0.995,
+      "step": 6967
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.77261047296892e-05,
+      "loss": 0.9619,
+      "step": 6968
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7725389357536488e-05,
+      "loss": 1.0608,
+      "step": 6969
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7724673887312657e-05,
+      "loss": 1.1289,
+      "step": 6970
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723958319026794e-05,
+      "loss": 0.9933,
+      "step": 6971
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7723242652687975e-05,
+      "loss": 1.0981,
+      "step": 6972
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7722526888305295e-05,
+      "loss": 1.0376,
+      "step": 6973
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772181102588783e-05,
+      "loss": 1.1241,
+      "step": 6974
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7721095065444678e-05,
+      "loss": 1.0949,
+      "step": 6975
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.772037900698492e-05,
+      "loss": 1.1246,
+      "step": 6976
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771966285051765e-05,
+      "loss": 1.0721,
+      "step": 6977
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771894659605196e-05,
+      "loss": 1.2141,
+      "step": 6978
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771823024359694e-05,
+      "loss": 1.0877,
+      "step": 6979
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771751379316168e-05,
+      "loss": 1.0704,
+      "step": 6980
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716797244755288e-05,
+      "loss": 1.0011,
+      "step": 6981
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7716080598386846e-05,
+      "loss": 1.0282,
+      "step": 6982
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771536385406546e-05,
+      "loss": 1.1344,
+      "step": 6983
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771464701180023e-05,
+      "loss": 1.0538,
+      "step": 6984
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771393007160025e-05,
+      "loss": 1.074,
+      "step": 6985
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771321303347462e-05,
+      "loss": 1.1321,
+      "step": 6986
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771249589743245e-05,
+      "loss": 1.1711,
+      "step": 6987
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.771177866348284e-05,
+      "loss": 1.0483,
+      "step": 6988
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7711061331634897e-05,
+      "loss": 1.0251,
+      "step": 6989
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7710343901897727e-05,
+      "loss": 1.0145,
+      "step": 6990
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7709626374280432e-05,
+      "loss": 1.0045,
+      "step": 6991
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770890874879213e-05,
+      "loss": 1.0177,
+      "step": 6992
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7708191025441922e-05,
+      "loss": 1.0617,
+      "step": 6993
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7707473204238924e-05,
+      "loss": 1.0163,
+      "step": 6994
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706755285192248e-05,
+      "loss": 1.1029,
+      "step": 6995
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7706037268311006e-05,
+      "loss": 0.9624,
+      "step": 6996
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7705319153604316e-05,
+      "loss": 1.0086,
+      "step": 6997
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7704600941081292e-05,
+      "loss": 0.8555,
+      "step": 6998
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703882630751056e-05,
+      "loss": 1.0834,
+      "step": 6999
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7703164222622717e-05,
+      "loss": 1.0347,
+      "step": 7000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7702445716705403e-05,
+      "loss": 1.0328,
+      "step": 7001
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7701727113008233e-05,
+      "loss": 1.1084,
+      "step": 7002
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.770100841154033e-05,
+      "loss": 1.0555,
+      "step": 7003
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7700289612310818e-05,
+      "loss": 1.0972,
+      "step": 7004
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7699570715328818e-05,
+      "loss": 1.1375,
+      "step": 7005
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769885172060346e-05,
+      "loss": 1.0874,
+      "step": 7006
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769813262814387e-05,
+      "loss": 0.9191,
+      "step": 7007
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7697413437959178e-05,
+      "loss": 1.1661,
+      "step": 7008
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7696694150058514e-05,
+      "loss": 1.0072,
+      "step": 7009
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7695974764451006e-05,
+      "loss": 1.2279,
+      "step": 7010
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769525528114579e-05,
+      "loss": 0.8949,
+      "step": 7011
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7694535700151997e-05,
+      "loss": 1.1112,
+      "step": 7012
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693816021478764e-05,
+      "loss": 1.0927,
+      "step": 7013
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7693096245135222e-05,
+      "loss": 1.1851,
+      "step": 7014
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7692376371130516e-05,
+      "loss": 1.1538,
+      "step": 7015
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.769165639947378e-05,
+      "loss": 1.1016,
+      "step": 7016
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690936330174153e-05,
+      "loss": 1.1207,
+      "step": 7017
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7690216163240782e-05,
+      "loss": 1.0683,
+      "step": 7018
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7689495898682797e-05,
+      "loss": 1.0481,
+      "step": 7019
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688775536509355e-05,
+      "loss": 0.9778,
+      "step": 7020
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7688055076729592e-05,
+      "loss": 1.0951,
+      "step": 7021
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768733451935266e-05,
+      "loss": 1.163,
+      "step": 7022
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7686613864387704e-05,
+      "loss": 1.0758,
+      "step": 7023
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768589311184387e-05,
+      "loss": 1.0889,
+      "step": 7024
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768517226173031e-05,
+      "loss": 1.1735,
+      "step": 7025
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7684451314056174e-05,
+      "loss": 1.1697,
+      "step": 7026
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683730268830612e-05,
+      "loss": 1.0763,
+      "step": 7027
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7683009126062783e-05,
+      "loss": 1.0598,
+      "step": 7028
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768228788576184e-05,
+      "loss": 0.9734,
+      "step": 7029
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7681566547936934e-05,
+      "loss": 1.1838,
+      "step": 7030
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.768084511259723e-05,
+      "loss": 1.0241,
+      "step": 7031
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7680123579751877e-05,
+      "loss": 1.0221,
+      "step": 7032
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7679401949410043e-05,
+      "loss": 1.2189,
+      "step": 7033
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7678680221580884e-05,
+      "loss": 1.1835,
+      "step": 7034
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677958396273564e-05,
+      "loss": 1.0464,
+      "step": 7035
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7677236473497244e-05,
+      "loss": 1.106,
+      "step": 7036
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7676514453261094e-05,
+      "loss": 1.0883,
+      "step": 7037
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675792335574277e-05,
+      "loss": 0.9752,
+      "step": 7038
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7675070120445955e-05,
+      "loss": 1.2426,
+      "step": 7039
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.76743478078853e-05,
+      "loss": 1.2244,
+      "step": 7040
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7673625397901487e-05,
+      "loss": 1.0409,
+      "step": 7041
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.7672902890503678e-05,
+      "loss": 1.0053,
+      "step": 7042
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767218028570105e-05,
+      "loss": 1.138,
+      "step": 7043
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7671457583502777e-05,
+      "loss": 1.1378,
+      "step": 7044
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.767073478391803e-05,
+      "loss": 0.994,
+      "step": 7045
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7670011886955984e-05,
+      "loss": 0.9673,
+      "step": 7046
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766928889262582e-05,
+      "loss": 1.0367,
+      "step": 7047
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7668565800936713e-05,
+      "loss": 1.0018,
+      "step": 7048
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667842611897847e-05,
+      "loss": 1.004,
+      "step": 7049
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7667119325518395e-05,
+      "loss": 1.0221,
+      "step": 7050
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7666395941807542e-05,
+      "loss": 1.0699,
+      "step": 7051
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7665672460774478e-05,
+      "loss": 0.9234,
+      "step": 7052
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664948882428376e-05,
+      "loss": 1.14,
+      "step": 7053
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7664225206778433e-05,
+      "loss": 0.9367,
+      "step": 7054
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7663501433833823e-05,
+      "loss": 1.2059,
+      "step": 7055
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7662777563603746e-05,
+      "loss": 0.9778,
+      "step": 7056
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7662053596097383e-05,
+      "loss": 1.0049,
+      "step": 7057
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.766132953132393e-05,
+      "loss": 0.9849,
+      "step": 7058
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7660605369292577e-05,
+      "loss": 1.1111,
+      "step": 7059
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659881110012515e-05,
+      "loss": 1.16,
+      "step": 7060
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7659156753492942e-05,
+      "loss": 1.039,
+      "step": 7061
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765843229974305e-05,
+      "loss": 1.1071,
+      "step": 7062
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7657707748772034e-05,
+      "loss": 0.995,
+      "step": 7063
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656983100589097e-05,
+      "loss": 1.0362,
+      "step": 7064
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7656258355203437e-05,
+      "loss": 1.1528,
+      "step": 7065
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7655533512624252e-05,
+      "loss": 1.0754,
+      "step": 7066
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7654808572860743e-05,
+      "loss": 1.1058,
+      "step": 7067
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765408353592212e-05,
+      "loss": 1.027,
+      "step": 7068
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7653358401817578e-05,
+      "loss": 1.0379,
+      "step": 7069
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7652633170556326e-05,
+      "loss": 1.0844,
+      "step": 7070
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.765190784214757e-05,
+      "loss": 1.1491,
+      "step": 7071
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7651182416600523e-05,
+      "loss": 1.0199,
+      "step": 7072
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7650456893924386e-05,
+      "loss": 0.9677,
+      "step": 7073
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649731274128376e-05,
+      "loss": 0.9494,
+      "step": 7074
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7649005557221695e-05,
+      "loss": 1.0429,
+      "step": 7075
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764827974321357e-05,
+      "loss": 1.1533,
+      "step": 7076
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7647553832113202e-05,
+      "loss": 1.1381,
+      "step": 7077
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7646827823929814e-05,
+      "loss": 1.1054,
+      "step": 7078
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7646101718672615e-05,
+      "loss": 1.0103,
+      "step": 7079
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7645375516350834e-05,
+      "loss": 1.1719,
+      "step": 7080
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7644649216973677e-05,
+      "loss": 0.9577,
+      "step": 7081
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7643922820550373e-05,
+      "loss": 0.9964,
+      "step": 7082
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.764319632709014e-05,
+      "loss": 0.9129,
+      "step": 7083
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76424697366022e-05,
+      "loss": 0.9489,
+      "step": 7084
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641743049095783e-05,
+      "loss": 1.0667,
+      "step": 7085
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7641016264580106e-05,
+      "loss": 0.983,
+      "step": 7086
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7640289383064397e-05,
+      "loss": 1.1363,
+      "step": 7087
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7639562404557886e-05,
+      "loss": 1.0635,
+      "step": 7088
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76388353290698e-05,
+      "loss": 1.0735,
+      "step": 7089
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763810815660937e-05,
+      "loss": 1.06,
+      "step": 7090
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7637380887185826e-05,
+      "loss": 1.0923,
+      "step": 7091
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76366535208084e-05,
+      "loss": 1.0012,
+      "step": 7092
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635926057486333e-05,
+      "loss": 1.0277,
+      "step": 7093
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7635198497228846e-05,
+      "loss": 1.0668,
+      "step": 7094
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763447084004519e-05,
+      "loss": 1.1603,
+      "step": 7095
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763374308594459e-05,
+      "loss": 0.9403,
+      "step": 7096
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7633015234936294e-05,
+      "loss": 1.1183,
+      "step": 7097
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7632287287029533e-05,
+      "loss": 1.0002,
+      "step": 7098
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7631559242233554e-05,
+      "loss": 1.0446,
+      "step": 7099
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76308311005576e-05,
+      "loss": 1.0963,
+      "step": 7100
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.763010286201091e-05,
+      "loss": 0.8858,
+      "step": 7101
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7629374526602735e-05,
+      "loss": 1.025,
+      "step": 7102
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762864609434231e-05,
+      "loss": 1.0283,
+      "step": 7103
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7627917565238893e-05,
+      "loss": 1.0558,
+      "step": 7104
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762718893930173e-05,
+      "loss": 1.0874,
+      "step": 7105
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762646021654007e-05,
+      "loss": 1.0488,
+      "step": 7106
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762573139696316e-05,
+      "loss": 1.0535,
+      "step": 7107
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762500248058026e-05,
+      "loss": 1.0919,
+      "step": 7108
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7624273467400613e-05,
+      "loss": 0.9414,
+      "step": 7109
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7623544357433482e-05,
+      "loss": 0.9646,
+      "step": 7110
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7622815150688125e-05,
+      "loss": 1.1795,
+      "step": 7111
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762208584717379e-05,
+      "loss": 1.1166,
+      "step": 7112
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7621356446899736e-05,
+      "loss": 1.2065,
+      "step": 7113
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.762062694987523e-05,
+      "loss": 0.9744,
+      "step": 7114
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619897356109526e-05,
+      "loss": 0.9935,
+      "step": 7115
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7619167665611892e-05,
+      "loss": 1.0786,
+      "step": 7116
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7618437878391586e-05,
+      "loss": 1.0695,
+      "step": 7117
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7617707994457875e-05,
+      "loss": 1.1336,
+      "step": 7118
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761697801382002e-05,
+      "loss": 1.1431,
+      "step": 7119
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7616247936487295e-05,
+      "loss": 0.9991,
+      "step": 7120
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.761551776246896e-05,
+      "loss": 1.1081,
+      "step": 7121
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614787491774296e-05,
+      "loss": 1.0494,
+      "step": 7122
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7614057124412562e-05,
+      "loss": 1.0232,
+      "step": 7123
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7613326660393033e-05,
+      "loss": 0.9923,
+      "step": 7124
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7612596099724987e-05,
+      "loss": 1.177,
+      "step": 7125
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611865442417693e-05,
+      "loss": 1.0677,
+      "step": 7126
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7611134688480426e-05,
+      "loss": 1.1338,
+      "step": 7127
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7610403837922466e-05,
+      "loss": 1.1152,
+      "step": 7128
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760967289075309e-05,
+      "loss": 1.092,
+      "step": 7129
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760894184698158e-05,
+      "loss": 1.04,
+      "step": 7130
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7608210706617205e-05,
+      "loss": 1.118,
+      "step": 7131
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760747946966926e-05,
+      "loss": 1.0823,
+      "step": 7132
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7606748136147017e-05,
+      "loss": 1.0277,
+      "step": 7133
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760601670605977e-05,
+      "loss": 1.1878,
+      "step": 7134
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76052851794168e-05,
+      "loss": 0.9802,
+      "step": 7135
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760455355622739e-05,
+      "loss": 1.1426,
+      "step": 7136
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.760382183650083e-05,
+      "loss": 0.9825,
+      "step": 7137
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7603090020246412e-05,
+      "loss": 0.9933,
+      "step": 7138
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7602358107473422e-05,
+      "loss": 1.0644,
+      "step": 7139
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7601626098191155e-05,
+      "loss": 1.1588,
+      "step": 7140
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.76008939924089e-05,
+      "loss": 1.2173,
+      "step": 7141
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7600161790135953e-05,
+      "loss": 1.1019,
+      "step": 7142
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7599429491381606e-05,
+      "loss": 0.9754,
+      "step": 7143
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7598697096155163e-05,
+      "loss": 1.1402,
+      "step": 7144
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7597964604465915e-05,
+      "loss": 0.9812,
+      "step": 7145
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759723201632316e-05,
+      "loss": 1.1405,
+      "step": 7146
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7596499331736202e-05,
+      "loss": 1.135,
+      "step": 7147
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7595766550714343e-05,
+      "loss": 1.0501,
+      "step": 7148
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759503367326688e-05,
+      "loss": 1.0396,
+      "step": 7149
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7594300699403116e-05,
+      "loss": 1.0928,
+      "step": 7150
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7593567629132363e-05,
+      "loss": 1.1279,
+      "step": 7151
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759283446246392e-05,
+      "loss": 1.0343,
+      "step": 7152
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75921011994071e-05,
+      "loss": 1.2727,
+      "step": 7153
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.759136783997121e-05,
+      "loss": 1.1252,
+      "step": 7154
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7590634384165556e-05,
+      "loss": 1.0029,
+      "step": 7155
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7589900831999453e-05,
+      "loss": 0.8747,
+      "step": 7156
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758916718348221e-05,
+      "loss": 1.0237,
+      "step": 7157
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7588433438623146e-05,
+      "loss": 1.1238,
+      "step": 7158
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758769959743157e-05,
+      "loss": 0.9864,
+      "step": 7159
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75869656599168e-05,
+      "loss": 1.049,
+      "step": 7160
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758623162608815e-05,
+      "loss": 0.995,
+      "step": 7161
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7585497495954946e-05,
+      "loss": 1.1259,
+      "step": 7162
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75847632695265e-05,
+      "loss": 1.1135,
+      "step": 7163
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7584028946812134e-05,
+      "loss": 0.8723,
+      "step": 7164
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7583294527821176e-05,
+      "loss": 1.1335,
+      "step": 7165
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7582560012562943e-05,
+      "loss": 1.0524,
+      "step": 7166
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758182540104676e-05,
+      "loss": 1.0301,
+      "step": 7167
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758109069328195e-05,
+      "loss": 1.1108,
+      "step": 7168
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.758035588927785e-05,
+      "loss": 1.2372,
+      "step": 7169
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757962098904378e-05,
+      "loss": 1.0864,
+      "step": 7170
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7578885992589065e-05,
+      "loss": 1.0765,
+      "step": 7171
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7578150899923047e-05,
+      "loss": 1.2118,
+      "step": 7172
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757741571105505e-05,
+      "loss": 0.9761,
+      "step": 7173
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7576680425994414e-05,
+      "loss": 1.173,
+      "step": 7174
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7575945044750464e-05,
+      "loss": 1.2053,
+      "step": 7175
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.757520956733254e-05,
+      "loss": 1.0987,
+      "step": 7176
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7574473993749976e-05,
+      "loss": 1.0178,
+      "step": 7177
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573738324012116e-05,
+      "loss": 1.0402,
+      "step": 7178
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7573002558128296e-05,
+      "loss": 1.1103,
+      "step": 7179
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7572266696107852e-05,
+      "loss": 1.0326,
+      "step": 7180
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7571530737960133e-05,
+      "loss": 1.1361,
+      "step": 7181
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7570794683694473e-05,
+      "loss": 1.1141,
+      "step": 7182
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7570058533320226e-05,
+      "loss": 0.987,
+      "step": 7183
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7569322286846727e-05,
+      "loss": 1.1818,
+      "step": 7184
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7568585944283333e-05,
+      "loss": 1.1046,
+      "step": 7185
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.756784950563938e-05,
+      "loss": 1.0579,
+      "step": 7186
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7567112970924224e-05,
+      "loss": 1.0232,
+      "step": 7187
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7566376340147218e-05,
+      "loss": 1.061,
+      "step": 7188
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7565639613317706e-05,
+      "loss": 1.1861,
+      "step": 7189
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564902790445045e-05,
+      "loss": 0.9894,
+      "step": 7190
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7564165871538584e-05,
+      "loss": 1.1343,
+      "step": 7191
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7563428856607684e-05,
+      "loss": 1.0059,
+      "step": 7192
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7562691745661697e-05,
+      "loss": 1.072,
+      "step": 7193
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561954538709983e-05,
+      "loss": 1.075,
+      "step": 7194
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7561217235761897e-05,
+      "loss": 1.042,
+      "step": 7195
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7560479836826805e-05,
+      "loss": 0.8835,
+      "step": 7196
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755974234191406e-05,
+      "loss": 0.9149,
+      "step": 7197
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755900475103303e-05,
+      "loss": 1.0672,
+      "step": 7198
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755826706419308e-05,
+      "loss": 0.9832,
+      "step": 7199
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7557529281403568e-05,
+      "loss": 1.2074,
+      "step": 7200
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556791402673862e-05,
+      "loss": 0.9942,
+      "step": 7201
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7556053428013332e-05,
+      "loss": 1.178,
+      "step": 7202
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7555315357431348e-05,
+      "loss": 1.1961,
+      "step": 7203
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7554577190937273e-05,
+      "loss": 0.9835,
+      "step": 7204
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7553838928540483e-05,
+      "loss": 1.0588,
+      "step": 7205
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755310057025035e-05,
+      "loss": 1.0075,
+      "step": 7206
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755236211607624e-05,
+      "loss": 1.0572,
+      "step": 7207
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7551623566027537e-05,
+      "loss": 1.0834,
+      "step": 7208
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.755088492011361e-05,
+      "loss": 1.0189,
+      "step": 7209
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7550146178343836e-05,
+      "loss": 0.953,
+      "step": 7210
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75494073407276e-05,
+      "loss": 0.9808,
+      "step": 7211
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7548668407274274e-05,
+      "loss": 1.018,
+      "step": 7212
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754792937799324e-05,
+      "loss": 0.9676,
+      "step": 7213
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7547190252893883e-05,
+      "loss": 1.1222,
+      "step": 7214
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7546451031985585e-05,
+      "loss": 1.0496,
+      "step": 7215
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754571171527773e-05,
+      "loss": 1.014,
+      "step": 7216
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75449723027797e-05,
+      "loss": 1.075,
+      "step": 7217
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754423279450088e-05,
+      "loss": 1.0607,
+      "step": 7218
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7543493190450665e-05,
+      "loss": 1.1217,
+      "step": 7219
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7542753490638443e-05,
+      "loss": 0.9563,
+      "step": 7220
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7542013695073598e-05,
+      "loss": 0.8974,
+      "step": 7221
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754127380376553e-05,
+      "loss": 1.0364,
+      "step": 7222
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.754053381672362e-05,
+      "loss": 1.0721,
+      "step": 7223
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539793733957276e-05,
+      "loss": 0.9717,
+      "step": 7224
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7539053555475883e-05,
+      "loss": 0.9973,
+      "step": 7225
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753831328128884e-05,
+      "loss": 1.1399,
+      "step": 7226
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7537572911405543e-05,
+      "loss": 1.0722,
+      "step": 7227
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536832445835394e-05,
+      "loss": 1.1142,
+      "step": 7228
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7536091884587795e-05,
+      "loss": 1.1086,
+      "step": 7229
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753535122767214e-05,
+      "loss": 1.1127,
+      "step": 7230
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7534610475097836e-05,
+      "loss": 1.0225,
+      "step": 7231
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533869626874285e-05,
+      "loss": 1.0756,
+      "step": 7232
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7533128683010895e-05,
+      "loss": 1.0472,
+      "step": 7233
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7532387643517066e-05,
+      "loss": 1.1604,
+      "step": 7234
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7531646508402208e-05,
+      "loss": 1.0051,
+      "step": 7235
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7530905277675735e-05,
+      "loss": 1.0158,
+      "step": 7236
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.753016395134705e-05,
+      "loss": 1.0036,
+      "step": 7237
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7529422529425562e-05,
+      "loss": 1.1268,
+      "step": 7238
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7528681011920693e-05,
+      "loss": 1.0534,
+      "step": 7239
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7527939398841844e-05,
+      "loss": 1.0055,
+      "step": 7240
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752719769019844e-05,
+      "loss": 1.1156,
+      "step": 7241
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752645588599989e-05,
+      "loss": 1.0072,
+      "step": 7242
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7525713986255612e-05,
+      "loss": 1.0867,
+      "step": 7243
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752497199097503e-05,
+      "loss": 1.0254,
+      "step": 7244
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7524229900167554e-05,
+      "loss": 1.1558,
+      "step": 7245
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7523487713842614e-05,
+      "loss": 1.006,
+      "step": 7246
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7522745432009626e-05,
+      "loss": 1.0843,
+      "step": 7247
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.752200305467801e-05,
+      "loss": 1.0223,
+      "step": 7248
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75212605818572e-05,
+      "loss": 0.9025,
+      "step": 7249
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7520518013556612e-05,
+      "loss": 1.0677,
+      "step": 7250
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751977534978568e-05,
+      "loss": 1.0181,
+      "step": 7251
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7519032590553828e-05,
+      "loss": 0.9832,
+      "step": 7252
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.751828973587048e-05,
+      "loss": 0.9734,
+      "step": 7253
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7517546785745074e-05,
+      "loss": 1.0442,
+      "step": 7254
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516803740187042e-05,
+      "loss": 1.0327,
+      "step": 7255
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7516060599205812e-05,
+      "loss": 1.0013,
+      "step": 7256
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7515317362810823e-05,
+      "loss": 1.1678,
+      "step": 7257
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.75145740310115e-05,
+      "loss": 0.9096,
+      "step": 7258
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513830603817293e-05,
+      "loss": 1.1089,
+      "step": 7259
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7513087081237634e-05,
+      "loss": 1.0537,
+      "step": 7260
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7512343463281953e-05,
+      "loss": 1.1139,
+      "step": 7261
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7511599749959703e-05,
+      "loss": 0.9849,
+      "step": 7262
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510855941280317e-05,
+      "loss": 1.1144,
+      "step": 7263
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7510112037253244e-05,
+      "loss": 0.9566,
+      "step": 7264
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750936803788792e-05,
+      "loss": 1.2435,
+      "step": 7265
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7508623943193796e-05,
+      "loss": 1.0019,
+      "step": 7266
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7507879753180312e-05,
+      "loss": 1.0884,
+      "step": 7267
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750713546785692e-05,
+      "loss": 1.0306,
+      "step": 7268
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750639108723307e-05,
+      "loss": 1.1519,
+      "step": 7269
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7505646611318206e-05,
+      "loss": 0.9499,
+      "step": 7270
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504902040121784e-05,
+      "loss": 1.1059,
+      "step": 7271
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7504157373653257e-05,
+      "loss": 0.9808,
+      "step": 7272
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.750341261192207e-05,
+      "loss": 1.0939,
+      "step": 7273
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7502667754937685e-05,
+      "loss": 1.1253,
+      "step": 7274
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7501922802709552e-05,
+      "loss": 1.1326,
+      "step": 7275
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7501177755247134e-05,
+      "loss": 1.1456,
+      "step": 7276
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7500432612559886e-05,
+      "loss": 1.0398,
+      "step": 7277
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7499687374657268e-05,
+      "loss": 1.0456,
+      "step": 7278
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749894204154874e-05,
+      "loss": 1.1345,
+      "step": 7279
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7498196613243763e-05,
+      "loss": 1.0128,
+      "step": 7280
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74974510897518e-05,
+      "loss": 1.0952,
+      "step": 7281
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749670547108232e-05,
+      "loss": 1.0228,
+      "step": 7282
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749595975724478e-05,
+      "loss": 1.1768,
+      "step": 7283
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749521394824865e-05,
+      "loss": 0.9754,
+      "step": 7284
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7494468044103402e-05,
+      "loss": 1.0769,
+      "step": 7285
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74937220448185e-05,
+      "loss": 1.067,
+      "step": 7286
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.749297595040342e-05,
+      "loss": 1.1023,
+      "step": 7287
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7492229760867627e-05,
+      "loss": 1.0042,
+      "step": 7288
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7491483476220595e-05,
+      "loss": 1.1329,
+      "step": 7289
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.74907370964718e-05,
+      "loss": 1.0743,
+      "step": 7290
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489990621630713e-05,
+      "loss": 1.1643,
+      "step": 7291
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7489244051706817e-05,
+      "loss": 1.0425,
+      "step": 7292
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7488497386709584e-05,
+      "loss": 1.0539,
+      "step": 7293
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487750626648495e-05,
+      "loss": 1.016,
+      "step": 7294
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7487003771533027e-05,
+      "loss": 1.1091,
+      "step": 7295
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7486256821372665e-05,
+      "loss": 1.0879,
+      "step": 7296
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7485509776176888e-05,
+      "loss": 1.1657,
+      "step": 7297
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748476263595518e-05,
+      "loss": 1.117,
+      "step": 7298
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748401540071703e-05,
+      "loss": 0.9933,
+      "step": 7299
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.748326807047192e-05,
+      "loss": 1.0421,
+      "step": 7300
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7482520645229332e-05,
+      "loss": 0.9689,
+      "step": 7301
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481773124998765e-05,
+      "loss": 1.0852,
+      "step": 7302
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7481025509789702e-05,
+      "loss": 1.2204,
+      "step": 7303
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7480277799611635e-05,
+      "loss": 1.0106,
+      "step": 7304
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7479529994474055e-05,
+      "loss": 1.0833,
+      "step": 7305
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478782094386457e-05,
+      "loss": 1.1094,
+      "step": 7306
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7478034099358338e-05,
+      "loss": 0.9669,
+      "step": 7307
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7477286009399184e-05,
+      "loss": 1.0763,
+      "step": 7308
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7476537824518502e-05,
+      "loss": 1.0835,
+      "step": 7309
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475789544725783e-05,
+      "loss": 0.9287,
+      "step": 7310
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7475041170030528e-05,
+      "loss": 1.0565,
+      "step": 7311
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7474292700442242e-05,
+      "loss": 1.1866,
+      "step": 7312
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7473544135970423e-05,
+      "loss": 1.0661,
+      "step": 7313
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7472795476624568e-05,
+      "loss": 0.9076,
+      "step": 7314
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747204672241419e-05,
+      "loss": 1.1277,
+      "step": 7315
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.747129787334879e-05,
+      "loss": 0.9912,
+      "step": 7316
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7470548929437875e-05,
+      "loss": 1.0642,
+      "step": 7317
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7469799890690954e-05,
+      "loss": 0.9795,
+      "step": 7318
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746905075711753e-05,
+      "loss": 1.1259,
+      "step": 7319
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746830152872712e-05,
+      "loss": 1.0337,
+      "step": 7320
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7467552205529235e-05,
+      "loss": 1.0469,
+      "step": 7321
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7466802787533385e-05,
+      "loss": 0.956,
+      "step": 7322
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746605327474908e-05,
+      "loss": 1.0603,
+      "step": 7323
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746530366718584e-05,
+      "loss": 0.9813,
+      "step": 7324
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7464553964853177e-05,
+      "loss": 1.0562,
+      "step": 7325
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463804167760613e-05,
+      "loss": 1.0038,
+      "step": 7326
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7463054275917662e-05,
+      "loss": 1.0026,
+      "step": 7327
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.746230428933385e-05,
+      "loss": 1.0702,
+      "step": 7328
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7461554208018688e-05,
+      "loss": 1.0513,
+      "step": 7329
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 1.7460804031981706e-05,
+      "loss": 1.1006,
+      "step": 7330
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7460053761232424e-05,
+      "loss": 1.1696,
+      "step": 7331
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745930339578037e-05,
+      "loss": 1.1712,
+      "step": 7332
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745855293563506e-05,
+      "loss": 1.0383,
+      "step": 7333
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7457802380806036e-05,
+      "loss": 0.9796,
+      "step": 7334
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745705173130281e-05,
+      "loss": 1.0257,
+      "step": 7335
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7456300987134925e-05,
+      "loss": 0.9721,
+      "step": 7336
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74555501483119e-05,
+      "loss": 1.1279,
+      "step": 7337
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454799214843278e-05,
+      "loss": 1.1628,
+      "step": 7338
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7454048186738578e-05,
+      "loss": 0.9552,
+      "step": 7339
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745329706400735e-05,
+      "loss": 0.9487,
+      "step": 7340
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7452545846659113e-05,
+      "loss": 0.9525,
+      "step": 7341
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7451794534703417e-05,
+      "loss": 1.0503,
+      "step": 7342
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.745104312814979e-05,
+      "loss": 1.0709,
+      "step": 7343
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7450291627007777e-05,
+      "loss": 0.9922,
+      "step": 7344
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7449540031286916e-05,
+      "loss": 1.0827,
+      "step": 7345
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744878834099675e-05,
+      "loss": 1.1063,
+      "step": 7346
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744803655614682e-05,
+      "loss": 1.0909,
+      "step": 7347
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7447284676746664e-05,
+      "loss": 1.1336,
+      "step": 7348
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7446532702805832e-05,
+      "loss": 1.0302,
+      "step": 7349
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445780634333877e-05,
+      "loss": 1.0161,
+      "step": 7350
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7445028471340336e-05,
+      "loss": 1.1997,
+      "step": 7351
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7444276213834758e-05,
+      "loss": 1.114,
+      "step": 7352
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.74435238618267e-05,
+      "loss": 1.1159,
+      "step": 7353
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7442771415325705e-05,
+      "loss": 1.0677,
+      "step": 7354
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744201887434133e-05,
+      "loss": 1.1232,
+      "step": 7355
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7441266238883127e-05,
+      "loss": 1.1435,
+      "step": 7356
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.744051350896065e-05,
+      "loss": 1.0366,
+      "step": 7357
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439760684583452e-05,
+      "loss": 1.1536,
+      "step": 7358
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7439007765761098e-05,
+      "loss": 0.9878,
+      "step": 7359
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7438254752503138e-05,
+      "loss": 1.1207,
+      "step": 7360
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.743750164481913e-05,
+      "loss": 1.1815,
+      "step": 7361
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7436748442718643e-05,
+      "loss": 1.1078,
+      "step": 7362
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7435995146211233e-05,
+      "loss": 0.9738,
+      "step": 7363
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7435241755306466e-05,
+      "loss": 1.1834,
+      "step": 7364
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7434488270013898e-05,
+      "loss": 1.0491,
+      "step": 7365
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7433734690343102e-05,
+      "loss": 1.1188,
+      "step": 7366
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432981016303645e-05,
+      "loss": 0.9669,
+      "step": 7367
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7432227247905093e-05,
+      "loss": 0.9524,
+      "step": 7368
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7431473385157008e-05,
+      "loss": 1.1299,
+      "step": 7369
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7430719428068972e-05,
+      "loss": 1.0199,
+      "step": 7370
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742996537665055e-05,
+      "loss": 1.0087,
+      "step": 7371
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742921123091131e-05,
+      "loss": 1.0182,
+      "step": 7372
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7428456990860833e-05,
+      "loss": 1.0109,
+      "step": 7373
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7427702656508693e-05,
+      "loss": 1.1009,
+      "step": 7374
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742694822786446e-05,
+      "loss": 1.1018,
+      "step": 7375
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7426193704937716e-05,
+      "loss": 1.0282,
+      "step": 7376
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742543908773804e-05,
+      "loss": 1.094,
+      "step": 7377
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742468437627501e-05,
+      "loss": 1.0806,
+      "step": 7378
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423929570558205e-05,
+      "loss": 1.0982,
+      "step": 7379
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7423174670597213e-05,
+      "loss": 1.0714,
+      "step": 7380
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.742241967640161e-05,
+      "loss": 1.0344,
+      "step": 7381
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7421664587980983e-05,
+      "loss": 1.092,
+      "step": 7382
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420909405344922e-05,
+      "loss": 1.19,
+      "step": 7383
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7420154128503006e-05,
+      "loss": 1.0671,
+      "step": 7384
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741939875746483e-05,
+      "loss": 1.04,
+      "step": 7385
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741864329223998e-05,
+      "loss": 1.2619,
+      "step": 7386
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7417887732838045e-05,
+      "loss": 1.0887,
+      "step": 7387
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741713207926862e-05,
+      "loss": 1.1293,
+      "step": 7388
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7416376331541292e-05,
+      "loss": 0.978,
+      "step": 7389
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7415620489665662e-05,
+      "loss": 1.0666,
+      "step": 7390
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414864553651322e-05,
+      "loss": 0.9898,
+      "step": 7391
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7414108523507868e-05,
+      "loss": 1.0161,
+      "step": 7392
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7413352399244897e-05,
+      "loss": 1.085,
+      "step": 7393
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7412596180872006e-05,
+      "loss": 1.1141,
+      "step": 7394
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7411839868398803e-05,
+      "loss": 1.1566,
+      "step": 7395
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741108346183488e-05,
+      "loss": 1.0617,
+      "step": 7396
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.741032696118984e-05,
+      "loss": 1.0509,
+      "step": 7397
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7409570366473295e-05,
+      "loss": 1.0969,
+      "step": 7398
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408813677694843e-05,
+      "loss": 1.0917,
+      "step": 7399
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7408056894864088e-05,
+      "loss": 1.0557,
+      "step": 7400
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7407300017990643e-05,
+      "loss": 1.1315,
+      "step": 7401
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7406543047084113e-05,
+      "loss": 0.9398,
+      "step": 7402
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405785982154107e-05,
+      "loss": 1.0728,
+      "step": 7403
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7405028823210237e-05,
+      "loss": 0.9902,
+      "step": 7404
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7404271570262112e-05,
+      "loss": 1.0914,
+      "step": 7405
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7403514223319354e-05,
+      "loss": 1.0448,
+      "step": 7406
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7402756782391565e-05,
+      "loss": 0.9034,
+      "step": 7407
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7401999247488368e-05,
+      "loss": 1.1715,
+      "step": 7408
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7401241618619377e-05,
+      "loss": 1.0943,
+      "step": 7409
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7400483895794212e-05,
+      "loss": 1.2454,
+      "step": 7410
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739972607902249e-05,
+      "loss": 1.0146,
+      "step": 7411
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739896816831383e-05,
+      "loss": 1.2489,
+      "step": 7412
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7398210163677858e-05,
+      "loss": 1.1383,
+      "step": 7413
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7397452065124195e-05,
+      "loss": 1.1204,
+      "step": 7414
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739669387266246e-05,
+      "loss": 0.9853,
+      "step": 7415
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739593558630228e-05,
+      "loss": 1.0294,
+      "step": 7416
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739517720605329e-05,
+      "loss": 0.9835,
+      "step": 7417
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7394418731925107e-05,
+      "loss": 1.1758,
+      "step": 7418
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739366016392736e-05,
+      "loss": 1.0957,
+      "step": 7419
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7392901502069686e-05,
+      "loss": 1.1086,
+      "step": 7420
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.739214274636171e-05,
+      "loss": 1.059,
+      "step": 7421
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7391383896813064e-05,
+      "loss": 1.0468,
+      "step": 7422
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7390624953433385e-05,
+      "loss": 1.094,
+      "step": 7423
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7389865916232302e-05,
+      "loss": 1.1896,
+      "step": 7424
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738910678521946e-05,
+      "loss": 1.0103,
+      "step": 7425
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7388347560404485e-05,
+      "loss": 1.0501,
+      "step": 7426
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7387588241797023e-05,
+      "loss": 1.0099,
+      "step": 7427
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738682882940671e-05,
+      "loss": 1.0185,
+      "step": 7428
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7386069323243187e-05,
+      "loss": 1.149,
+      "step": 7429
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7385309723316097e-05,
+      "loss": 1.1055,
+      "step": 7430
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738455002963508e-05,
+      "loss": 1.0351,
+      "step": 7431
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738379024220978e-05,
+      "loss": 1.072,
+      "step": 7432
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7383030361049845e-05,
+      "loss": 1.1481,
+      "step": 7433
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7382270386164922e-05,
+      "loss": 0.9332,
+      "step": 7434
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7381510317564653e-05,
+      "loss": 1.0781,
+      "step": 7435
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.738075015525869e-05,
+      "loss": 1.0687,
+      "step": 7436
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737998989925669e-05,
+      "loss": 1.0033,
+      "step": 7437
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7379229549568292e-05,
+      "loss": 1.1445,
+      "step": 7438
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7378469106203158e-05,
+      "loss": 1.173,
+      "step": 7439
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7377708569170938e-05,
+      "loss": 0.9509,
+      "step": 7440
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376947938481282e-05,
+      "loss": 1.0675,
+      "step": 7441
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7376187214143853e-05,
+      "loss": 0.9908,
+      "step": 7442
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7375426396168308e-05,
+      "loss": 0.9901,
+      "step": 7443
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73746654845643e-05,
+      "loss": 0.9668,
+      "step": 7444
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.737390447934149e-05,
+      "loss": 1.093,
+      "step": 7445
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7373143380509544e-05,
+      "loss": 1.0024,
+      "step": 7446
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7372382188078117e-05,
+      "loss": 1.0653,
+      "step": 7447
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7371620902056875e-05,
+      "loss": 1.1253,
+      "step": 7448
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370859522455483e-05,
+      "loss": 1.1305,
+      "step": 7449
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7370098049283605e-05,
+      "loss": 1.1431,
+      "step": 7450
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736933648255091e-05,
+      "loss": 1.0157,
+      "step": 7451
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7368574822267063e-05,
+      "loss": 0.9864,
+      "step": 7452
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367813068441738e-05,
+      "loss": 1.1357,
+      "step": 7453
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7367051221084596e-05,
+      "loss": 1.0758,
+      "step": 7454
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7366289280205316e-05,
+      "loss": 1.1466,
+      "step": 7455
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7365527245813568e-05,
+      "loss": 1.0988,
+      "step": 7456
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736476511791903e-05,
+      "loss": 1.1131,
+      "step": 7457
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7364002896531368e-05,
+      "loss": 1.0422,
+      "step": 7458
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7363240581660265e-05,
+      "loss": 1.1005,
+      "step": 7459
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7362478173315394e-05,
+      "loss": 1.1103,
+      "step": 7460
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736171567150644e-05,
+      "loss": 0.8796,
+      "step": 7461
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7360953076243076e-05,
+      "loss": 1.201,
+      "step": 7462
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.736019038753499e-05,
+      "loss": 1.0004,
+      "step": 7463
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7359427605391858e-05,
+      "loss": 0.9893,
+      "step": 7464
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7358664729823362e-05,
+      "loss": 1.1083,
+      "step": 7465
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357901760839188e-05,
+      "loss": 0.9645,
+      "step": 7466
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7357138698449027e-05,
+      "loss": 1.0701,
+      "step": 7467
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735637554266256e-05,
+      "loss": 1.0778,
+      "step": 7468
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7355612293489476e-05,
+      "loss": 0.9554,
+      "step": 7469
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7354848950939468e-05,
+      "loss": 1.0804,
+      "step": 7470
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735408551502222e-05,
+      "loss": 0.9981,
+      "step": 7471
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735332198574743e-05,
+      "loss": 0.9573,
+      "step": 7472
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735255836312479e-05,
+      "loss": 1.1039,
+      "step": 7473
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7351794647163986e-05,
+      "loss": 1.0065,
+      "step": 7474
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.735103083787472e-05,
+      "loss": 1.1205,
+      "step": 7475
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7350266935266688e-05,
+      "loss": 1.0057,
+      "step": 7476
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7349502939349587e-05,
+      "loss": 1.1292,
+      "step": 7477
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7348738850133114e-05,
+      "loss": 0.9679,
+      "step": 7478
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734797466762697e-05,
+      "loss": 0.9172,
+      "step": 7479
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7347210391840853e-05,
+      "loss": 0.9884,
+      "step": 7480
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7346446022784474e-05,
+      "loss": 0.9951,
+      "step": 7481
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7345681560467526e-05,
+      "loss": 1.0361,
+      "step": 7482
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734491700489972e-05,
+      "loss": 1.0604,
+      "step": 7483
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734415235609076e-05,
+      "loss": 1.0172,
+      "step": 7484
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7343387614050354e-05,
+      "loss": 0.9067,
+      "step": 7485
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7342622778788207e-05,
+      "loss": 0.9305,
+      "step": 7486
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.734185785031403e-05,
+      "loss": 0.931,
+      "step": 7487
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7341092828637537e-05,
+      "loss": 0.9785,
+      "step": 7488
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7340327713768435e-05,
+      "loss": 1.1143,
+      "step": 7489
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7339562505716436e-05,
+      "loss": 0.9659,
+      "step": 7490
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733879720449126e-05,
+      "loss": 1.1085,
+      "step": 7491
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7338031810102616e-05,
+      "loss": 1.0782,
+      "step": 7492
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7337266322560224e-05,
+      "loss": 1.0317,
+      "step": 7493
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7336500741873803e-05,
+      "loss": 0.9744,
+      "step": 7494
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7335735068053066e-05,
+      "loss": 1.0657,
+      "step": 7495
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334969301107735e-05,
+      "loss": 1.1254,
+      "step": 7496
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7334203441047537e-05,
+      "loss": 1.0077,
+      "step": 7497
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733343748788219e-05,
+      "loss": 0.8985,
+      "step": 7498
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7332671441621415e-05,
+      "loss": 1.0264,
+      "step": 7499
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733190530227494e-05,
+      "loss": 1.1014,
+      "step": 7500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7331139069852487e-05,
+      "loss": 1.0412,
+      "step": 7501
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.733037274436379e-05,
+      "loss": 1.0535,
+      "step": 7502
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732960632581857e-05,
+      "loss": 1.1595,
+      "step": 7503
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328839814226564e-05,
+      "loss": 1.008,
+      "step": 7504
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7328073209597494e-05,
+      "loss": 1.0359,
+      "step": 7505
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73273065119411e-05,
+      "loss": 1.0144,
+      "step": 7506
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732653972126711e-05,
+      "loss": 1.0991,
+      "step": 7507
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325772837585256e-05,
+      "loss": 1.153,
+      "step": 7508
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7325005860905278e-05,
+      "loss": 0.987,
+      "step": 7509
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732423879123691e-05,
+      "loss": 1.0344,
+      "step": 7510
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732347162858989e-05,
+      "loss": 0.955,
+      "step": 7511
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732270437297396e-05,
+      "loss": 1.2335,
+      "step": 7512
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.732193702439886e-05,
+      "loss": 1.0829,
+      "step": 7513
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7321169582874323e-05,
+      "loss": 1.0242,
+      "step": 7514
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73204020484101e-05,
+      "loss": 1.0702,
+      "step": 7515
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7319634421015933e-05,
+      "loss": 1.1534,
+      "step": 7516
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318866700701563e-05,
+      "loss": 1.0668,
+      "step": 7517
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7318098887476738e-05,
+      "loss": 1.0596,
+      "step": 7518
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7317330981351207e-05,
+      "loss": 1.0223,
+      "step": 7519
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731656298233472e-05,
+      "loss": 1.0509,
+      "step": 7520
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731579489043702e-05,
+      "loss": 1.0071,
+      "step": 7521
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7315026705667862e-05,
+      "loss": 1.115,
+      "step": 7522
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7314258428036997e-05,
+      "loss": 1.0028,
+      "step": 7523
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7313490057554173e-05,
+      "loss": 1.105,
+      "step": 7524
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7312721594229156e-05,
+      "loss": 1.0635,
+      "step": 7525
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731195303807169e-05,
+      "loss": 1.0642,
+      "step": 7526
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.731118438909154e-05,
+      "loss": 1.0603,
+      "step": 7527
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7310415647298455e-05,
+      "loss": 0.9779,
+      "step": 7528
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.73096468127022e-05,
+      "loss": 1.0102,
+      "step": 7529
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308877885312535e-05,
+      "loss": 1.1128,
+      "step": 7530
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7308108865139218e-05,
+      "loss": 0.9842,
+      "step": 7531
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730733975219202e-05,
+      "loss": 1.166,
+      "step": 7532
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7306570546480686e-05,
+      "loss": 1.0358,
+      "step": 7533
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305801248015003e-05,
+      "loss": 1.0056,
+      "step": 7534
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7305031856804725e-05,
+      "loss": 1.0016,
+      "step": 7535
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7304262372859616e-05,
+      "loss": 1.1372,
+      "step": 7536
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7303492796189455e-05,
+      "loss": 1.1231,
+      "step": 7537
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7302723126804003e-05,
+      "loss": 0.9768,
+      "step": 7538
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7301953364713037e-05,
+      "loss": 1.0365,
+      "step": 7539
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730118350992632e-05,
+      "loss": 1.2003,
+      "step": 7540
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.730041356245363e-05,
+      "loss": 1.0823,
+      "step": 7541
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7299643522304748e-05,
+      "loss": 1.0444,
+      "step": 7542
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729887338948944e-05,
+      "loss": 1.1125,
+      "step": 7543
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7298103164017482e-05,
+      "loss": 0.9484,
+      "step": 7544
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729733284589866e-05,
+      "loss": 1.1138,
+      "step": 7545
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7296562435142745e-05,
+      "loss": 1.0884,
+      "step": 7546
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7295791931759522e-05,
+      "loss": 0.9449,
+      "step": 7547
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729502133575877e-05,
+      "loss": 0.9977,
+      "step": 7548
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7294250647150273e-05,
+      "loss": 1.0232,
+      "step": 7549
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729347986594381e-05,
+      "loss": 1.0271,
+      "step": 7550
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7292708992149173e-05,
+      "loss": 1.1385,
+      "step": 7551
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7291938025776145e-05,
+      "loss": 1.141,
+      "step": 7552
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.729116696683451e-05,
+      "loss": 1.1191,
+      "step": 7553
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7290395815334063e-05,
+      "loss": 1.0171,
+      "step": 7554
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7289624571284583e-05,
+      "loss": 1.0733,
+      "step": 7555
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288853234695873e-05,
+      "loss": 1.0929,
+      "step": 7556
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7288081805577718e-05,
+      "loss": 1.0396,
+      "step": 7557
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7287310283939908e-05,
+      "loss": 1.0712,
+      "step": 7558
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7286538669792247e-05,
+      "loss": 0.9966,
+      "step": 7559
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728576696314452e-05,
+      "loss": 1.0821,
+      "step": 7560
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728499516400653e-05,
+      "loss": 1.075,
+      "step": 7561
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728422327238807e-05,
+      "loss": 0.9996,
+      "step": 7562
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7283451288298943e-05,
+      "loss": 0.9201,
+      "step": 7563
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.728267921174895e-05,
+      "loss": 1.0602,
+      "step": 7564
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281907042747886e-05,
+      "loss": 0.9977,
+      "step": 7565
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7281134781305558e-05,
+      "loss": 1.1268,
+      "step": 7566
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7280362427431772e-05,
+      "loss": 1.0761,
+      "step": 7567
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7279589981136328e-05,
+      "loss": 1.2006,
+      "step": 7568
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278817442429033e-05,
+      "loss": 1.1668,
+      "step": 7569
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7278044811319697e-05,
+      "loss": 0.9779,
+      "step": 7570
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7277272087818128e-05,
+      "loss": 0.9235,
+      "step": 7571
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7276499271934128e-05,
+      "loss": 1.1074,
+      "step": 7572
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7275726363677515e-05,
+      "loss": 0.9602,
+      "step": 7573
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72749533630581e-05,
+      "loss": 1.0735,
+      "step": 7574
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7274180270085695e-05,
+      "loss": 1.0151,
+      "step": 7575
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727340708477011e-05,
+      "loss": 1.0615,
+      "step": 7576
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727263380712117e-05,
+      "loss": 1.0322,
+      "step": 7577
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7271860437148683e-05,
+      "loss": 0.9574,
+      "step": 7578
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727108697486247e-05,
+      "loss": 1.052,
+      "step": 7579
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.727031342027235e-05,
+      "loss": 1.1365,
+      "step": 7580
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726953977338814e-05,
+      "loss": 1.0074,
+      "step": 7581
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7268766034219667e-05,
+      "loss": 0.9926,
+      "step": 7582
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7267992202776746e-05,
+      "loss": 1.1469,
+      "step": 7583
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726721827906921e-05,
+      "loss": 1.0581,
+      "step": 7584
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7266444263106873e-05,
+      "loss": 1.1488,
+      "step": 7585
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726567015489957e-05,
+      "loss": 0.9993,
+      "step": 7586
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726489595445712e-05,
+      "loss": 1.104,
+      "step": 7587
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7264121661789356e-05,
+      "loss": 1.0236,
+      "step": 7588
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7263347276906112e-05,
+      "loss": 1.0253,
+      "step": 7589
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7262572799817206e-05,
+      "loss": 0.9561,
+      "step": 7590
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726179823053248e-05,
+      "loss": 0.9999,
+      "step": 7591
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7261023569061765e-05,
+      "loss": 1.0605,
+      "step": 7592
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.726024881541489e-05,
+      "loss": 1.1355,
+      "step": 7593
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7259473969601697e-05,
+      "loss": 1.0491,
+      "step": 7594
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7258699031632017e-05,
+      "loss": 1.029,
+      "step": 7595
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257924001515693e-05,
+      "loss": 1.0254,
+      "step": 7596
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7257148879262557e-05,
+      "loss": 1.0371,
+      "step": 7597
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7256373664882455e-05,
+      "loss": 1.0483,
+      "step": 7598
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7255598358385226e-05,
+      "loss": 0.9846,
+      "step": 7599
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725482295978071e-05,
+      "loss": 1.0803,
+      "step": 7600
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7254047469078755e-05,
+      "loss": 0.9831,
+      "step": 7601
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.72532718862892e-05,
+      "loss": 1.006,
+      "step": 7602
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7252496211421897e-05,
+      "loss": 1.0563,
+      "step": 7603
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.725172044448669e-05,
+      "loss": 1.1199,
+      "step": 7604
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250944585493424e-05,
+      "loss": 1.1361,
+      "step": 7605
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7250168634451952e-05,
+      "loss": 1.0343,
+      "step": 7606
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724939259137212e-05,
+      "loss": 1.044,
+      "step": 7607
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724861645626379e-05,
+      "loss": 0.9671,
+      "step": 7608
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7247840229136802e-05,
+      "loss": 0.9842,
+      "step": 7609
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7247063910001016e-05,
+      "loss": 1.1634,
+      "step": 7610
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724628749886629e-05,
+      "loss": 1.0692,
+      "step": 7611
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7245510995742477e-05,
+      "loss": 1.0822,
+      "step": 7612
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7244734400639432e-05,
+      "loss": 1.0146,
+      "step": 7613
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7243957713567018e-05,
+      "loss": 1.0141,
+      "step": 7614
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.724318093453509e-05,
+      "loss": 1.0517,
+      "step": 7615
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7242404063553513e-05,
+      "loss": 1.1109,
+      "step": 7616
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 1.7241627100632147e-05,
+      "loss": 1.0799,
+      "step": 7617
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7240850045780856e-05,
+      "loss": 0.9356,
+      "step": 7618
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.724007289900951e-05,
+      "loss": 1.1169,
+      "step": 7619
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7239295660327965e-05,
+      "loss": 0.9688,
+      "step": 7620
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723851832974609e-05,
+      "loss": 0.9394,
+      "step": 7621
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723774090727376e-05,
+      "loss": 1.0666,
+      "step": 7622
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7236963392920837e-05,
+      "loss": 1.0645,
+      "step": 7623
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723618578669719e-05,
+      "loss": 0.9901,
+      "step": 7624
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7235408088612697e-05,
+      "loss": 0.9269,
+      "step": 7625
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7234630298677225e-05,
+      "loss": 1.1396,
+      "step": 7626
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7233852416900655e-05,
+      "loss": 0.9903,
+      "step": 7627
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723307444329285e-05,
+      "loss": 1.016,
+      "step": 7628
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.72322963778637e-05,
+      "loss": 0.9395,
+      "step": 7629
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7231518220623072e-05,
+      "loss": 1.0355,
+      "step": 7630
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.723073997158085e-05,
+      "loss": 1.0195,
+      "step": 7631
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722996163074691e-05,
+      "loss": 1.0117,
+      "step": 7632
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722918319813113e-05,
+      "loss": 1.0315,
+      "step": 7633
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7228404673743403e-05,
+      "loss": 0.9756,
+      "step": 7634
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7227626057593604e-05,
+      "loss": 0.953,
+      "step": 7635
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7226847349691614e-05,
+      "loss": 1.1133,
+      "step": 7636
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.722606855004733e-05,
+      "loss": 1.0263,
+      "step": 7637
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7225289658670627e-05,
+      "loss": 1.0257,
+      "step": 7638
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7224510675571396e-05,
+      "loss": 1.0854,
+      "step": 7639
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7223731600759528e-05,
+      "loss": 1.0775,
+      "step": 7640
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222952434244915e-05,
+      "loss": 1.0269,
+      "step": 7641
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7222173176037443e-05,
+      "loss": 1.03,
+      "step": 7642
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7221393826147007e-05,
+      "loss": 1.0619,
+      "step": 7643
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7220614384583503e-05,
+      "loss": 1.229,
+      "step": 7644
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721983485135682e-05,
+      "loss": 1.0165,
+      "step": 7645
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721905522647686e-05,
+      "loss": 1.0913,
+      "step": 7646
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7218275509953515e-05,
+      "loss": 0.9805,
+      "step": 7647
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7217495701796686e-05,
+      "loss": 1.0971,
+      "step": 7648
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7216715802016272e-05,
+      "loss": 1.0144,
+      "step": 7649
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7215935810622177e-05,
+      "loss": 1.0418,
+      "step": 7650
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7215155727624295e-05,
+      "loss": 1.0819,
+      "step": 7651
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.721437555303253e-05,
+      "loss": 1.1032,
+      "step": 7652
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7213595286856795e-05,
+      "loss": 1.0641,
+      "step": 7653
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212814929106985e-05,
+      "loss": 1.1301,
+      "step": 7654
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7212034479793014e-05,
+      "loss": 1.0691,
+      "step": 7655
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7211253938924784e-05,
+      "loss": 1.0272,
+      "step": 7656
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7210473306512208e-05,
+      "loss": 1.0862,
+      "step": 7657
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7209692582565192e-05,
+      "loss": 1.0169,
+      "step": 7658
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720891176709365e-05,
+      "loss": 1.0064,
+      "step": 7659
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720813086010749e-05,
+      "loss": 1.1547,
+      "step": 7660
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720734986161663e-05,
+      "loss": 1.0368,
+      "step": 7661
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7206568771630984e-05,
+      "loss": 1.0129,
+      "step": 7662
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7205787590160467e-05,
+      "loss": 1.0754,
+      "step": 7663
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7205006317214993e-05,
+      "loss": 1.0537,
+      "step": 7664
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720422495280448e-05,
+      "loss": 1.1795,
+      "step": 7665
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7203443496938854e-05,
+      "loss": 1.0244,
+      "step": 7666
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7202661949628028e-05,
+      "loss": 1.1222,
+      "step": 7667
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201880310881926e-05,
+      "loss": 1.1862,
+      "step": 7668
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7201098580710472e-05,
+      "loss": 0.9492,
+      "step": 7669
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.720031675912359e-05,
+      "loss": 0.9964,
+      "step": 7670
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71995348461312e-05,
+      "loss": 1.0983,
+      "step": 7671
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7198752841743234e-05,
+      "loss": 1.0775,
+      "step": 7672
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197970745969616e-05,
+      "loss": 0.9447,
+      "step": 7673
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7197188558820276e-05,
+      "loss": 1.1318,
+      "step": 7674
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7196406280305144e-05,
+      "loss": 1.1167,
+      "step": 7675
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7195623910434147e-05,
+      "loss": 1.0737,
+      "step": 7676
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.719484144921722e-05,
+      "loss": 0.9893,
+      "step": 7677
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7194058896664298e-05,
+      "loss": 1.1576,
+      "step": 7678
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7193276252785312e-05,
+      "loss": 1.0514,
+      "step": 7679
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7192493517590195e-05,
+      "loss": 0.9338,
+      "step": 7680
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7191710691088894e-05,
+      "loss": 1.0264,
+      "step": 7681
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190927773291332e-05,
+      "loss": 1.1518,
+      "step": 7682
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7190144764207457e-05,
+      "loss": 1.0829,
+      "step": 7683
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718936166384721e-05,
+      "loss": 1.0377,
+      "step": 7684
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718857847222053e-05,
+      "loss": 1.0295,
+      "step": 7685
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187795189337357e-05,
+      "loss": 1.1043,
+      "step": 7686
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7187011815207636e-05,
+      "loss": 1.0353,
+      "step": 7687
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7186228349841312e-05,
+      "loss": 1.1224,
+      "step": 7688
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7185444793248334e-05,
+      "loss": 1.1434,
+      "step": 7689
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7184661145438645e-05,
+      "loss": 1.1384,
+      "step": 7690
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718387740642219e-05,
+      "loss": 0.9824,
+      "step": 7691
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7183093576208927e-05,
+      "loss": 1.0384,
+      "step": 7692
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71823096548088e-05,
+      "loss": 1.0159,
+      "step": 7693
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.718152564223176e-05,
+      "loss": 1.0608,
+      "step": 7694
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7180741538487768e-05,
+      "loss": 1.1276,
+      "step": 7695
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717995734358677e-05,
+      "loss": 1.0805,
+      "step": 7696
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717917305753872e-05,
+      "loss": 1.0602,
+      "step": 7697
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717838868035358e-05,
+      "loss": 0.9471,
+      "step": 7698
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7177604212041302e-05,
+      "loss": 0.8775,
+      "step": 7699
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176819652611852e-05,
+      "loss": 1.1062,
+      "step": 7700
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7176035002075182e-05,
+      "loss": 1.0942,
+      "step": 7701
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7175250260441257e-05,
+      "loss": 0.9381,
+      "step": 7702
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7174465427720038e-05,
+      "loss": 1.1605,
+      "step": 7703
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7173680503921486e-05,
+      "loss": 1.2261,
+      "step": 7704
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717289548905557e-05,
+      "loss": 0.9231,
+      "step": 7705
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7172110383132252e-05,
+      "loss": 1.1653,
+      "step": 7706
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7171325186161502e-05,
+      "loss": 0.9937,
+      "step": 7707
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.717053989815328e-05,
+      "loss": 1.0103,
+      "step": 7708
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7169754519117562e-05,
+      "loss": 1.1584,
+      "step": 7709
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716896904906432e-05,
+      "loss": 1.1371,
+      "step": 7710
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7168183488003517e-05,
+      "loss": 1.0873,
+      "step": 7711
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7167397835945133e-05,
+      "loss": 1.2073,
+      "step": 7712
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7166612092899144e-05,
+      "loss": 1.0326,
+      "step": 7713
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716582625887551e-05,
+      "loss": 1.1556,
+      "step": 7714
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716504033388422e-05,
+      "loss": 0.986,
+      "step": 7715
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716425431793525e-05,
+      "loss": 1.1003,
+      "step": 7716
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7163468211038572e-05,
+      "loss": 1.0896,
+      "step": 7717
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.716268201320417e-05,
+      "loss": 1.0964,
+      "step": 7718
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7161895724442027e-05,
+      "loss": 0.998,
+      "step": 7719
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7161109344762117e-05,
+      "loss": 0.8967,
+      "step": 7720
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7160322874174427e-05,
+      "loss": 1.0436,
+      "step": 7721
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7159536312688944e-05,
+      "loss": 1.0716,
+      "step": 7722
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7158749660315647e-05,
+      "loss": 1.244,
+      "step": 7723
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7157962917064527e-05,
+      "loss": 1.0625,
+      "step": 7724
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715717608294557e-05,
+      "loss": 1.117,
+      "step": 7725
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7156389157968762e-05,
+      "loss": 1.032,
+      "step": 7726
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7155602142144093e-05,
+      "loss": 0.9893,
+      "step": 7727
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715481503548156e-05,
+      "loss": 1.0335,
+      "step": 7728
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.715402783799115e-05,
+      "loss": 1.0761,
+      "step": 7729
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7153240549682858e-05,
+      "loss": 1.1019,
+      "step": 7730
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7152453170566676e-05,
+      "loss": 1.0592,
+      "step": 7731
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7151665700652604e-05,
+      "loss": 1.0496,
+      "step": 7732
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150878139950634e-05,
+      "loss": 1.055,
+      "step": 7733
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7150090488470765e-05,
+      "loss": 1.0913,
+      "step": 7734
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7149302746222997e-05,
+      "loss": 1.0274,
+      "step": 7735
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7148514913217325e-05,
+      "loss": 1.1683,
+      "step": 7736
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7147726989463763e-05,
+      "loss": 1.0475,
+      "step": 7737
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71469389749723e-05,
+      "loss": 1.0875,
+      "step": 7738
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7146150869752948e-05,
+      "loss": 1.1114,
+      "step": 7739
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7145362673815705e-05,
+      "loss": 0.9719,
+      "step": 7740
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7144574387170582e-05,
+      "loss": 1.0244,
+      "step": 7741
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7143786009827586e-05,
+      "loss": 1.1356,
+      "step": 7742
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714299754179672e-05,
+      "loss": 1.0253,
+      "step": 7743
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7142208983088e-05,
+      "loss": 1.1765,
+      "step": 7744
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.714142033371143e-05,
+      "loss": 1.0254,
+      "step": 7745
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7140631593677032e-05,
+      "loss": 0.9945,
+      "step": 7746
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139842762994807e-05,
+      "loss": 1.021,
+      "step": 7747
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7139053841674774e-05,
+      "loss": 1.1523,
+      "step": 7748
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7138264829726946e-05,
+      "loss": 1.0046,
+      "step": 7749
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7137475727161344e-05,
+      "loss": 1.0096,
+      "step": 7750
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713668653398798e-05,
+      "loss": 1.1066,
+      "step": 7751
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135897250216878e-05,
+      "loss": 1.0705,
+      "step": 7752
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7135107875858055e-05,
+      "loss": 1.0852,
+      "step": 7753
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.713431841092153e-05,
+      "loss": 0.8819,
+      "step": 7754
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7133528855417324e-05,
+      "loss": 1.0338,
+      "step": 7755
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7132739209355468e-05,
+      "loss": 1.0477,
+      "step": 7756
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131949472745976e-05,
+      "loss": 1.1388,
+      "step": 7757
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7131159645598877e-05,
+      "loss": 1.0738,
+      "step": 7758
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7130369727924204e-05,
+      "loss": 1.0808,
+      "step": 7759
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7129579719731978e-05,
+      "loss": 1.0774,
+      "step": 7760
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712878962103223e-05,
+      "loss": 1.0652,
+      "step": 7761
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7127999431834988e-05,
+      "loss": 0.9941,
+      "step": 7762
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7127209152150284e-05,
+      "loss": 0.9667,
+      "step": 7763
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712641878198815e-05,
+      "loss": 1.0513,
+      "step": 7764
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7125628321358626e-05,
+      "loss": 1.0514,
+      "step": 7765
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124837770271735e-05,
+      "loss": 1.1258,
+      "step": 7766
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7124047128737525e-05,
+      "loss": 0.9734,
+      "step": 7767
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7123256396766023e-05,
+      "loss": 0.979,
+      "step": 7768
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712246557436727e-05,
+      "loss": 1.1195,
+      "step": 7769
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.712167466155131e-05,
+      "loss": 1.1257,
+      "step": 7770
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120883658328177e-05,
+      "loss": 0.9287,
+      "step": 7771
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7120092564707915e-05,
+      "loss": 1.2268,
+      "step": 7772
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7119301380700567e-05,
+      "loss": 1.0271,
+      "step": 7773
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7118510106316177e-05,
+      "loss": 1.0945,
+      "step": 7774
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7117718741564785e-05,
+      "loss": 1.0557,
+      "step": 7775
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116927286456445e-05,
+      "loss": 0.917,
+      "step": 7776
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7116135741001203e-05,
+      "loss": 0.9525,
+      "step": 7777
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71153441052091e-05,
+      "loss": 1.1355,
+      "step": 7778
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7114552379090198e-05,
+      "loss": 0.9895,
+      "step": 7779
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7113760562654533e-05,
+      "loss": 1.1467,
+      "step": 7780
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7112968655912166e-05,
+      "loss": 1.139,
+      "step": 7781
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711217665887315e-05,
+      "loss": 1.0331,
+      "step": 7782
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711138457154754e-05,
+      "loss": 0.9608,
+      "step": 7783
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.711059239394538e-05,
+      "loss": 0.9961,
+      "step": 7784
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109800126076744e-05,
+      "loss": 0.8983,
+      "step": 7785
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7109007767951675e-05,
+      "loss": 1.1536,
+      "step": 7786
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7108215319580236e-05,
+      "loss": 0.999,
+      "step": 7787
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710742278097249e-05,
+      "loss": 1.0485,
+      "step": 7788
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.71066301521385e-05,
+      "loss": 0.974,
+      "step": 7789
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710583743308832e-05,
+      "loss": 1.0881,
+      "step": 7790
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710504462383202e-05,
+      "loss": 1.1679,
+      "step": 7791
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710425172437966e-05,
+      "loss": 0.9805,
+      "step": 7792
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7103458734741307e-05,
+      "loss": 1.1536,
+      "step": 7793
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7102665654927033e-05,
+      "loss": 1.0192,
+      "step": 7794
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7101872484946894e-05,
+      "loss": 1.1484,
+      "step": 7795
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710107922481097e-05,
+      "loss": 1.0457,
+      "step": 7796
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.710028587452933e-05,
+      "loss": 0.9213,
+      "step": 7797
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709949243411204e-05,
+      "loss": 1.1139,
+      "step": 7798
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709869890356917e-05,
+      "loss": 1.11,
+      "step": 7799
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7097905282910808e-05,
+      "loss": 0.9257,
+      "step": 7800
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7097111572147015e-05,
+      "loss": 1.0881,
+      "step": 7801
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709631777128787e-05,
+      "loss": 1.0175,
+      "step": 7802
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7095523880343452e-05,
+      "loss": 1.0022,
+      "step": 7803
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7094729899323842e-05,
+      "loss": 0.9337,
+      "step": 7804
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.709393582823911e-05,
+      "loss": 1.0139,
+      "step": 7805
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7093141667099344e-05,
+      "loss": 1.0377,
+      "step": 7806
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7092347415914625e-05,
+      "loss": 1.0094,
+      "step": 7807
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7091553074695037e-05,
+      "loss": 1.1446,
+      "step": 7808
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7090758643450656e-05,
+      "loss": 0.9866,
+      "step": 7809
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089964122191576e-05,
+      "loss": 1.0445,
+      "step": 7810
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7089169510927883e-05,
+      "loss": 1.1074,
+      "step": 7811
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7088374809669656e-05,
+      "loss": 1.032,
+      "step": 7812
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708758001842699e-05,
+      "loss": 1.0854,
+      "step": 7813
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.708678513720997e-05,
+      "loss": 1.0584,
+      "step": 7814
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085990166028695e-05,
+      "loss": 1.0652,
+      "step": 7815
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7085195104893247e-05,
+      "loss": 1.0092,
+      "step": 7816
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7084399953813725e-05,
+      "loss": 1.1991,
+      "step": 7817
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7083604712800222e-05,
+      "loss": 1.1005,
+      "step": 7818
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082809381862835e-05,
+      "loss": 1.0542,
+      "step": 7819
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7082013961011658e-05,
+      "loss": 0.9869,
+      "step": 7820
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7081218450256786e-05,
+      "loss": 1.1369,
+      "step": 7821
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7080422849608327e-05,
+      "loss": 1.023,
+      "step": 7822
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707962715907637e-05,
+      "loss": 1.2086,
+      "step": 7823
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707883137867102e-05,
+      "loss": 1.0531,
+      "step": 7824
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7078035508402382e-05,
+      "loss": 1.0347,
+      "step": 7825
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707723954828056e-05,
+      "loss": 1.0094,
+      "step": 7826
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707644349831565e-05,
+      "loss": 1.0381,
+      "step": 7827
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7075647358517766e-05,
+      "loss": 1.1014,
+      "step": 7828
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7074851128897016e-05,
+      "loss": 0.9569,
+      "step": 7829
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7074054809463496e-05,
+      "loss": 1.1007,
+      "step": 7830
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707325840022733e-05,
+      "loss": 1.2016,
+      "step": 7831
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7072461901198617e-05,
+      "loss": 0.9566,
+      "step": 7832
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7071665312387474e-05,
+      "loss": 0.9588,
+      "step": 7833
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.707086863380401e-05,
+      "loss": 1.0416,
+      "step": 7834
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7070071865458338e-05,
+      "loss": 1.0867,
+      "step": 7835
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706927500736058e-05,
+      "loss": 1.0546,
+      "step": 7836
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7068478059520847e-05,
+      "loss": 0.8949,
+      "step": 7837
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706768102194925e-05,
+      "loss": 1.0074,
+      "step": 7838
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7066883894655915e-05,
+      "loss": 1.0114,
+      "step": 7839
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7066086677650963e-05,
+      "loss": 0.9192,
+      "step": 7840
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7065289370944507e-05,
+      "loss": 1.1368,
+      "step": 7841
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7064491974546674e-05,
+      "loss": 1.0279,
+      "step": 7842
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706369448846758e-05,
+      "loss": 1.0844,
+      "step": 7843
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706289691271736e-05,
+      "loss": 0.9987,
+      "step": 7844
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706209924730613e-05,
+      "loss": 1.1557,
+      "step": 7845
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7061301492244017e-05,
+      "loss": 1.0715,
+      "step": 7846
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.706050364754115e-05,
+      "loss": 1.0769,
+      "step": 7847
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7059705713207656e-05,
+      "loss": 1.0945,
+      "step": 7848
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7058907689253667e-05,
+      "loss": 0.9022,
+      "step": 7849
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7058109575689314e-05,
+      "loss": 1.0396,
+      "step": 7850
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7057311372524725e-05,
+      "loss": 1.1712,
+      "step": 7851
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7056513079770037e-05,
+      "loss": 1.0921,
+      "step": 7852
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705571469743538e-05,
+      "loss": 1.0671,
+      "step": 7853
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705491622553089e-05,
+      "loss": 0.9912,
+      "step": 7854
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.705411766406671e-05,
+      "loss": 1.05,
+      "step": 7855
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7053319013052964e-05,
+      "loss": 1.1314,
+      "step": 7856
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7052520272499804e-05,
+      "loss": 0.9737,
+      "step": 7857
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7051721442417362e-05,
+      "loss": 1.0145,
+      "step": 7858
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050922522815782e-05,
+      "loss": 1.0187,
+      "step": 7859
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7050123513705202e-05,
+      "loss": 1.1317,
+      "step": 7860
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7049324415095775e-05,
+      "loss": 0.9659,
+      "step": 7861
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7048525226997633e-05,
+      "loss": 1.1566,
+      "step": 7862
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704772594942093e-05,
+      "loss": 1.0738,
+      "step": 7863
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046926582375808e-05,
+      "loss": 1.0442,
+      "step": 7864
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7046127125872416e-05,
+      "loss": 1.082,
+      "step": 7865
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7045327579920906e-05,
+      "loss": 1.2115,
+      "step": 7866
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.704452794453142e-05,
+      "loss": 1.177,
+      "step": 7867
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7043728219714116e-05,
+      "loss": 1.2211,
+      "step": 7868
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042928405479147e-05,
+      "loss": 1.0255,
+      "step": 7869
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7042128501836657e-05,
+      "loss": 1.0574,
+      "step": 7870
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7041328508796814e-05,
+      "loss": 1.071,
+      "step": 7871
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7040528426369766e-05,
+      "loss": 1.0503,
+      "step": 7872
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7039728254565668e-05,
+      "loss": 1.0891,
+      "step": 7873
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703892799339468e-05,
+      "loss": 1.0287,
+      "step": 7874
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7038127642866964e-05,
+      "loss": 0.9716,
+      "step": 7875
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7037327202992677e-05,
+      "loss": 1.0925,
+      "step": 7876
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703652667378198e-05,
+      "loss": 0.886,
+      "step": 7877
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7035726055245035e-05,
+      "loss": 1.2035,
+      "step": 7878
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703492534739201e-05,
+      "loss": 1.0025,
+      "step": 7879
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7034124550233066e-05,
+      "loss": 0.9789,
+      "step": 7880
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.703332366377837e-05,
+      "loss": 0.9229,
+      "step": 7881
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7032522688038087e-05,
+      "loss": 1.0214,
+      "step": 7882
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7031721623022388e-05,
+      "loss": 1.1316,
+      "step": 7883
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030920468741438e-05,
+      "loss": 1.0651,
+      "step": 7884
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7030119225205414e-05,
+      "loss": 1.0797,
+      "step": 7885
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702931789242448e-05,
+      "loss": 1.0914,
+      "step": 7886
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7028516470408816e-05,
+      "loss": 1.1766,
+      "step": 7887
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702771495916859e-05,
+      "loss": 1.0653,
+      "step": 7888
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7026913358713978e-05,
+      "loss": 0.9255,
+      "step": 7889
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702611166905516e-05,
+      "loss": 0.9733,
+      "step": 7890
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702530989020231e-05,
+      "loss": 0.9797,
+      "step": 7891
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7024508022165603e-05,
+      "loss": 1.0579,
+      "step": 7892
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7023706064955227e-05,
+      "loss": 1.2342,
+      "step": 7893
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7022904018581355e-05,
+      "loss": 0.9882,
+      "step": 7894
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.702210188305417e-05,
+      "loss": 1.0155,
+      "step": 7895
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7021299658383856e-05,
+      "loss": 1.0151,
+      "step": 7896
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.70204973445806e-05,
+      "loss": 1.1081,
+      "step": 7897
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701969494165458e-05,
+      "loss": 0.9694,
+      "step": 7898
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7018892449615987e-05,
+      "loss": 1.0331,
+      "step": 7899
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701808986847501e-05,
+      "loss": 1.1717,
+      "step": 7900
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7017287198241836e-05,
+      "loss": 0.9928,
+      "step": 7901
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.701648443892665e-05,
+      "loss": 1.0984,
+      "step": 7902
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7015681590539653e-05,
+      "loss": 0.9925,
+      "step": 7903
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014878653091024e-05,
+      "loss": 1.1168,
+      "step": 7904
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 1.7014075626590964e-05,
+      "loss": 1.0805,
+      "step": 7905
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7013272511049667e-05,
+      "loss": 1.0272,
+      "step": 7906
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7012469306477326e-05,
+      "loss": 1.0177,
+      "step": 7907
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.701166601288414e-05,
+      "loss": 0.9457,
+      "step": 7908
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7010862630280302e-05,
+      "loss": 1.0075,
+      "step": 7909
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7010059158676016e-05,
+      "loss": 1.0888,
+      "step": 7910
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7009255598081477e-05,
+      "loss": 0.9839,
+      "step": 7911
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700845194850689e-05,
+      "loss": 1.0606,
+      "step": 7912
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7007648209962453e-05,
+      "loss": 1.1242,
+      "step": 7913
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006844382458373e-05,
+      "loss": 0.932,
+      "step": 7914
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7006040466004855e-05,
+      "loss": 1.1111,
+      "step": 7915
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7005236460612098e-05,
+      "loss": 0.9883,
+      "step": 7916
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7004432366290315e-05,
+      "loss": 1.0601,
+      "step": 7917
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.700362818304971e-05,
+      "loss": 1.047,
+      "step": 7918
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002823910900494e-05,
+      "loss": 0.9791,
+      "step": 7919
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7002019549852877e-05,
+      "loss": 1.0801,
+      "step": 7920
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7001215099917067e-05,
+      "loss": 1.0265,
+      "step": 7921
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.7000410561103283e-05,
+      "loss": 1.1628,
+      "step": 7922
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699960593342173e-05,
+      "loss": 1.1756,
+      "step": 7923
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6998801216882627e-05,
+      "loss": 1.0637,
+      "step": 7924
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699799641149619e-05,
+      "loss": 1.154,
+      "step": 7925
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6997191517272632e-05,
+      "loss": 1.0813,
+      "step": 7926
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6996386534222176e-05,
+      "loss": 0.9991,
+      "step": 7927
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6995581462355037e-05,
+      "loss": 1.0816,
+      "step": 7928
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.699477630168144e-05,
+      "loss": 1.176,
+      "step": 7929
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993971052211596e-05,
+      "loss": 0.9778,
+      "step": 7930
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6993165713955743e-05,
+      "loss": 0.9958,
+      "step": 7931
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6992360286924087e-05,
+      "loss": 1.1608,
+      "step": 7932
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6991554771126865e-05,
+      "loss": 0.9985,
+      "step": 7933
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6990749166574298e-05,
+      "loss": 1.1188,
+      "step": 7934
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989943473276616e-05,
+      "loss": 1.0719,
+      "step": 7935
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6989137691244043e-05,
+      "loss": 1.1247,
+      "step": 7936
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6988331820486812e-05,
+      "loss": 0.9632,
+      "step": 7937
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698752586101515e-05,
+      "loss": 1.0011,
+      "step": 7938
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6986719812839287e-05,
+      "loss": 0.9663,
+      "step": 7939
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.698591367596946e-05,
+      "loss": 1.1018,
+      "step": 7940
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6985107450415902e-05,
+      "loss": 1.0672,
+      "step": 7941
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6984301136188846e-05,
+      "loss": 0.9079,
+      "step": 7942
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6983494733298528e-05,
+      "loss": 1.2001,
+      "step": 7943
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6982688241755187e-05,
+      "loss": 1.0443,
+      "step": 7944
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981881661569058e-05,
+      "loss": 1.0131,
+      "step": 7945
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6981074992750378e-05,
+      "loss": 1.1376,
+      "step": 7946
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6980268235309395e-05,
+      "loss": 1.1266,
+      "step": 7947
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6979461389256348e-05,
+      "loss": 1.0795,
+      "step": 7948
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6978654454601476e-05,
+      "loss": 0.9396,
+      "step": 7949
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6977847431355025e-05,
+      "loss": 1.0335,
+      "step": 7950
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697704031952724e-05,
+      "loss": 1.0932,
+      "step": 7951
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6976233119128366e-05,
+      "loss": 1.0436,
+      "step": 7952
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6975425830168655e-05,
+      "loss": 1.1071,
+      "step": 7953
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697461845265835e-05,
+      "loss": 1.0433,
+      "step": 7954
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973810986607698e-05,
+      "loss": 1.0905,
+      "step": 7955
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6973003432026956e-05,
+      "loss": 1.0377,
+      "step": 7956
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.697219578892637e-05,
+      "loss": 1.1648,
+      "step": 7957
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6971388057316197e-05,
+      "loss": 1.0303,
+      "step": 7958
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6970580237206693e-05,
+      "loss": 0.9985,
+      "step": 7959
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6969772328608104e-05,
+      "loss": 1.0103,
+      "step": 7960
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6968964331530696e-05,
+      "loss": 1.0346,
+      "step": 7961
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6968156245984718e-05,
+      "loss": 1.0805,
+      "step": 7962
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6967348071980434e-05,
+      "loss": 1.0426,
+      "step": 7963
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69665398095281e-05,
+      "loss": 1.15,
+      "step": 7964
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696573145863798e-05,
+      "loss": 1.0117,
+      "step": 7965
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964923019320334e-05,
+      "loss": 1.0741,
+      "step": 7966
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6964114491585424e-05,
+      "loss": 1.1405,
+      "step": 7967
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696330587544351e-05,
+      "loss": 1.0096,
+      "step": 7968
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.696249717090487e-05,
+      "loss": 1.0084,
+      "step": 7969
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6961688377979756e-05,
+      "loss": 1.1054,
+      "step": 7970
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960879496678448e-05,
+      "loss": 1.2367,
+      "step": 7971
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6960070527011203e-05,
+      "loss": 1.0799,
+      "step": 7972
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6959261468988298e-05,
+      "loss": 1.0308,
+      "step": 7973
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695845232262e-05,
+      "loss": 1.0916,
+      "step": 7974
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6957643087916578e-05,
+      "loss": 1.18,
+      "step": 7975
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6956833764888312e-05,
+      "loss": 1.0862,
+      "step": 7976
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6956024353545474e-05,
+      "loss": 1.1433,
+      "step": 7977
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6955214853898337e-05,
+      "loss": 1.0296,
+      "step": 7978
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695440526595718e-05,
+      "loss": 0.9061,
+      "step": 7979
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695359558973228e-05,
+      "loss": 1.0823,
+      "step": 7980
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.695278582523391e-05,
+      "loss": 1.0779,
+      "step": 7981
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6951975972472355e-05,
+      "loss": 1.0319,
+      "step": 7982
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69511660314579e-05,
+      "loss": 0.9704,
+      "step": 7983
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6950356002200818e-05,
+      "loss": 1.0961,
+      "step": 7984
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.69495458847114e-05,
+      "loss": 1.2353,
+      "step": 7985
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6948735678999918e-05,
+      "loss": 1.071,
+      "step": 7986
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694792538507667e-05,
+      "loss": 0.898,
+      "step": 7987
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694711500295194e-05,
+      "loss": 1.0033,
+      "step": 7988
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694630453263601e-05,
+      "loss": 1.1901,
+      "step": 7989
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6945493974139172e-05,
+      "loss": 1.0626,
+      "step": 7990
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694468332747172e-05,
+      "loss": 1.1067,
+      "step": 7991
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694387259264394e-05,
+      "loss": 1.0719,
+      "step": 7992
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6943061769666117e-05,
+      "loss": 1.1233,
+      "step": 7993
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6942250858548562e-05,
+      "loss": 1.0545,
+      "step": 7994
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6941439859301552e-05,
+      "loss": 0.9801,
+      "step": 7995
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.694062877193539e-05,
+      "loss": 1.0222,
+      "step": 7996
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6939817596460374e-05,
+      "loss": 1.0851,
+      "step": 7997
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6939006332886796e-05,
+      "loss": 1.087,
+      "step": 7998
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6938194981224962e-05,
+      "loss": 1.1838,
+      "step": 7999
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6937383541485166e-05,
+      "loss": 1.0196,
+      "step": 8000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693657201367771e-05,
+      "loss": 1.0958,
+      "step": 8001
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6935760397812898e-05,
+      "loss": 0.9531,
+      "step": 8002
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934948693901033e-05,
+      "loss": 1.1255,
+      "step": 8003
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6934136901952415e-05,
+      "loss": 1.1443,
+      "step": 8004
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6933325021977355e-05,
+      "loss": 1.0131,
+      "step": 8005
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693251305398616e-05,
+      "loss": 1.0779,
+      "step": 8006
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6931700997989132e-05,
+      "loss": 1.1245,
+      "step": 8007
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6930888853996583e-05,
+      "loss": 1.1372,
+      "step": 8008
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.693007662201882e-05,
+      "loss": 1.06,
+      "step": 8009
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692926430206616e-05,
+      "loss": 1.1331,
+      "step": 8010
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6928451894148906e-05,
+      "loss": 1.0726,
+      "step": 8011
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6927639398277386e-05,
+      "loss": 0.9779,
+      "step": 8012
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6926826814461898e-05,
+      "loss": 1.0377,
+      "step": 8013
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692601414271277e-05,
+      "loss": 1.0986,
+      "step": 8014
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692520138304031e-05,
+      "loss": 1.0473,
+      "step": 8015
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692438853545484e-05,
+      "loss": 1.0122,
+      "step": 8016
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692357559996668e-05,
+      "loss": 1.0175,
+      "step": 8017
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692276257658614e-05,
+      "loss": 0.9681,
+      "step": 8018
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.692194946532356e-05,
+      "loss": 1.0867,
+      "step": 8019
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6921136266189243e-05,
+      "loss": 1.1788,
+      "step": 8020
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6920322979193528e-05,
+      "loss": 1.0318,
+      "step": 8021
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6919509604346723e-05,
+      "loss": 1.1042,
+      "step": 8022
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6918696141659166e-05,
+      "loss": 1.0681,
+      "step": 8023
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691788259114118e-05,
+      "loss": 1.0166,
+      "step": 8024
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691706895280309e-05,
+      "loss": 0.9816,
+      "step": 8025
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6916255226655232e-05,
+      "loss": 1.0917,
+      "step": 8026
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691544141270793e-05,
+      "loss": 1.0849,
+      "step": 8027
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6914627510971514e-05,
+      "loss": 0.9877,
+      "step": 8028
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691381352145632e-05,
+      "loss": 1.1846,
+      "step": 8029
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.691299944417268e-05,
+      "loss": 1.1148,
+      "step": 8030
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6912185279130932e-05,
+      "loss": 1.0655,
+      "step": 8031
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6911371026341403e-05,
+      "loss": 1.0361,
+      "step": 8032
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6910556685814436e-05,
+      "loss": 0.9897,
+      "step": 8033
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6909742257560364e-05,
+      "loss": 1.0571,
+      "step": 8034
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908927741589533e-05,
+      "loss": 1.0485,
+      "step": 8035
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6908113137912282e-05,
+      "loss": 0.9346,
+      "step": 8036
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6907298446538945e-05,
+      "loss": 0.9635,
+      "step": 8037
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690648366747987e-05,
+      "loss": 1.0556,
+      "step": 8038
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6905668800745398e-05,
+      "loss": 1.0812,
+      "step": 8039
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904853846345878e-05,
+      "loss": 0.9876,
+      "step": 8040
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6904038804291648e-05,
+      "loss": 1.1994,
+      "step": 8041
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690322367459306e-05,
+      "loss": 1.0899,
+      "step": 8042
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6902408457260458e-05,
+      "loss": 1.0371,
+      "step": 8043
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6901593152304193e-05,
+      "loss": 0.9614,
+      "step": 8044
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.690077775973462e-05,
+      "loss": 1.1196,
+      "step": 8045
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899962279562083e-05,
+      "loss": 0.9115,
+      "step": 8046
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6899146711796935e-05,
+      "loss": 1.0473,
+      "step": 8047
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6898331056449534e-05,
+      "loss": 1.054,
+      "step": 8048
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6897515313530228e-05,
+      "loss": 0.9601,
+      "step": 8049
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6896699483049382e-05,
+      "loss": 1.0504,
+      "step": 8050
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689588356501734e-05,
+      "loss": 0.9883,
+      "step": 8051
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689506755944447e-05,
+      "loss": 1.1466,
+      "step": 8052
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6894251466341126e-05,
+      "loss": 0.9383,
+      "step": 8053
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689343528571767e-05,
+      "loss": 1.1006,
+      "step": 8054
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6892619017584465e-05,
+      "loss": 1.0037,
+      "step": 8055
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689180266195187e-05,
+      "loss": 1.101,
+      "step": 8056
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6890986218830247e-05,
+      "loss": 1.0269,
+      "step": 8057
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.689016968822996e-05,
+      "loss": 1.0907,
+      "step": 8058
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688935307016138e-05,
+      "loss": 1.0227,
+      "step": 8059
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6888536364634874e-05,
+      "loss": 1.0325,
+      "step": 8060
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6887719571660804e-05,
+      "loss": 0.949,
+      "step": 8061
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688690269124954e-05,
+      "loss": 0.9952,
+      "step": 8062
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6886085723411456e-05,
+      "loss": 1.1344,
+      "step": 8063
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6885268668156917e-05,
+      "loss": 1.1217,
+      "step": 8064
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6884451525496303e-05,
+      "loss": 1.0744,
+      "step": 8065
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688363429543998e-05,
+      "loss": 1.0166,
+      "step": 8066
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688281697799833e-05,
+      "loss": 1.0754,
+      "step": 8067
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.688199957318172e-05,
+      "loss": 1.1964,
+      "step": 8068
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6881182081000532e-05,
+      "loss": 1.0746,
+      "step": 8069
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6880364501465145e-05,
+      "loss": 1.0959,
+      "step": 8070
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6879546834585933e-05,
+      "loss": 1.0447,
+      "step": 8071
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687872908037328e-05,
+      "loss": 1.0775,
+      "step": 8072
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877911238837565e-05,
+      "loss": 0.8999,
+      "step": 8073
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6877093309989172e-05,
+      "loss": 1.0312,
+      "step": 8074
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6876275293838485e-05,
+      "loss": 1.0553,
+      "step": 8075
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687545719039588e-05,
+      "loss": 0.9041,
+      "step": 8076
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6874638999671754e-05,
+      "loss": 1.0236,
+      "step": 8077
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687382072167649e-05,
+      "loss": 0.9435,
+      "step": 8078
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6873002356420474e-05,
+      "loss": 1.0408,
+      "step": 8079
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687218390391409e-05,
+      "loss": 1.0174,
+      "step": 8080
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.687136536416774e-05,
+      "loss": 1.068,
+      "step": 8081
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6870546737191805e-05,
+      "loss": 1.0729,
+      "step": 8082
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6869728022996682e-05,
+      "loss": 1.0173,
+      "step": 8083
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868909221592765e-05,
+      "loss": 1.0005,
+      "step": 8084
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6868090332990442e-05,
+      "loss": 1.0587,
+      "step": 8085
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6867271357200117e-05,
+      "loss": 1.048,
+      "step": 8086
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686645229423218e-05,
+      "loss": 1.0465,
+      "step": 8087
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6865633144097032e-05,
+      "loss": 1.095,
+      "step": 8088
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6864813906805068e-05,
+      "loss": 1.0834,
+      "step": 8089
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6863994582366695e-05,
+      "loss": 1.0216,
+      "step": 8090
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.686317517079231e-05,
+      "loss": 0.9924,
+      "step": 8091
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6862355672092312e-05,
+      "loss": 1.0268,
+      "step": 8092
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6861536086277112e-05,
+      "loss": 1.1665,
+      "step": 8093
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6860716413357106e-05,
+      "loss": 1.0146,
+      "step": 8094
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6859896653342703e-05,
+      "loss": 1.1439,
+      "step": 8095
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685907680624431e-05,
+      "loss": 0.8444,
+      "step": 8096
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6858256872072336e-05,
+      "loss": 1.0941,
+      "step": 8097
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6857436850837186e-05,
+      "loss": 1.0133,
+      "step": 8098
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6856616742549276e-05,
+      "loss": 1.1257,
+      "step": 8099
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685579654721901e-05,
+      "loss": 1.0376,
+      "step": 8100
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6854976264856802e-05,
+      "loss": 1.0806,
+      "step": 8101
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685415589547307e-05,
+      "loss": 1.054,
+      "step": 8102
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6853335439078223e-05,
+      "loss": 1.0162,
+      "step": 8103
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6852514895682678e-05,
+      "loss": 1.0167,
+      "step": 8104
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6851694265296852e-05,
+      "loss": 1.0072,
+      "step": 8105
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.685087354793116e-05,
+      "loss": 0.953,
+      "step": 8106
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6850052743596027e-05,
+      "loss": 1.0578,
+      "step": 8107
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6849231852301868e-05,
+      "loss": 1.1277,
+      "step": 8108
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6848410874059102e-05,
+      "loss": 1.0939,
+      "step": 8109
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6847589808878156e-05,
+      "loss": 1.0306,
+      "step": 8110
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6846768656769447e-05,
+      "loss": 1.0733,
+      "step": 8111
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845947417743408e-05,
+      "loss": 0.9683,
+      "step": 8112
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6845126091810457e-05,
+      "loss": 1.1899,
+      "step": 8113
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6844304678981023e-05,
+      "loss": 1.0107,
+      "step": 8114
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6843483179265534e-05,
+      "loss": 1.0162,
+      "step": 8115
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6842661592674415e-05,
+      "loss": 0.9562,
+      "step": 8116
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841839919218103e-05,
+      "loss": 1.0899,
+      "step": 8117
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6841018158907023e-05,
+      "loss": 1.0351,
+      "step": 8118
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.684019631175161e-05,
+      "loss": 1.0372,
+      "step": 8119
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6839374377762294e-05,
+      "loss": 1.2601,
+      "step": 8120
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683855235694951e-05,
+      "loss": 0.9908,
+      "step": 8121
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68377302493237e-05,
+      "loss": 1.0866,
+      "step": 8122
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683690805489529e-05,
+      "loss": 1.0078,
+      "step": 8123
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6836085773674722e-05,
+      "loss": 0.9895,
+      "step": 8124
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6835263405672434e-05,
+      "loss": 0.9297,
+      "step": 8125
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6834440950898867e-05,
+      "loss": 1.0225,
+      "step": 8126
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683361840936446e-05,
+      "loss": 0.9447,
+      "step": 8127
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6832795781079657e-05,
+      "loss": 1.1724,
+      "step": 8128
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6831973066054902e-05,
+      "loss": 1.0277,
+      "step": 8129
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.683115026430063e-05,
+      "loss": 1.0235,
+      "step": 8130
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68303273758273e-05,
+      "loss": 1.0701,
+      "step": 8131
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682950440064535e-05,
+      "loss": 1.012,
+      "step": 8132
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682868133876523e-05,
+      "loss": 1.0073,
+      "step": 8133
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827858190197383e-05,
+      "loss": 1.0046,
+      "step": 8134
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6827034954952266e-05,
+      "loss": 0.9984,
+      "step": 8135
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6826211633040327e-05,
+      "loss": 1.0489,
+      "step": 8136
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6825388224472015e-05,
+      "loss": 1.1163,
+      "step": 8137
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682456472925779e-05,
+      "loss": 1.0854,
+      "step": 8138
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6823741147408092e-05,
+      "loss": 1.0707,
+      "step": 8139
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822917478933395e-05,
+      "loss": 1.1729,
+      "step": 8140
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6822093723844138e-05,
+      "loss": 1.139,
+      "step": 8141
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.682126988215079e-05,
+      "loss": 1.0467,
+      "step": 8142
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6820445953863808e-05,
+      "loss": 1.1733,
+      "step": 8143
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6819621938993642e-05,
+      "loss": 1.1041,
+      "step": 8144
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6818797837550763e-05,
+      "loss": 1.1327,
+      "step": 8145
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6817973649545627e-05,
+      "loss": 1.0537,
+      "step": 8146
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68171493749887e-05,
+      "loss": 1.0215,
+      "step": 8147
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.681632501389044e-05,
+      "loss": 1.2143,
+      "step": 8148
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6815500566261324e-05,
+      "loss": 1.0022,
+      "step": 8149
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6814676032111806e-05,
+      "loss": 1.1398,
+      "step": 8150
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813851411452357e-05,
+      "loss": 0.9842,
+      "step": 8151
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6813026704293453e-05,
+      "loss": 1.1013,
+      "step": 8152
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6812201910645546e-05,
+      "loss": 1.0424,
+      "step": 8153
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6811377030519126e-05,
+      "loss": 0.8927,
+      "step": 8154
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6810552063924648e-05,
+      "loss": 1.0528,
+      "step": 8155
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6809727010872596e-05,
+      "loss": 1.0956,
+      "step": 8156
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808901871373442e-05,
+      "loss": 1.0553,
+      "step": 8157
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6808076645437656e-05,
+      "loss": 1.0679,
+      "step": 8158
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6807251333075717e-05,
+      "loss": 0.94,
+      "step": 8159
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.68064259342981e-05,
+      "loss": 1.0822,
+      "step": 8160
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6805600449115287e-05,
+      "loss": 0.9767,
+      "step": 8161
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6804774877537756e-05,
+      "loss": 0.9156,
+      "step": 8162
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803949219575982e-05,
+      "loss": 1.0581,
+      "step": 8163
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6803123475240454e-05,
+      "loss": 1.1821,
+      "step": 8164
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6802297644541653e-05,
+      "loss": 1.0727,
+      "step": 8165
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6801471727490058e-05,
+      "loss": 1.0249,
+      "step": 8166
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6800645724096157e-05,
+      "loss": 1.0591,
+      "step": 8167
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679981963437044e-05,
+      "loss": 1.1596,
+      "step": 8168
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798993458323382e-05,
+      "loss": 1.1491,
+      "step": 8169
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6798167195965485e-05,
+      "loss": 1.0893,
+      "step": 8170
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679734084730723e-05,
+      "loss": 1.0871,
+      "step": 8171
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6796514412359107e-05,
+      "loss": 1.0358,
+      "step": 8172
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679568789113161e-05,
+      "loss": 1.0504,
+      "step": 8173
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794861283635234e-05,
+      "loss": 1.0162,
+      "step": 8174
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6794034589880465e-05,
+      "loss": 0.916,
+      "step": 8175
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.67932078098778e-05,
+      "loss": 1.053,
+      "step": 8176
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679238094363774e-05,
+      "loss": 0.987,
+      "step": 8177
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679155399117078e-05,
+      "loss": 0.9485,
+      "step": 8178
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.679072695248741e-05,
+      "loss": 1.0394,
+      "step": 8179
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678989982759814e-05,
+      "loss": 1.0859,
+      "step": 8180
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678907261651346e-05,
+      "loss": 1.0775,
+      "step": 8181
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6788245319243885e-05,
+      "loss": 1.1754,
+      "step": 8182
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.67874179357999e-05,
+      "loss": 1.0375,
+      "step": 8183
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678659046619202e-05,
+      "loss": 1.151,
+      "step": 8184
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6785762910430745e-05,
+      "loss": 1.0149,
+      "step": 8185
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.678493526852658e-05,
+      "loss": 1.0685,
+      "step": 8186
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6784107540490036e-05,
+      "loss": 0.9625,
+      "step": 8187
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6783279726331618e-05,
+      "loss": 0.9432,
+      "step": 8188
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6782451826061838e-05,
+      "loss": 1.0442,
+      "step": 8189
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6781623839691196e-05,
+      "loss": 1.082,
+      "step": 8190
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6780795767230212e-05,
+      "loss": 0.9738,
+      "step": 8191
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 1.6779967608689397e-05,
+      "loss": 1.1205,
+      "step": 8192
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6779139364079267e-05,
+      "loss": 0.9914,
+      "step": 8193
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6778311033410325e-05,
+      "loss": 0.9953,
+      "step": 8194
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67774826166931e-05,
+      "loss": 1.1056,
+      "step": 8195
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6776654113938097e-05,
+      "loss": 1.06,
+      "step": 8196
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677582552515584e-05,
+      "loss": 0.8851,
+      "step": 8197
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6774996850356847e-05,
+      "loss": 1.0127,
+      "step": 8198
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.677416808955164e-05,
+      "loss": 1.2509,
+      "step": 8199
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6773339242750736e-05,
+      "loss": 0.9118,
+      "step": 8200
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6772510309964656e-05,
+      "loss": 0.9896,
+      "step": 8201
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6771681291203922e-05,
+      "loss": 0.9967,
+      "step": 8202
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770852186479065e-05,
+      "loss": 0.9743,
+      "step": 8203
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6770022995800607e-05,
+      "loss": 1.0865,
+      "step": 8204
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676919371917907e-05,
+      "loss": 0.9136,
+      "step": 8205
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676836435662499e-05,
+      "loss": 1.1524,
+      "step": 8206
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6767534908148883e-05,
+      "loss": 0.9872,
+      "step": 8207
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676670537376129e-05,
+      "loss": 1.1337,
+      "step": 8208
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765875753472737e-05,
+      "loss": 1.0413,
+      "step": 8209
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6765046047293755e-05,
+      "loss": 1.0092,
+      "step": 8210
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676421625523488e-05,
+      "loss": 1.0433,
+      "step": 8211
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6763386377306643e-05,
+      "loss": 1.2108,
+      "step": 8212
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.676255641351958e-05,
+      "loss": 0.9391,
+      "step": 8213
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6761726363884226e-05,
+      "loss": 1.098,
+      "step": 8214
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760896228411125e-05,
+      "loss": 1.13,
+      "step": 8215
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6760066007110804e-05,
+      "loss": 1.0488,
+      "step": 8216
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6759235699993807e-05,
+      "loss": 0.9615,
+      "step": 8217
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675840530707068e-05,
+      "loss": 1.0246,
+      "step": 8218
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6757574828351957e-05,
+      "loss": 1.1329,
+      "step": 8219
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6756744263848187e-05,
+      "loss": 1.1319,
+      "step": 8220
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755913613569907e-05,
+      "loss": 1.1786,
+      "step": 8221
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6755082877527664e-05,
+      "loss": 1.0789,
+      "step": 8222
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6754252055732005e-05,
+      "loss": 1.0093,
+      "step": 8223
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6753421148193478e-05,
+      "loss": 1.0845,
+      "step": 8224
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6752590154922633e-05,
+      "loss": 1.0883,
+      "step": 8225
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675175907593001e-05,
+      "loss": 1.0555,
+      "step": 8226
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6750927911226172e-05,
+      "loss": 1.0044,
+      "step": 8227
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.675009666082166e-05,
+      "loss": 1.015,
+      "step": 8228
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674926532472703e-05,
+      "loss": 1.0074,
+      "step": 8229
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674843390295284e-05,
+      "loss": 1.2731,
+      "step": 8230
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674760239550964e-05,
+      "loss": 0.8819,
+      "step": 8231
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674677080240798e-05,
+      "loss": 1.0348,
+      "step": 8232
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6745939123658426e-05,
+      "loss": 1.1234,
+      "step": 8233
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6745107359271538e-05,
+      "loss": 1.124,
+      "step": 8234
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674427550925786e-05,
+      "loss": 1.0374,
+      "step": 8235
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6743443573627968e-05,
+      "loss": 0.9484,
+      "step": 8236
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674261155239242e-05,
+      "loss": 1.1102,
+      "step": 8237
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.674177944556177e-05,
+      "loss": 0.9363,
+      "step": 8238
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6740947253146584e-05,
+      "loss": 1.1146,
+      "step": 8239
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6740114975157435e-05,
+      "loss": 1.0185,
+      "step": 8240
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673928261160488e-05,
+      "loss": 1.1293,
+      "step": 8241
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6738450162499485e-05,
+      "loss": 1.1375,
+      "step": 8242
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6737617627851826e-05,
+      "loss": 1.0376,
+      "step": 8243
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6736785007672463e-05,
+      "loss": 0.985,
+      "step": 8244
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6735952301971968e-05,
+      "loss": 0.9626,
+      "step": 8245
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673511951076091e-05,
+      "loss": 0.9234,
+      "step": 8246
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6734286634049875e-05,
+      "loss": 0.9488,
+      "step": 8247
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6733453671849417e-05,
+      "loss": 1.1667,
+      "step": 8248
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673262062417012e-05,
+      "loss": 1.0018,
+      "step": 8249
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673178749102256e-05,
+      "loss": 1.1642,
+      "step": 8250
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6730954272417306e-05,
+      "loss": 1.1687,
+      "step": 8251
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.673012096836494e-05,
+      "loss": 1.0701,
+      "step": 8252
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6729287578876042e-05,
+      "loss": 1.0402,
+      "step": 8253
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6728454103961194e-05,
+      "loss": 0.9409,
+      "step": 8254
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672762054363097e-05,
+      "loss": 1.1048,
+      "step": 8255
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6726786897895953e-05,
+      "loss": 1.0431,
+      "step": 8256
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725953166766734e-05,
+      "loss": 1.0575,
+      "step": 8257
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6725119350253882e-05,
+      "loss": 0.9996,
+      "step": 8258
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6724285448367998e-05,
+      "loss": 0.9734,
+      "step": 8259
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6723451461119654e-05,
+      "loss": 0.9848,
+      "step": 8260
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672261738851945e-05,
+      "loss": 0.9921,
+      "step": 8261
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6721783230577964e-05,
+      "loss": 1.1381,
+      "step": 8262
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672094898730579e-05,
+      "loss": 1.13,
+      "step": 8263
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.672011465871352e-05,
+      "loss": 1.0519,
+      "step": 8264
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671928024481174e-05,
+      "loss": 1.0366,
+      "step": 8265
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6718445745611047e-05,
+      "loss": 0.9136,
+      "step": 8266
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6717611161122032e-05,
+      "loss": 1.0438,
+      "step": 8267
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6716776491355295e-05,
+      "loss": 1.0207,
+      "step": 8268
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715941736321427e-05,
+      "loss": 0.9402,
+      "step": 8269
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6715106896031023e-05,
+      "loss": 1.0518,
+      "step": 8270
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671427197049469e-05,
+      "loss": 1.0723,
+      "step": 8271
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6713436959723017e-05,
+      "loss": 1.0121,
+      "step": 8272
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.671260186372661e-05,
+      "loss": 1.0415,
+      "step": 8273
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6711766682516067e-05,
+      "loss": 1.0593,
+      "step": 8274
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710931416101995e-05,
+      "loss": 0.909,
+      "step": 8275
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6710096064494993e-05,
+      "loss": 1.0389,
+      "step": 8276
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6709260627705665e-05,
+      "loss": 0.9968,
+      "step": 8277
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6708425105744622e-05,
+      "loss": 1.0552,
+      "step": 8278
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6707589498622464e-05,
+      "loss": 1.1146,
+      "step": 8279
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.67067538063498e-05,
+      "loss": 0.9536,
+      "step": 8280
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705918028937246e-05,
+      "loss": 0.9807,
+      "step": 8281
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6705082166395405e-05,
+      "loss": 1.1051,
+      "step": 8282
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670424621873489e-05,
+      "loss": 0.9633,
+      "step": 8283
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6703410185966308e-05,
+      "loss": 1.0203,
+      "step": 8284
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670257406810028e-05,
+      "loss": 1.0901,
+      "step": 8285
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.670173786514742e-05,
+      "loss": 1.0257,
+      "step": 8286
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700901577118337e-05,
+      "loss": 1.0881,
+      "step": 8287
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6700065204023647e-05,
+      "loss": 0.9515,
+      "step": 8288
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669922874587398e-05,
+      "loss": 0.9922,
+      "step": 8289
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6698392202679938e-05,
+      "loss": 0.8971,
+      "step": 8290
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6697555574452152e-05,
+      "loss": 1.0986,
+      "step": 8291
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669671886120124e-05,
+      "loss": 1.0419,
+      "step": 8292
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669588206293782e-05,
+      "loss": 1.2009,
+      "step": 8293
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669504517967252e-05,
+      "loss": 0.9815,
+      "step": 8294
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.669420821141596e-05,
+      "loss": 0.9796,
+      "step": 8295
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6693371158178767e-05,
+      "loss": 0.9303,
+      "step": 8296
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6692534019971566e-05,
+      "loss": 1.1342,
+      "step": 8297
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6691696796804985e-05,
+      "loss": 0.9554,
+      "step": 8298
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6690859488689653e-05,
+      "loss": 1.0749,
+      "step": 8299
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6690022095636198e-05,
+      "loss": 1.0444,
+      "step": 8300
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6689184617655252e-05,
+      "loss": 1.0368,
+      "step": 8301
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6688347054757444e-05,
+      "loss": 1.0267,
+      "step": 8302
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668750940695341e-05,
+      "loss": 1.0777,
+      "step": 8303
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668667167425378e-05,
+      "loss": 1.092,
+      "step": 8304
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.668583385666919e-05,
+      "loss": 1.2011,
+      "step": 8305
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6684995954210276e-05,
+      "loss": 1.0025,
+      "step": 8306
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6684157966887677e-05,
+      "loss": 1.0676,
+      "step": 8307
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6683319894712027e-05,
+      "loss": 1.1079,
+      "step": 8308
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6682481737693967e-05,
+      "loss": 1.1279,
+      "step": 8309
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6681643495844137e-05,
+      "loss": 0.9955,
+      "step": 8310
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6680805169173176e-05,
+      "loss": 1.1811,
+      "step": 8311
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6679966757691734e-05,
+      "loss": 0.9608,
+      "step": 8312
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667912826141045e-05,
+      "loss": 1.0987,
+      "step": 8313
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667828968033996e-05,
+      "loss": 1.1181,
+      "step": 8314
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667745101449092e-05,
+      "loss": 1.0238,
+      "step": 8315
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6676612263873974e-05,
+      "loss": 1.0744,
+      "step": 8316
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6675773428499772e-05,
+      "loss": 0.9795,
+      "step": 8317
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674934508378955e-05,
+      "loss": 1.1485,
+      "step": 8318
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6674095503522178e-05,
+      "loss": 1.034,
+      "step": 8319
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6673256413940097e-05,
+      "loss": 0.9908,
+      "step": 8320
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.667241723964335e-05,
+      "loss": 1.1064,
+      "step": 8321
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6671577980642605e-05,
+      "loss": 1.0252,
+      "step": 8322
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6670738636948505e-05,
+      "loss": 1.1512,
+      "step": 8323
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666989920857171e-05,
+      "loss": 1.0937,
+      "step": 8324
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666905969552288e-05,
+      "loss": 1.094,
+      "step": 8325
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6668220097812667e-05,
+      "loss": 1.0252,
+      "step": 8326
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6667380415451728e-05,
+      "loss": 1.159,
+      "step": 8327
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6666540648450728e-05,
+      "loss": 1.0821,
+      "step": 8328
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6665700796820325e-05,
+      "loss": 1.0858,
+      "step": 8329
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6664860860571174e-05,
+      "loss": 0.9942,
+      "step": 8330
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666402083971395e-05,
+      "loss": 0.9999,
+      "step": 8331
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6663180734259308e-05,
+      "loss": 1.1289,
+      "step": 8332
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6662340544217917e-05,
+      "loss": 1.0758,
+      "step": 8333
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6661500269600443e-05,
+      "loss": 1.0703,
+      "step": 8334
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.666065991041755e-05,
+      "loss": 1.0218,
+      "step": 8335
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665981946667991e-05,
+      "loss": 1.1495,
+      "step": 8336
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6658978938398184e-05,
+      "loss": 1.1451,
+      "step": 8337
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6658138325583052e-05,
+      "loss": 1.1678,
+      "step": 8338
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6657297628245183e-05,
+      "loss": 1.004,
+      "step": 8339
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6656456846395245e-05,
+      "loss": 1.2401,
+      "step": 8340
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665561598004391e-05,
+      "loss": 1.0116,
+      "step": 8341
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665477502920186e-05,
+      "loss": 1.1482,
+      "step": 8342
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665393399387977e-05,
+      "loss": 1.0541,
+      "step": 8343
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665309287408831e-05,
+      "loss": 1.1136,
+      "step": 8344
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6652251669838163e-05,
+      "loss": 1.0273,
+      "step": 8345
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6651410381140007e-05,
+      "loss": 1.1615,
+      "step": 8346
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.665056900800452e-05,
+      "loss": 1.0278,
+      "step": 8347
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6649727550442384e-05,
+      "loss": 1.047,
+      "step": 8348
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6648886008464282e-05,
+      "loss": 0.9298,
+      "step": 8349
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6648044382080893e-05,
+      "loss": 1.1124,
+      "step": 8350
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6647202671302905e-05,
+      "loss": 1.1487,
+      "step": 8351
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664636087614101e-05,
+      "loss": 1.0257,
+      "step": 8352
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664551899660588e-05,
+      "loss": 1.0536,
+      "step": 8353
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6644677032708208e-05,
+      "loss": 1.0361,
+      "step": 8354
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6643834984458687e-05,
+      "loss": 1.0367,
+      "step": 8355
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6642992851868002e-05,
+      "loss": 1.0113,
+      "step": 8356
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6642150634946842e-05,
+      "loss": 0.9815,
+      "step": 8357
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6641308333705907e-05,
+      "loss": 1.1108,
+      "step": 8358
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.664046594815588e-05,
+      "loss": 1.1309,
+      "step": 8359
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663962347830746e-05,
+      "loss": 1.0624,
+      "step": 8360
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663878092417134e-05,
+      "loss": 1.0922,
+      "step": 8361
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663793828575822e-05,
+      "loss": 1.0707,
+      "step": 8362
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663709556307879e-05,
+      "loss": 1.0473,
+      "step": 8363
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6636252756143755e-05,
+      "loss": 1.0519,
+      "step": 8364
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6635409864963806e-05,
+      "loss": 1.0632,
+      "step": 8365
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6634566889549653e-05,
+      "loss": 1.0381,
+      "step": 8366
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6633723829911988e-05,
+      "loss": 1.0981,
+      "step": 8367
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6632880686061522e-05,
+      "loss": 0.97,
+      "step": 8368
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.663203745800895e-05,
+      "loss": 1.0138,
+      "step": 8369
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6631194145764983e-05,
+      "loss": 1.0684,
+      "step": 8370
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6630350749340323e-05,
+      "loss": 1.1402,
+      "step": 8371
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6629507268745678e-05,
+      "loss": 1.1381,
+      "step": 8372
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6628663703991757e-05,
+      "loss": 0.9856,
+      "step": 8373
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6627820055089266e-05,
+      "loss": 1.0944,
+      "step": 8374
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662697632204892e-05,
+      "loss": 1.1406,
+      "step": 8375
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662613250488142e-05,
+      "loss": 1.0202,
+      "step": 8376
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6625288603597486e-05,
+      "loss": 1.0173,
+      "step": 8377
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662444461820783e-05,
+      "loss": 1.0584,
+      "step": 8378
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662360054872316e-05,
+      "loss": 1.1252,
+      "step": 8379
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6622756395154205e-05,
+      "loss": 1.0533,
+      "step": 8380
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6621912157511665e-05,
+      "loss": 0.99,
+      "step": 8381
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662106783580627e-05,
+      "loss": 0.9829,
+      "step": 8382
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.662022343004873e-05,
+      "loss": 0.9721,
+      "step": 8383
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661937894024977e-05,
+      "loss": 1.0356,
+      "step": 8384
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6618534366420104e-05,
+      "loss": 0.998,
+      "step": 8385
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661768970857046e-05,
+      "loss": 1.0182,
+      "step": 8386
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661684496671156e-05,
+      "loss": 0.9256,
+      "step": 8387
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6616000140854123e-05,
+      "loss": 0.9362,
+      "step": 8388
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6615155231008877e-05,
+      "loss": 1.0941,
+      "step": 8389
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661431023718655e-05,
+      "loss": 1.0111,
+      "step": 8390
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6613465159397866e-05,
+      "loss": 1.1659,
+      "step": 8391
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661261999765355e-05,
+      "loss": 1.029,
+      "step": 8392
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.661177475196434e-05,
+      "loss": 1.0755,
+      "step": 8393
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610929422340958e-05,
+      "loss": 1.0293,
+      "step": 8394
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6610084008794135e-05,
+      "loss": 1.0834,
+      "step": 8395
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6609238511334607e-05,
+      "loss": 1.0843,
+      "step": 8396
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6608392929973112e-05,
+      "loss": 0.9706,
+      "step": 8397
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6607547264720373e-05,
+      "loss": 0.9917,
+      "step": 8398
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6606701515587133e-05,
+      "loss": 0.9411,
+      "step": 8399
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6605855682584123e-05,
+      "loss": 1.1157,
+      "step": 8400
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.660500976572209e-05,
+      "loss": 1.1562,
+      "step": 8401
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6604163765011764e-05,
+      "loss": 0.9908,
+      "step": 8402
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6603317680463885e-05,
+      "loss": 1.0226,
+      "step": 8403
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6602471512089198e-05,
+      "loss": 1.2041,
+      "step": 8404
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6601625259898445e-05,
+      "loss": 1.1371,
+      "step": 8405
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6600778923902365e-05,
+      "loss": 1.0647,
+      "step": 8406
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599932504111704e-05,
+      "loss": 1.0065,
+      "step": 8407
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6599086000537205e-05,
+      "loss": 1.0337,
+      "step": 8408
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6598239413189617e-05,
+      "loss": 1.0803,
+      "step": 8409
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6597392742079686e-05,
+      "loss": 0.988,
+      "step": 8410
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6596545987218163e-05,
+      "loss": 1.0544,
+      "step": 8411
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6595699148615788e-05,
+      "loss": 1.0698,
+      "step": 8412
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6594852226283325e-05,
+      "loss": 1.1395,
+      "step": 8413
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659400522023151e-05,
+      "loss": 1.1291,
+      "step": 8414
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6593158130471115e-05,
+      "loss": 1.1996,
+      "step": 8415
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6592310957012872e-05,
+      "loss": 1.1178,
+      "step": 8416
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.659146369986755e-05,
+      "loss": 1.1323,
+      "step": 8417
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65906163590459e-05,
+      "loss": 1.0837,
+      "step": 8418
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6589768934558675e-05,
+      "loss": 1.0081,
+      "step": 8419
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658892142641664e-05,
+      "loss": 1.2152,
+      "step": 8420
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6588073834630553e-05,
+      "loss": 1.0873,
+      "step": 8421
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658722615921117e-05,
+      "loss": 1.0452,
+      "step": 8422
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658637840016925e-05,
+      "loss": 1.056,
+      "step": 8423
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6585530557515562e-05,
+      "loss": 1.1036,
+      "step": 8424
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6584682631260867e-05,
+      "loss": 1.1348,
+      "step": 8425
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6583834621415925e-05,
+      "loss": 1.0893,
+      "step": 8426
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65829865279915e-05,
+      "loss": 1.0202,
+      "step": 8427
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658213835099837e-05,
+      "loss": 0.9726,
+      "step": 8428
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658129009044729e-05,
+      "loss": 1.0928,
+      "step": 8429
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.658044174634903e-05,
+      "loss": 1.117,
+      "step": 8430
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6579593318714365e-05,
+      "loss": 1.1795,
+      "step": 8431
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6578744807554064e-05,
+      "loss": 0.9488,
+      "step": 8432
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6577896212878894e-05,
+      "loss": 1.07,
+      "step": 8433
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6577047534699634e-05,
+      "loss": 0.9925,
+      "step": 8434
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6576198773027052e-05,
+      "loss": 1.067,
+      "step": 8435
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6575349927871926e-05,
+      "loss": 0.9821,
+      "step": 8436
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6574500999245034e-05,
+      "loss": 1.046,
+      "step": 8437
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6573651987157146e-05,
+      "loss": 1.0754,
+      "step": 8438
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6572802891619044e-05,
+      "loss": 1.1065,
+      "step": 8439
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657195371264151e-05,
+      "loss": 1.0321,
+      "step": 8440
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6571104450235317e-05,
+      "loss": 1.1206,
+      "step": 8441
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.657025510441125e-05,
+      "loss": 1.0837,
+      "step": 8442
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6569405675180094e-05,
+      "loss": 1.1581,
+      "step": 8443
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6568556162552628e-05,
+      "loss": 1.1871,
+      "step": 8444
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6567706566539637e-05,
+      "loss": 1.1462,
+      "step": 8445
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6566856887151902e-05,
+      "loss": 1.1235,
+      "step": 8446
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656600712440022e-05,
+      "loss": 1.124,
+      "step": 8447
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6565157278295373e-05,
+      "loss": 0.9504,
+      "step": 8448
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6564307348848147e-05,
+      "loss": 1.1193,
+      "step": 8449
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6563457336069335e-05,
+      "loss": 1.0294,
+      "step": 8450
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6562607239969725e-05,
+      "loss": 1.0456,
+      "step": 8451
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656175706056011e-05,
+      "loss": 0.9971,
+      "step": 8452
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6560906797851283e-05,
+      "loss": 1.1483,
+      "step": 8453
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.656005645185404e-05,
+      "loss": 1.0009,
+      "step": 8454
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655920602257917e-05,
+      "loss": 1.1041,
+      "step": 8455
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655835551003747e-05,
+      "loss": 1.0432,
+      "step": 8456
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6557504914239743e-05,
+      "loss": 0.9628,
+      "step": 8457
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655665423519678e-05,
+      "loss": 0.9667,
+      "step": 8458
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655580347291939e-05,
+      "loss": 1.1401,
+      "step": 8459
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655495262741836e-05,
+      "loss": 1.0818,
+      "step": 8460
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6554101698704502e-05,
+      "loss": 1.0144,
+      "step": 8461
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.655325068678861e-05,
+      "loss": 1.1274,
+      "step": 8462
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6552399591681494e-05,
+      "loss": 1.1217,
+      "step": 8463
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6551548413393952e-05,
+      "loss": 1.0595,
+      "step": 8464
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6550697151936795e-05,
+      "loss": 1.0916,
+      "step": 8465
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654984580732083e-05,
+      "loss": 1.1564,
+      "step": 8466
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548994379556857e-05,
+      "loss": 1.0402,
+      "step": 8467
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6548142868655693e-05,
+      "loss": 1.1052,
+      "step": 8468
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6547291274628145e-05,
+      "loss": 0.9578,
+      "step": 8469
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6546439597485018e-05,
+      "loss": 1.0228,
+      "step": 8470
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6545587837237134e-05,
+      "loss": 1.1039,
+      "step": 8471
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6544735993895294e-05,
+      "loss": 1.083,
+      "step": 8472
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6543884067470324e-05,
+      "loss": 1.0874,
+      "step": 8473
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654303205797303e-05,
+      "loss": 0.985,
+      "step": 8474
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6542179965414233e-05,
+      "loss": 1.0816,
+      "step": 8475
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6541327789804746e-05,
+      "loss": 1.072,
+      "step": 8476
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.654047553115539e-05,
+      "loss": 1.0084,
+      "step": 8477
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.6539623189476984e-05,
+      "loss": 1.0381,
+      "step": 8478
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.653877076478035e-05,
+      "loss": 1.0972,
+      "step": 8479
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 1.65379182570763e-05,
+      "loss": 1.0321,
+      "step": 8480
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6537065666375665e-05,
+      "loss": 1.024,
+      "step": 8481
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653621299268927e-05,
+      "loss": 1.0962,
+      "step": 8482
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6535360236027935e-05,
+      "loss": 1.0554,
+      "step": 8483
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6534507396402483e-05,
+      "loss": 1.0037,
+      "step": 8484
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6533654473823747e-05,
+      "loss": 1.229,
+      "step": 8485
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6532801468302546e-05,
+      "loss": 0.9722,
+      "step": 8486
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6531948379849718e-05,
+      "loss": 1.0811,
+      "step": 8487
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.653109520847609e-05,
+      "loss": 1.0128,
+      "step": 8488
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6530241954192487e-05,
+      "loss": 0.9043,
+      "step": 8489
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6529388617009748e-05,
+      "loss": 1.0097,
+      "step": 8490
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.65285351969387e-05,
+      "loss": 1.0023,
+      "step": 8491
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6527681693990183e-05,
+      "loss": 1.0334,
+      "step": 8492
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6526828108175027e-05,
+      "loss": 1.1619,
+      "step": 8493
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525974439504068e-05,
+      "loss": 1.0272,
+      "step": 8494
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6525120687988146e-05,
+      "loss": 0.9262,
+      "step": 8495
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6524266853638097e-05,
+      "loss": 1.1616,
+      "step": 8496
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6523412936464764e-05,
+      "loss": 0.9757,
+      "step": 8497
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6522558936478978e-05,
+      "loss": 1.1334,
+      "step": 8498
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652170485369159e-05,
+      "loss": 1.0758,
+      "step": 8499
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.652085068811344e-05,
+      "loss": 1.0437,
+      "step": 8500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519996439755366e-05,
+      "loss": 1.0134,
+      "step": 8501
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6519142108628215e-05,
+      "loss": 0.9218,
+      "step": 8502
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651828769474284e-05,
+      "loss": 1.0573,
+      "step": 8503
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651743319811008e-05,
+      "loss": 0.997,
+      "step": 8504
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6516578618740777e-05,
+      "loss": 1.0051,
+      "step": 8505
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651572395664579e-05,
+      "loss": 0.9862,
+      "step": 8506
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6514869211835966e-05,
+      "loss": 1.0391,
+      "step": 8507
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6514014384322155e-05,
+      "loss": 0.9919,
+      "step": 8508
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6513159474115207e-05,
+      "loss": 1.1173,
+      "step": 8509
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6512304481225976e-05,
+      "loss": 1.1235,
+      "step": 8510
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6511449405665315e-05,
+      "loss": 1.0865,
+      "step": 8511
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.651059424744408e-05,
+      "loss": 1.1964,
+      "step": 8512
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650973900657313e-05,
+      "loss": 1.0453,
+      "step": 8513
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6508883683063314e-05,
+      "loss": 1.1357,
+      "step": 8514
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6508028276925498e-05,
+      "loss": 1.0507,
+      "step": 8515
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6507172788170537e-05,
+      "loss": 1.026,
+      "step": 8516
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6506317216809292e-05,
+      "loss": 1.2141,
+      "step": 8517
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6505461562852624e-05,
+      "loss": 1.1524,
+      "step": 8518
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6504605826311396e-05,
+      "loss": 1.1717,
+      "step": 8519
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650375000719647e-05,
+      "loss": 1.0752,
+      "step": 8520
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.650289410551871e-05,
+      "loss": 1.0365,
+      "step": 8521
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6502038121288985e-05,
+      "loss": 0.9728,
+      "step": 8522
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6501182054518155e-05,
+      "loss": 1.0602,
+      "step": 8523
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6500325905217094e-05,
+      "loss": 1.1958,
+      "step": 8524
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6499469673396667e-05,
+      "loss": 1.1039,
+      "step": 8525
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6498613359067743e-05,
+      "loss": 1.0165,
+      "step": 8526
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6497756962241195e-05,
+      "loss": 0.9073,
+      "step": 8527
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496900482927894e-05,
+      "loss": 0.9948,
+      "step": 8528
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6496043921138715e-05,
+      "loss": 0.9916,
+      "step": 8529
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6495187276884522e-05,
+      "loss": 1.0813,
+      "step": 8530
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.64943305501762e-05,
+      "loss": 1.0583,
+      "step": 8531
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6493473741024626e-05,
+      "loss": 1.1258,
+      "step": 8532
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6492616849440665e-05,
+      "loss": 1.0903,
+      "step": 8533
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649175987543521e-05,
+      "loss": 0.9085,
+      "step": 8534
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.649090281901913e-05,
+      "loss": 1.012,
+      "step": 8535
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6490045680203306e-05,
+      "loss": 1.1365,
+      "step": 8536
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6489188458998618e-05,
+      "loss": 1.1188,
+      "step": 8537
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6488331155415954e-05,
+      "loss": 1.0706,
+      "step": 8538
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6487473769466197e-05,
+      "loss": 0.961,
+      "step": 8539
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6486616301160223e-05,
+      "loss": 1.1612,
+      "step": 8540
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6485758750508924e-05,
+      "loss": 1.1786,
+      "step": 8541
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6484901117523184e-05,
+      "loss": 0.8793,
+      "step": 8542
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648404340221389e-05,
+      "loss": 0.9769,
+      "step": 8543
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6483185604591935e-05,
+      "loss": 1.052,
+      "step": 8544
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6482327724668206e-05,
+      "loss": 1.0446,
+      "step": 8545
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6481469762453587e-05,
+      "loss": 1.1788,
+      "step": 8546
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.648061171795898e-05,
+      "loss": 0.9231,
+      "step": 8547
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6479753591195267e-05,
+      "loss": 1.1287,
+      "step": 8548
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647889538217335e-05,
+      "loss": 1.1012,
+      "step": 8549
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6478037090904123e-05,
+      "loss": 1.0658,
+      "step": 8550
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6477178717398477e-05,
+      "loss": 0.9673,
+      "step": 8551
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647632026166731e-05,
+      "loss": 0.9248,
+      "step": 8552
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6475461723721523e-05,
+      "loss": 0.994,
+      "step": 8553
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6474603103572012e-05,
+      "loss": 0.9592,
+      "step": 8554
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6473744401229676e-05,
+      "loss": 1.1752,
+      "step": 8555
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647288561670542e-05,
+      "loss": 0.8835,
+      "step": 8556
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6472026750010143e-05,
+      "loss": 1.1315,
+      "step": 8557
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6471167801154748e-05,
+      "loss": 0.9175,
+      "step": 8558
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.647030877015014e-05,
+      "loss": 1.0108,
+      "step": 8559
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6469449657007227e-05,
+      "loss": 1.0528,
+      "step": 8560
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6468590461736907e-05,
+      "loss": 1.1021,
+      "step": 8561
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6467731184350097e-05,
+      "loss": 0.9987,
+      "step": 8562
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6466871824857696e-05,
+      "loss": 0.9702,
+      "step": 8563
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646601238327062e-05,
+      "loss": 1.1419,
+      "step": 8564
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6465152859599774e-05,
+      "loss": 0.9513,
+      "step": 8565
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6464293253856074e-05,
+      "loss": 1.1197,
+      "step": 8566
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6463433566050434e-05,
+      "loss": 1.0649,
+      "step": 8567
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646257379619376e-05,
+      "loss": 1.0644,
+      "step": 8568
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646171394429697e-05,
+      "loss": 0.9669,
+      "step": 8569
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.646085401037098e-05,
+      "loss": 1.0616,
+      "step": 8570
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6459993994426708e-05,
+      "loss": 1.142,
+      "step": 8571
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645913389647507e-05,
+      "loss": 1.0859,
+      "step": 8572
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6458273716526987e-05,
+      "loss": 1.0468,
+      "step": 8573
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6457413454593375e-05,
+      "loss": 1.067,
+      "step": 8574
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6456553110685155e-05,
+      "loss": 1.0892,
+      "step": 8575
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6455692684813253e-05,
+      "loss": 1.0932,
+      "step": 8576
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6454832176988586e-05,
+      "loss": 1.0302,
+      "step": 8577
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453971587222083e-05,
+      "loss": 1.0376,
+      "step": 8578
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6453110915524662e-05,
+      "loss": 1.0533,
+      "step": 8579
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645225016190726e-05,
+      "loss": 1.0928,
+      "step": 8580
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.645138932638079e-05,
+      "loss": 1.0294,
+      "step": 8581
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6450528408956195e-05,
+      "loss": 0.9266,
+      "step": 8582
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6449667409644394e-05,
+      "loss": 1.1641,
+      "step": 8583
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6448806328456318e-05,
+      "loss": 1.0016,
+      "step": 8584
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447945165402904e-05,
+      "loss": 0.9443,
+      "step": 8585
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6447083920495072e-05,
+      "loss": 0.9074,
+      "step": 8586
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6446222593743772e-05,
+      "loss": 1.093,
+      "step": 8587
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6445361185159922e-05,
+      "loss": 1.1092,
+      "step": 8588
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.644449969475447e-05,
+      "loss": 1.0594,
+      "step": 8589
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6443638122538346e-05,
+      "loss": 1.2772,
+      "step": 8590
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6442776468522484e-05,
+      "loss": 1.0565,
+      "step": 8591
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441914732717836e-05,
+      "loss": 1.0664,
+      "step": 8592
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6441052915135324e-05,
+      "loss": 1.0168,
+      "step": 8593
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6440191015785897e-05,
+      "loss": 1.1197,
+      "step": 8594
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6439329034680496e-05,
+      "loss": 0.9961,
+      "step": 8595
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6438466971830064e-05,
+      "loss": 1.0226,
+      "step": 8596
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643760482724555e-05,
+      "loss": 1.159,
+      "step": 8597
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6436742600937888e-05,
+      "loss": 1.0163,
+      "step": 8598
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643588029291803e-05,
+      "loss": 1.2276,
+      "step": 8599
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6435017903196918e-05,
+      "loss": 1.0888,
+      "step": 8600
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6434155431785507e-05,
+      "loss": 1.1036,
+      "step": 8601
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6433292878694736e-05,
+      "loss": 1.0722,
+      "step": 8602
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643243024393557e-05,
+      "loss": 1.0616,
+      "step": 8603
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6431567527518947e-05,
+      "loss": 1.1448,
+      "step": 8604
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.643070472945582e-05,
+      "loss": 1.0228,
+      "step": 8605
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6429841849757143e-05,
+      "loss": 0.9462,
+      "step": 8606
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642897888843387e-05,
+      "loss": 1.1077,
+      "step": 8607
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6428115845496962e-05,
+      "loss": 1.0655,
+      "step": 8608
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6427252720957368e-05,
+      "loss": 1.0296,
+      "step": 8609
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6426389514826045e-05,
+      "loss": 0.932,
+      "step": 8610
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642552622711396e-05,
+      "loss": 0.9747,
+      "step": 8611
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642466285783206e-05,
+      "loss": 1.0202,
+      "step": 8612
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6423799406991312e-05,
+      "loss": 1.1835,
+      "step": 8613
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422935874602678e-05,
+      "loss": 1.0599,
+      "step": 8614
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6422072260677114e-05,
+      "loss": 1.0086,
+      "step": 8615
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642120856522559e-05,
+      "loss": 0.991,
+      "step": 8616
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.642034478825907e-05,
+      "loss": 1.0183,
+      "step": 8617
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6419480929788513e-05,
+      "loss": 0.9784,
+      "step": 8618
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641861698982489e-05,
+      "loss": 0.9929,
+      "step": 8619
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6417752968379168e-05,
+      "loss": 0.9876,
+      "step": 8620
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6416888865462313e-05,
+      "loss": 1.0105,
+      "step": 8621
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6416024681085304e-05,
+      "loss": 0.9946,
+      "step": 8622
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6415160415259097e-05,
+      "loss": 1.059,
+      "step": 8623
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6414296067994678e-05,
+      "loss": 1.1078,
+      "step": 8624
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6413431639303005e-05,
+      "loss": 1.1326,
+      "step": 8625
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6412567129195065e-05,
+      "loss": 0.9743,
+      "step": 8626
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6411702537681824e-05,
+      "loss": 1.1866,
+      "step": 8627
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.641083786477426e-05,
+      "loss": 1.0667,
+      "step": 8628
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640997311048335e-05,
+      "loss": 1.0507,
+      "step": 8629
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640910827482007e-05,
+      "loss": 1.0637,
+      "step": 8630
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6408243357795405e-05,
+      "loss": 1.0952,
+      "step": 8631
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640737835942033e-05,
+      "loss": 1.0132,
+      "step": 8632
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6406513279705826e-05,
+      "loss": 1.0228,
+      "step": 8633
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640564811866287e-05,
+      "loss": 1.1125,
+      "step": 8634
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6404782876302455e-05,
+      "loss": 0.9473,
+      "step": 8635
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640391755263556e-05,
+      "loss": 1.1596,
+      "step": 8636
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640305214767317e-05,
+      "loss": 1.1172,
+      "step": 8637
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6402186661426272e-05,
+      "loss": 1.0033,
+      "step": 8638
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.640132109390585e-05,
+      "loss": 1.0161,
+      "step": 8639
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6400455445122895e-05,
+      "loss": 1.0749,
+      "step": 8640
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6399589715088397e-05,
+      "loss": 1.0683,
+      "step": 8641
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639872390381334e-05,
+      "loss": 1.0665,
+      "step": 8642
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6397858011308723e-05,
+      "loss": 0.961,
+      "step": 8643
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6396992037585536e-05,
+      "loss": 1.0333,
+      "step": 8644
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639612598265477e-05,
+      "loss": 1.0739,
+      "step": 8645
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6395259846527418e-05,
+      "loss": 1.0941,
+      "step": 8646
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639439362921448e-05,
+      "loss": 1.0516,
+      "step": 8647
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6393527330726947e-05,
+      "loss": 1.0075,
+      "step": 8648
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6392660951075824e-05,
+      "loss": 1.0459,
+      "step": 8649
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63917944902721e-05,
+      "loss": 1.1618,
+      "step": 8650
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.639092794832678e-05,
+      "loss": 1.1039,
+      "step": 8651
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6390061325250868e-05,
+      "loss": 0.9862,
+      "step": 8652
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6389194621055357e-05,
+      "loss": 1.0526,
+      "step": 8653
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6388327835751255e-05,
+      "loss": 1.0653,
+      "step": 8654
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6387460969349566e-05,
+      "loss": 1.1048,
+      "step": 8655
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6386594021861287e-05,
+      "loss": 0.9764,
+      "step": 8656
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6385726993297434e-05,
+      "loss": 0.9506,
+      "step": 8657
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6384859883669007e-05,
+      "loss": 1.0645,
+      "step": 8658
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383992692987015e-05,
+      "loss": 1.109,
+      "step": 8659
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6383125421262466e-05,
+      "loss": 0.9484,
+      "step": 8660
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6382258068506374e-05,
+      "loss": 1.0349,
+      "step": 8661
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6381390634729744e-05,
+      "loss": 1.077,
+      "step": 8662
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6380523119943595e-05,
+      "loss": 0.9268,
+      "step": 8663
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637965552415893e-05,
+      "loss": 1.2138,
+      "step": 8664
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637878784738677e-05,
+      "loss": 0.9223,
+      "step": 8665
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637792008963813e-05,
+      "loss": 1.0055,
+      "step": 8666
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6377052250924022e-05,
+      "loss": 0.9711,
+      "step": 8667
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6376184331255463e-05,
+      "loss": 0.9617,
+      "step": 8668
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6375316330643474e-05,
+      "loss": 0.867,
+      "step": 8669
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6374448249099073e-05,
+      "loss": 1.0578,
+      "step": 8670
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637358008663328e-05,
+      "loss": 1.035,
+      "step": 8671
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6372711843257115e-05,
+      "loss": 1.0086,
+      "step": 8672
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6371843518981602e-05,
+      "loss": 0.9735,
+      "step": 8673
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637097511381776e-05,
+      "loss": 1.1286,
+      "step": 8674
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.637010662777662e-05,
+      "loss": 1.1009,
+      "step": 8675
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63692380608692e-05,
+      "loss": 1.0285,
+      "step": 8676
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6368369413106526e-05,
+      "loss": 1.0282,
+      "step": 8677
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6367500684499632e-05,
+      "loss": 1.0207,
+      "step": 8678
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636663187505954e-05,
+      "loss": 1.0087,
+      "step": 8679
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6365762984797284e-05,
+      "loss": 0.9601,
+      "step": 8680
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364894013723892e-05,
+      "loss": 1.0809,
+      "step": 8681
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6364024961850398e-05,
+      "loss": 0.952,
+      "step": 8682
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6363155829187826e-05,
+      "loss": 0.8727,
+      "step": 8683
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636228661574722e-05,
+      "loss": 1.0369,
+      "step": 8684
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6361417321539604e-05,
+      "loss": 1.1086,
+      "step": 8685
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.636054794657602e-05,
+      "loss": 1.1281,
+      "step": 8686
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6359678490867507e-05,
+      "loss": 1.0653,
+      "step": 8687
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6358808954425095e-05,
+      "loss": 1.1504,
+      "step": 8688
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6357939337259827e-05,
+      "loss": 0.8628,
+      "step": 8689
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.635706963938274e-05,
+      "loss": 1.0562,
+      "step": 8690
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6356199860804878e-05,
+      "loss": 1.137,
+      "step": 8691
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6355330001537278e-05,
+      "loss": 1.0812,
+      "step": 8692
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6354460061590986e-05,
+      "loss": 1.0935,
+      "step": 8693
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6353590040977046e-05,
+      "loss": 0.9363,
+      "step": 8694
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6352719939706502e-05,
+      "loss": 1.0203,
+      "step": 8695
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6351849757790395e-05,
+      "loss": 1.1077,
+      "step": 8696
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6350979495239784e-05,
+      "loss": 1.1295,
+      "step": 8697
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.63501091520657e-05,
+      "loss": 1.0718,
+      "step": 8698
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6349238728279202e-05,
+      "loss": 1.0728,
+      "step": 8699
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6348368223891337e-05,
+      "loss": 1.0559,
+      "step": 8700
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6347497638913158e-05,
+      "loss": 1.0733,
+      "step": 8701
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6346626973355716e-05,
+      "loss": 1.0156,
+      "step": 8702
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634575622723006e-05,
+      "loss": 1.1542,
+      "step": 8703
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6344885400547246e-05,
+      "loss": 1.0038,
+      "step": 8704
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634401449331833e-05,
+      "loss": 1.0028,
+      "step": 8705
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634314350555437e-05,
+      "loss": 0.9666,
+      "step": 8706
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634227243726642e-05,
+      "loss": 0.8902,
+      "step": 8707
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.634140128846554e-05,
+      "loss": 1.0422,
+      "step": 8708
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6340530059162784e-05,
+      "loss": 1.0127,
+      "step": 8709
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6339658749369216e-05,
+      "loss": 1.1294,
+      "step": 8710
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6338787359095896e-05,
+      "loss": 1.0062,
+      "step": 8711
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6337915888353887e-05,
+      "loss": 1.0811,
+      "step": 8712
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633704433715425e-05,
+      "loss": 0.994,
+      "step": 8713
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633617270550805e-05,
+      "loss": 0.8922,
+      "step": 8714
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6335300993426352e-05,
+      "loss": 0.9487,
+      "step": 8715
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6334429200920227e-05,
+      "loss": 1.0912,
+      "step": 8716
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633355732800073e-05,
+      "loss": 0.867,
+      "step": 8717
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6332685374678942e-05,
+      "loss": 1.1213,
+      "step": 8718
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6331813340965924e-05,
+      "loss": 0.9682,
+      "step": 8719
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.633094122687275e-05,
+      "loss": 1.0666,
+      "step": 8720
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6330069032410486e-05,
+      "loss": 1.0951,
+      "step": 8721
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6329196757590217e-05,
+      "loss": 1.022,
+      "step": 8722
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6328324402423e-05,
+      "loss": 1.213,
+      "step": 8723
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632745196691992e-05,
+      "loss": 1.1178,
+      "step": 8724
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6326579451092048e-05,
+      "loss": 1.1392,
+      "step": 8725
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632570685495046e-05,
+      "loss": 1.0027,
+      "step": 8726
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6324834178506237e-05,
+      "loss": 1.2934,
+      "step": 8727
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6323961421770452e-05,
+      "loss": 0.9238,
+      "step": 8728
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632308858475419e-05,
+      "loss": 1.1247,
+      "step": 8729
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6322215667468527e-05,
+      "loss": 0.9199,
+      "step": 8730
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6321342669924545e-05,
+      "loss": 1.2013,
+      "step": 8731
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.632046959213333e-05,
+      "loss": 1.0263,
+      "step": 8732
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631959643410596e-05,
+      "loss": 1.1259,
+      "step": 8733
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6318723195853526e-05,
+      "loss": 1.0118,
+      "step": 8734
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6317849877387107e-05,
+      "loss": 0.9712,
+      "step": 8735
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631697647871779e-05,
+      "loss": 0.9631,
+      "step": 8736
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6316102999856668e-05,
+      "loss": 1.0162,
+      "step": 8737
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6315229440814823e-05,
+      "loss": 1.0418,
+      "step": 8738
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6314355801603353e-05,
+      "loss": 0.9048,
+      "step": 8739
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631348208223334e-05,
+      "loss": 1.0501,
+      "step": 8740
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6312608282715876e-05,
+      "loss": 1.0924,
+      "step": 8741
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.631173440306206e-05,
+      "loss": 1.1043,
+      "step": 8742
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6310860443282984e-05,
+      "loss": 1.0582,
+      "step": 8743
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309986403389734e-05,
+      "loss": 0.9912,
+      "step": 8744
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6309112283393415e-05,
+      "loss": 1.1661,
+      "step": 8745
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6308238083305123e-05,
+      "loss": 0.9281,
+      "step": 8746
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630736380313595e-05,
+      "loss": 1.0582,
+      "step": 8747
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6306489442897e-05,
+      "loss": 1.0483,
+      "step": 8748
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6305615002599374e-05,
+      "loss": 1.1513,
+      "step": 8749
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6304740482254166e-05,
+      "loss": 1.0605,
+      "step": 8750
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630386588187248e-05,
+      "loss": 1.0718,
+      "step": 8751
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302991201465426e-05,
+      "loss": 1.144,
+      "step": 8752
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6302116441044097e-05,
+      "loss": 1.0825,
+      "step": 8753
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6301241600619602e-05,
+      "loss": 1.062,
+      "step": 8754
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.630036668020305e-05,
+      "loss": 1.1525,
+      "step": 8755
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6299491679805544e-05,
+      "loss": 1.0343,
+      "step": 8756
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6298616599438194e-05,
+      "loss": 1.0083,
+      "step": 8757
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6297741439112106e-05,
+      "loss": 1.043,
+      "step": 8758
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6296866198838393e-05,
+      "loss": 1.028,
+      "step": 8759
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6295990878628164e-05,
+      "loss": 0.989,
+      "step": 8760
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629511547849253e-05,
+      "loss": 1.022,
+      "step": 8761
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6294239998442608e-05,
+      "loss": 1.0951,
+      "step": 8762
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629336443848951e-05,
+      "loss": 1.2036,
+      "step": 8763
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6292488798644348e-05,
+      "loss": 1.1808,
+      "step": 8764
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.629161307891824e-05,
+      "loss": 1.1938,
+      "step": 8765
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6290737279322305e-05,
+      "loss": 1.1389,
+      "step": 8766
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 1.6289861399867657e-05,
+      "loss": 1.0641,
+      "step": 8767
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6288985440565416e-05,
+      "loss": 1.1049,
+      "step": 8768
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6288109401426707e-05,
+      "loss": 1.0771,
+      "step": 8769
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6287233282462644e-05,
+      "loss": 0.9959,
+      "step": 8770
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6286357083684354e-05,
+      "loss": 1.0306,
+      "step": 8771
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6285480805102957e-05,
+      "loss": 1.1029,
+      "step": 8772
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628460444672958e-05,
+      "loss": 1.109,
+      "step": 8773
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6283728008575342e-05,
+      "loss": 1.0423,
+      "step": 8774
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6282851490651376e-05,
+      "loss": 1.0615,
+      "step": 8775
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.628197489296881e-05,
+      "loss": 0.9722,
+      "step": 8776
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6281098215538765e-05,
+      "loss": 0.978,
+      "step": 8777
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6280221458372372e-05,
+      "loss": 1.0927,
+      "step": 8778
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6279344621480768e-05,
+      "loss": 1.1454,
+      "step": 8779
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6278467704875072e-05,
+      "loss": 1.1321,
+      "step": 8780
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6277590708566426e-05,
+      "loss": 1.0441,
+      "step": 8781
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6276713632565962e-05,
+      "loss": 1.084,
+      "step": 8782
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627583647688481e-05,
+      "loss": 0.992,
+      "step": 8783
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627495924153411e-05,
+      "loss": 0.9936,
+      "step": 8784
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6274081926524992e-05,
+      "loss": 1.0559,
+      "step": 8785
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62732045318686e-05,
+      "loss": 1.0419,
+      "step": 8786
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627232705757607e-05,
+      "loss": 0.9943,
+      "step": 8787
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627144950365854e-05,
+      "loss": 1.2075,
+      "step": 8788
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.627057187012715e-05,
+      "loss": 0.9123,
+      "step": 8789
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626969415699304e-05,
+      "loss": 1.0287,
+      "step": 8790
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6268816364267355e-05,
+      "loss": 1.2132,
+      "step": 8791
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6267938491961236e-05,
+      "loss": 1.0278,
+      "step": 8792
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.626706054008583e-05,
+      "loss": 1.0935,
+      "step": 8793
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6266182508652285e-05,
+      "loss": 1.0623,
+      "step": 8794
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6265304397671736e-05,
+      "loss": 1.0699,
+      "step": 8795
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6264426207155345e-05,
+      "loss": 0.9433,
+      "step": 8796
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6263547937114247e-05,
+      "loss": 0.9557,
+      "step": 8797
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6262669587559604e-05,
+      "loss": 1.0672,
+      "step": 8798
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6261791158502554e-05,
+      "loss": 0.97,
+      "step": 8799
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260912649954254e-05,
+      "loss": 1.0488,
+      "step": 8800
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6260034061925857e-05,
+      "loss": 1.0653,
+      "step": 8801
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6259155394428517e-05,
+      "loss": 0.9926,
+      "step": 8802
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625827664747339e-05,
+      "loss": 1.0793,
+      "step": 8803
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625739782107162e-05,
+      "loss": 1.0287,
+      "step": 8804
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625651891523438e-05,
+      "loss": 0.9461,
+      "step": 8805
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6255639929972815e-05,
+      "loss": 1.1088,
+      "step": 8806
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6254760865298087e-05,
+      "loss": 1.026,
+      "step": 8807
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625388172122136e-05,
+      "loss": 1.007,
+      "step": 8808
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.625300249775379e-05,
+      "loss": 1.1036,
+      "step": 8809
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6252123194906536e-05,
+      "loss": 1.0068,
+      "step": 8810
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6251243812690764e-05,
+      "loss": 1.0425,
+      "step": 8811
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6250364351117638e-05,
+      "loss": 1.047,
+      "step": 8812
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624948481019832e-05,
+      "loss": 1.0707,
+      "step": 8813
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6248605189943977e-05,
+      "loss": 1.0082,
+      "step": 8814
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6247725490365776e-05,
+      "loss": 1.0474,
+      "step": 8815
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6246845711474883e-05,
+      "loss": 1.0158,
+      "step": 8816
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624596585328247e-05,
+      "loss": 1.0041,
+      "step": 8817
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62450859157997e-05,
+      "loss": 0.9811,
+      "step": 8818
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6244205899037748e-05,
+      "loss": 1.032,
+      "step": 8819
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6243325803007786e-05,
+      "loss": 1.0717,
+      "step": 8820
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6242445627720984e-05,
+      "loss": 1.1629,
+      "step": 8821
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624156537318852e-05,
+      "loss": 0.9542,
+      "step": 8822
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.624068503942156e-05,
+      "loss": 1.0815,
+      "step": 8823
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6239804626431288e-05,
+      "loss": 0.978,
+      "step": 8824
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6238924134228878e-05,
+      "loss": 1.0516,
+      "step": 8825
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.62380435628255e-05,
+      "loss": 1.048,
+      "step": 8826
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623716291223235e-05,
+      "loss": 0.9658,
+      "step": 8827
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6236282182460587e-05,
+      "loss": 1.0655,
+      "step": 8828
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6235401373521407e-05,
+      "loss": 0.9946,
+      "step": 8829
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623452048542599e-05,
+      "loss": 1.0427,
+      "step": 8830
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6233639518185507e-05,
+      "loss": 1.0301,
+      "step": 8831
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.623275847181115e-05,
+      "loss": 0.9929,
+      "step": 8832
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6231877346314106e-05,
+      "loss": 1.1091,
+      "step": 8833
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6230996141705556e-05,
+      "loss": 1.0506,
+      "step": 8834
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6230114857996688e-05,
+      "loss": 1.0087,
+      "step": 8835
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622923349519869e-05,
+      "loss": 0.952,
+      "step": 8836
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622835205332275e-05,
+      "loss": 1.0193,
+      "step": 8837
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6227470532380056e-05,
+      "loss": 1.1314,
+      "step": 8838
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6226588932381805e-05,
+      "loss": 1.0971,
+      "step": 8839
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622570725333918e-05,
+      "loss": 1.1454,
+      "step": 8840
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6224825495263375e-05,
+      "loss": 0.998,
+      "step": 8841
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223943658165593e-05,
+      "loss": 1.0539,
+      "step": 8842
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6223061742057017e-05,
+      "loss": 0.9128,
+      "step": 8843
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.622217974694885e-05,
+      "loss": 1.0876,
+      "step": 8844
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6221297672852282e-05,
+      "loss": 1.1452,
+      "step": 8845
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6220415519778518e-05,
+      "loss": 1.0859,
+      "step": 8846
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6219533287738752e-05,
+      "loss": 0.994,
+      "step": 8847
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6218650976744186e-05,
+      "loss": 1.0383,
+      "step": 8848
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621776858680602e-05,
+      "loss": 1.0951,
+      "step": 8849
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6216886117935452e-05,
+      "loss": 0.9638,
+      "step": 8850
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.621600357014369e-05,
+      "loss": 1.0462,
+      "step": 8851
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6215120943441934e-05,
+      "loss": 1.1373,
+      "step": 8852
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6214238237841392e-05,
+      "loss": 0.9868,
+      "step": 8853
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6213355453353267e-05,
+      "loss": 0.9714,
+      "step": 8854
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6212472589988766e-05,
+      "loss": 0.9825,
+      "step": 8855
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6211589647759098e-05,
+      "loss": 1.1072,
+      "step": 8856
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6210706626675466e-05,
+      "loss": 1.0237,
+      "step": 8857
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620982352674909e-05,
+      "loss": 1.0152,
+      "step": 8858
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6208940347991173e-05,
+      "loss": 1.0637,
+      "step": 8859
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620805709041293e-05,
+      "loss": 0.9266,
+      "step": 8860
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6207173754025568e-05,
+      "loss": 1.113,
+      "step": 8861
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6206290338840308e-05,
+      "loss": 1.1427,
+      "step": 8862
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620540684486836e-05,
+      "loss": 1.1972,
+      "step": 8863
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6204523272120942e-05,
+      "loss": 0.985,
+      "step": 8864
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6203639620609272e-05,
+      "loss": 1.0091,
+      "step": 8865
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6202755890344565e-05,
+      "loss": 1.1161,
+      "step": 8866
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6201872081338038e-05,
+      "loss": 1.0205,
+      "step": 8867
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.620098819360091e-05,
+      "loss": 0.9209,
+      "step": 8868
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6200104227144406e-05,
+      "loss": 1.0763,
+      "step": 8869
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6199220181979747e-05,
+      "loss": 1.0673,
+      "step": 8870
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6198336058118155e-05,
+      "loss": 1.0419,
+      "step": 8871
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6197451855570854e-05,
+      "loss": 1.2513,
+      "step": 8872
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619656757434907e-05,
+      "loss": 1.1188,
+      "step": 8873
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619568321446402e-05,
+      "loss": 1.0886,
+      "step": 8874
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6194798775926944e-05,
+      "loss": 1.0987,
+      "step": 8875
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619391425874906e-05,
+      "loss": 1.0435,
+      "step": 8876
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6193029662941597e-05,
+      "loss": 1.0705,
+      "step": 8877
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.619214498851579e-05,
+      "loss": 1.1521,
+      "step": 8878
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6191260235482867e-05,
+      "loss": 1.0451,
+      "step": 8879
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6190375403854056e-05,
+      "loss": 1.0195,
+      "step": 8880
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6189490493640597e-05,
+      "loss": 1.1309,
+      "step": 8881
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6188605504853717e-05,
+      "loss": 0.8925,
+      "step": 8882
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6187720437504656e-05,
+      "loss": 1.0336,
+      "step": 8883
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6186835291604643e-05,
+      "loss": 1.0996,
+      "step": 8884
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185950067164922e-05,
+      "loss": 1.0045,
+      "step": 8885
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6185064764196727e-05,
+      "loss": 0.9609,
+      "step": 8886
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6184179382711294e-05,
+      "loss": 1.0687,
+      "step": 8887
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6183293922719867e-05,
+      "loss": 1.1214,
+      "step": 8888
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6182408384233682e-05,
+      "loss": 1.1076,
+      "step": 8889
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618152276726399e-05,
+      "loss": 1.0677,
+      "step": 8890
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.618063707182202e-05,
+      "loss": 0.9211,
+      "step": 8891
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6179751297919028e-05,
+      "loss": 1.0337,
+      "step": 8892
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6178865445566252e-05,
+      "loss": 1.1819,
+      "step": 8893
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617797951477494e-05,
+      "loss": 1.1093,
+      "step": 8894
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6177093505556336e-05,
+      "loss": 1.1145,
+      "step": 8895
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6176207417921687e-05,
+      "loss": 0.9186,
+      "step": 8896
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6175321251882247e-05,
+      "loss": 0.8967,
+      "step": 8897
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617443500744926e-05,
+      "loss": 1.0759,
+      "step": 8898
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6173548684633978e-05,
+      "loss": 1.0067,
+      "step": 8899
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6172662283447654e-05,
+      "loss": 1.0461,
+      "step": 8900
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6171775803901543e-05,
+      "loss": 1.1438,
+      "step": 8901
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6170889246006894e-05,
+      "loss": 1.129,
+      "step": 8902
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.617000260977496e-05,
+      "loss": 1.0356,
+      "step": 8903
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6169115895217004e-05,
+      "loss": 1.0529,
+      "step": 8904
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6168229102344275e-05,
+      "loss": 1.052,
+      "step": 8905
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6167342231168036e-05,
+      "loss": 1.0925,
+      "step": 8906
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6166455281699538e-05,
+      "loss": 1.0499,
+      "step": 8907
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616556825395005e-05,
+      "loss": 1.0119,
+      "step": 8908
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616468114793083e-05,
+      "loss": 1.0374,
+      "step": 8909
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6163793963653135e-05,
+      "loss": 1.1055,
+      "step": 8910
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616290670112823e-05,
+      "loss": 1.2852,
+      "step": 8911
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6162019360367376e-05,
+      "loss": 0.9909,
+      "step": 8912
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.616113194138184e-05,
+      "loss": 1.1537,
+      "step": 8913
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6160244444182894e-05,
+      "loss": 1.0721,
+      "step": 8914
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615935686878179e-05,
+      "loss": 1.0686,
+      "step": 8915
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615846921518981e-05,
+      "loss": 1.0482,
+      "step": 8916
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615758148341821e-05,
+      "loss": 1.0463,
+      "step": 8917
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6156693673478273e-05,
+      "loss": 1.0246,
+      "step": 8918
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6155805785381257e-05,
+      "loss": 1.1175,
+      "step": 8919
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.615491781913844e-05,
+      "loss": 1.0561,
+      "step": 8920
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6154029774761092e-05,
+      "loss": 1.1283,
+      "step": 8921
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6153141652260487e-05,
+      "loss": 1.0676,
+      "step": 8922
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61522534516479e-05,
+      "loss": 1.0166,
+      "step": 8923
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6151365172934604e-05,
+      "loss": 1.0212,
+      "step": 8924
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6150476816131876e-05,
+      "loss": 1.0795,
+      "step": 8925
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6149588381250996e-05,
+      "loss": 1.1603,
+      "step": 8926
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6148699868303243e-05,
+      "loss": 1.0371,
+      "step": 8927
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6147811277299893e-05,
+      "loss": 1.1301,
+      "step": 8928
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146922608252228e-05,
+      "loss": 0.8842,
+      "step": 8929
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6146033861171528e-05,
+      "loss": 0.9974,
+      "step": 8930
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6145145036069077e-05,
+      "loss": 1.1034,
+      "step": 8931
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614425613295616e-05,
+      "loss": 1.1471,
+      "step": 8932
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6143367151844058e-05,
+      "loss": 1.0099,
+      "step": 8933
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6142478092744056e-05,
+      "loss": 0.9275,
+      "step": 8934
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.614158895566744e-05,
+      "loss": 0.9238,
+      "step": 8935
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61406997406255e-05,
+      "loss": 1.0594,
+      "step": 8936
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6139810447629528e-05,
+      "loss": 1.0953,
+      "step": 8937
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138921076690803e-05,
+      "loss": 1.1624,
+      "step": 8938
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6138031627820622e-05,
+      "loss": 1.058,
+      "step": 8939
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6137142101030275e-05,
+      "loss": 0.9846,
+      "step": 8940
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6136252496331057e-05,
+      "loss": 1.046,
+      "step": 8941
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6135362813734253e-05,
+      "loss": 1.1186,
+      "step": 8942
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6134473053251168e-05,
+      "loss": 1.1314,
+      "step": 8943
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.613358321489309e-05,
+      "loss": 1.0416,
+      "step": 8944
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6132693298671315e-05,
+      "loss": 0.9838,
+      "step": 8945
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6131803304597143e-05,
+      "loss": 1.0424,
+      "step": 8946
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6130913232681872e-05,
+      "loss": 1.0279,
+      "step": 8947
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61300230829368e-05,
+      "loss": 1.0183,
+      "step": 8948
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6129132855373225e-05,
+      "loss": 0.9843,
+      "step": 8949
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6128242550002454e-05,
+      "loss": 1.0623,
+      "step": 8950
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612735216683579e-05,
+      "loss": 1.0559,
+      "step": 8951
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6126461705884528e-05,
+      "loss": 1.0634,
+      "step": 8952
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612557116715997e-05,
+      "loss": 1.0006,
+      "step": 8953
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6124680550673436e-05,
+      "loss": 1.0732,
+      "step": 8954
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6123789856436222e-05,
+      "loss": 1.0764,
+      "step": 8955
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6122899084459632e-05,
+      "loss": 0.9704,
+      "step": 8956
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.612200823475498e-05,
+      "loss": 1.114,
+      "step": 8957
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6121117307333576e-05,
+      "loss": 1.0262,
+      "step": 8958
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6120226302206722e-05,
+      "loss": 1.0935,
+      "step": 8959
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611933521938574e-05,
+      "loss": 1.09,
+      "step": 8960
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611844405888193e-05,
+      "loss": 1.0293,
+      "step": 8961
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6117552820706616e-05,
+      "loss": 0.9423,
+      "step": 8962
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6116661504871106e-05,
+      "loss": 1.0008,
+      "step": 8963
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.611577011138672e-05,
+      "loss": 1.0447,
+      "step": 8964
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6114878640264765e-05,
+      "loss": 1.0671,
+      "step": 8965
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113987091516567e-05,
+      "loss": 1.117,
+      "step": 8966
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6113095465153438e-05,
+      "loss": 1.0059,
+      "step": 8967
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6112203761186696e-05,
+      "loss": 1.0497,
+      "step": 8968
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6111311979627667e-05,
+      "loss": 1.074,
+      "step": 8969
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6110420120487667e-05,
+      "loss": 1.1289,
+      "step": 8970
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610952818377802e-05,
+      "loss": 1.0838,
+      "step": 8971
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6108636169510048e-05,
+      "loss": 0.9341,
+      "step": 8972
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6107744077695075e-05,
+      "loss": 1.1,
+      "step": 8973
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6106851908344427e-05,
+      "loss": 1.0103,
+      "step": 8974
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105959661469427e-05,
+      "loss": 1.0781,
+      "step": 8975
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6105067337081406e-05,
+      "loss": 1.1572,
+      "step": 8976
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6104174935191685e-05,
+      "loss": 1.0425,
+      "step": 8977
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.61032824558116e-05,
+      "loss": 1.0236,
+      "step": 8978
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610238989895248e-05,
+      "loss": 0.9797,
+      "step": 8979
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6101497264625646e-05,
+      "loss": 1.1285,
+      "step": 8980
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.610060455284244e-05,
+      "loss": 0.9781,
+      "step": 8981
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6099711763614193e-05,
+      "loss": 0.9844,
+      "step": 8982
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6098818896952236e-05,
+      "loss": 0.9105,
+      "step": 8983
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6097925952867907e-05,
+      "loss": 1.0296,
+      "step": 8984
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609703293137254e-05,
+      "loss": 1.0994,
+      "step": 8985
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6096139832477465e-05,
+      "loss": 0.9875,
+      "step": 8986
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609524665619403e-05,
+      "loss": 1.0685,
+      "step": 8987
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609435340253357e-05,
+      "loss": 0.8793,
+      "step": 8988
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6093460071507424e-05,
+      "loss": 1.1377,
+      "step": 8989
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.609256666312693e-05,
+      "loss": 1.0244,
+      "step": 8990
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6091673177403434e-05,
+      "loss": 1.052,
+      "step": 8991
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6090779614348275e-05,
+      "loss": 1.1305,
+      "step": 8992
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6089885973972793e-05,
+      "loss": 1.0166,
+      "step": 8993
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088992256288344e-05,
+      "loss": 1.0965,
+      "step": 8994
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6088098461306266e-05,
+      "loss": 0.9474,
+      "step": 8995
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60872045890379e-05,
+      "loss": 1.1265,
+      "step": 8996
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6086310639494608e-05,
+      "loss": 1.0919,
+      "step": 8997
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6085416612687724e-05,
+      "loss": 1.0057,
+      "step": 8998
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6084522508628606e-05,
+      "loss": 1.0508,
+      "step": 8999
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60836283273286e-05,
+      "loss": 1.1121,
+      "step": 9000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608273406879906e-05,
+      "loss": 1.0633,
+      "step": 9001
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6081839733051338e-05,
+      "loss": 1.0332,
+      "step": 9002
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.608094532009678e-05,
+      "loss": 1.0106,
+      "step": 9003
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6080050829946757e-05,
+      "loss": 0.9731,
+      "step": 9004
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6079156262612604e-05,
+      "loss": 0.9463,
+      "step": 9005
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6078261618105693e-05,
+      "loss": 1.1523,
+      "step": 9006
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6077366896437374e-05,
+      "loss": 0.9,
+      "step": 9007
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6076472097619007e-05,
+      "loss": 1.1019,
+      "step": 9008
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607557722166195e-05,
+      "loss": 0.9792,
+      "step": 9009
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6074682268577564e-05,
+      "loss": 0.9818,
+      "step": 9010
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607378723837721e-05,
+      "loss": 1.0227,
+      "step": 9011
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607289213107225e-05,
+      "loss": 1.0886,
+      "step": 9012
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071996946674048e-05,
+      "loss": 1.039,
+      "step": 9013
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6071101685193962e-05,
+      "loss": 1.0057,
+      "step": 9014
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.607020634664337e-05,
+      "loss": 0.9045,
+      "step": 9015
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6069310931033628e-05,
+      "loss": 0.9469,
+      "step": 9016
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60684154383761e-05,
+      "loss": 1.0083,
+      "step": 9017
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6067519868682164e-05,
+      "loss": 1.0374,
+      "step": 9018
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606662422196318e-05,
+      "loss": 0.9812,
+      "step": 9019
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6065728498230527e-05,
+      "loss": 1.1078,
+      "step": 9020
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606483269749557e-05,
+      "loss": 1.0857,
+      "step": 9021
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606393681976968e-05,
+      "loss": 0.9503,
+      "step": 9022
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6063040865064234e-05,
+      "loss": 0.8403,
+      "step": 9023
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.60621448333906e-05,
+      "loss": 1.1405,
+      "step": 9024
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606124872476016e-05,
+      "loss": 0.9349,
+      "step": 9025
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.606035253918428e-05,
+      "loss": 1.0907,
+      "step": 9026
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6059456276674348e-05,
+      "loss": 0.98,
+      "step": 9027
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6058559937241733e-05,
+      "loss": 1.0453,
+      "step": 9028
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6057663520897825e-05,
+      "loss": 0.9006,
+      "step": 9029
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6056767027653987e-05,
+      "loss": 1.1646,
+      "step": 9030
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6055870457521612e-05,
+      "loss": 1.1989,
+      "step": 9031
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605497381051208e-05,
+      "loss": 1.12,
+      "step": 9032
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605407708663677e-05,
+      "loss": 1.1899,
+      "step": 9033
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.605318028590707e-05,
+      "loss": 1.1357,
+      "step": 9034
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6052283408334356e-05,
+      "loss": 1.1049,
+      "step": 9035
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6051386453930027e-05,
+      "loss": 1.0742,
+      "step": 9036
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6050489422705458e-05,
+      "loss": 1.0123,
+      "step": 9037
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6049592314672043e-05,
+      "loss": 1.0287,
+      "step": 9038
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6048695129841166e-05,
+      "loss": 1.093,
+      "step": 9039
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604779786822422e-05,
+      "loss": 1.1496,
+      "step": 9040
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604690052983259e-05,
+      "loss": 1.1263,
+      "step": 9041
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6046003114677677e-05,
+      "loss": 1.0546,
+      "step": 9042
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6045105622770865e-05,
+      "loss": 1.0039,
+      "step": 9043
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604420805412355e-05,
+      "loss": 1.0531,
+      "step": 9044
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6043310408747126e-05,
+      "loss": 0.9712,
+      "step": 9045
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.604241268665299e-05,
+      "loss": 1.0891,
+      "step": 9046
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6041514887852536e-05,
+      "loss": 0.9284,
+      "step": 9047
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6040617012357164e-05,
+      "loss": 1.0957,
+      "step": 9048
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6039719060178266e-05,
+      "loss": 0.949,
+      "step": 9049
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6038821031327253e-05,
+      "loss": 1.1533,
+      "step": 9050
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6037922925815513e-05,
+      "loss": 1.0769,
+      "step": 9051
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603702474365445e-05,
+      "loss": 0.8894,
+      "step": 9052
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603612648485547e-05,
+      "loss": 0.9655,
+      "step": 9053
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.6035228149429977e-05,
+      "loss": 1.197,
+      "step": 9054
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 1.603432973738937e-05,
+      "loss": 1.0982,
+      "step": 9055
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6033431248745058e-05,
+      "loss": 1.0956,
+      "step": 9056
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6032532683508443e-05,
+      "loss": 1.0578,
+      "step": 9057
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603163404169093e-05,
+      "loss": 1.1858,
+      "step": 9058
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.603073532330394e-05,
+      "loss": 1.0787,
+      "step": 9059
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602983652835887e-05,
+      "loss": 1.0305,
+      "step": 9060
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6028937656867128e-05,
+      "loss": 1.1015,
+      "step": 9061
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6028038708840135e-05,
+      "loss": 0.9749,
+      "step": 9062
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6027139684289293e-05,
+      "loss": 1.0703,
+      "step": 9063
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6026240583226025e-05,
+      "loss": 1.0677,
+      "step": 9064
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.602534140566174e-05,
+      "loss": 1.0562,
+      "step": 9065
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6024442151607846e-05,
+      "loss": 1.1881,
+      "step": 9066
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6023542821075766e-05,
+      "loss": 1.0214,
+      "step": 9067
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6022643414076914e-05,
+      "loss": 1.0461,
+      "step": 9068
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6021743930622716e-05,
+      "loss": 1.1018,
+      "step": 9069
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6020844370724574e-05,
+      "loss": 1.1582,
+      "step": 9070
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6019944734393926e-05,
+      "loss": 1.1003,
+      "step": 9071
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601904502164218e-05,
+      "loss": 0.9184,
+      "step": 9072
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601814523248076e-05,
+      "loss": 1.061,
+      "step": 9073
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6017245366921093e-05,
+      "loss": 1.1027,
+      "step": 9074
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6016345424974598e-05,
+      "loss": 1.142,
+      "step": 9075
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6015445406652697e-05,
+      "loss": 1.105,
+      "step": 9076
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6014545311966822e-05,
+      "loss": 0.9674,
+      "step": 9077
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.60136451409284e-05,
+      "loss": 1.0574,
+      "step": 9078
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6012744893548853e-05,
+      "loss": 1.0941,
+      "step": 9079
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.601184456983961e-05,
+      "loss": 0.9907,
+      "step": 9080
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010944169812098e-05,
+      "loss": 0.9713,
+      "step": 9081
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6010043693477758e-05,
+      "loss": 1.0823,
+      "step": 9082
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600914314084801e-05,
+      "loss": 1.0607,
+      "step": 9083
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600824251193429e-05,
+      "loss": 0.9946,
+      "step": 9084
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6007341806748034e-05,
+      "loss": 1.0635,
+      "step": 9085
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6006441025300673e-05,
+      "loss": 1.0448,
+      "step": 9086
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.600554016760364e-05,
+      "loss": 1.1047,
+      "step": 9087
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6004639233668376e-05,
+      "loss": 0.9544,
+      "step": 9088
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6003738223506314e-05,
+      "loss": 1.1182,
+      "step": 9089
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6002837137128895e-05,
+      "loss": 0.9544,
+      "step": 9090
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6001935974547556e-05,
+      "loss": 1.0477,
+      "step": 9091
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6001034735773738e-05,
+      "loss": 1.1337,
+      "step": 9092
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.6000133420818884e-05,
+      "loss": 0.9628,
+      "step": 9093
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599923202969443e-05,
+      "loss": 1.0558,
+      "step": 9094
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5998330562411823e-05,
+      "loss": 0.9907,
+      "step": 9095
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5997429018982506e-05,
+      "loss": 1.0195,
+      "step": 9096
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5996527399417928e-05,
+      "loss": 0.9321,
+      "step": 9097
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5995625703729526e-05,
+      "loss": 0.9527,
+      "step": 9098
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5994723931928753e-05,
+      "loss": 1.1244,
+      "step": 9099
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5993822084027054e-05,
+      "loss": 1.1808,
+      "step": 9100
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599292016003588e-05,
+      "loss": 1.2016,
+      "step": 9101
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599201815996668e-05,
+      "loss": 0.8999,
+      "step": 9102
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59911160838309e-05,
+      "loss": 1.0129,
+      "step": 9103
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.599021393164e-05,
+      "loss": 0.9635,
+      "step": 9104
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5989311703405426e-05,
+      "loss": 0.9861,
+      "step": 9105
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5988409399138634e-05,
+      "loss": 1.0429,
+      "step": 9106
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5987507018851075e-05,
+      "loss": 1.1004,
+      "step": 9107
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5986604562554213e-05,
+      "loss": 1.1819,
+      "step": 9108
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5985702030259494e-05,
+      "loss": 0.9882,
+      "step": 9109
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5984799421978377e-05,
+      "loss": 1.1366,
+      "step": 9110
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598389673772233e-05,
+      "loss": 1.0105,
+      "step": 9111
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982993977502804e-05,
+      "loss": 1.1161,
+      "step": 9112
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5982091141331258e-05,
+      "loss": 0.9894,
+      "step": 9113
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.598118822921916e-05,
+      "loss": 1.002,
+      "step": 9114
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5980285241177964e-05,
+      "loss": 1.114,
+      "step": 9115
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597938217721914e-05,
+      "loss": 1.0565,
+      "step": 9116
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5978479037354145e-05,
+      "loss": 1.0763,
+      "step": 9117
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5977575821594453e-05,
+      "loss": 1.0507,
+      "step": 9118
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5976672529951526e-05,
+      "loss": 1.0244,
+      "step": 9119
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5975769162436825e-05,
+      "loss": 1.0874,
+      "step": 9120
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.597486571906183e-05,
+      "loss": 1.0834,
+      "step": 9121
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973962199837996e-05,
+      "loss": 1.0702,
+      "step": 9122
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5973058604776804e-05,
+      "loss": 0.9735,
+      "step": 9123
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5972154933889722e-05,
+      "loss": 0.9414,
+      "step": 9124
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5971251187188217e-05,
+      "loss": 1.0193,
+      "step": 9125
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5970347364683768e-05,
+      "loss": 1.0331,
+      "step": 9126
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5969443466387848e-05,
+      "loss": 1.1702,
+      "step": 9127
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5968539492311922e-05,
+      "loss": 1.1018,
+      "step": 9128
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596763544246748e-05,
+      "loss": 1.0065,
+      "step": 9129
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5966731316865992e-05,
+      "loss": 1.0381,
+      "step": 9130
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5965827115518932e-05,
+      "loss": 1.0939,
+      "step": 9131
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964922838437782e-05,
+      "loss": 0.8948,
+      "step": 9132
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5964018485634025e-05,
+      "loss": 1.0661,
+      "step": 9133
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5963114057119136e-05,
+      "loss": 1.0962,
+      "step": 9134
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59622095529046e-05,
+      "loss": 1.0552,
+      "step": 9135
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5961304973001895e-05,
+      "loss": 0.9698,
+      "step": 9136
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.596040031742251e-05,
+      "loss": 1.0207,
+      "step": 9137
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5959495586177924e-05,
+      "loss": 1.1141,
+      "step": 9138
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5958590779279625e-05,
+      "loss": 1.033,
+      "step": 9139
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5957685896739097e-05,
+      "loss": 1.1638,
+      "step": 9140
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5956780938567832e-05,
+      "loss": 1.0683,
+      "step": 9141
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5955875904777317e-05,
+      "loss": 1.0191,
+      "step": 9142
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954970795379035e-05,
+      "loss": 1.0196,
+      "step": 9143
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5954065610384482e-05,
+      "loss": 1.0656,
+      "step": 9144
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5953160349805148e-05,
+      "loss": 1.1652,
+      "step": 9145
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.595225501365252e-05,
+      "loss": 1.0381,
+      "step": 9146
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59513496019381e-05,
+      "loss": 0.9899,
+      "step": 9147
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5950444114673376e-05,
+      "loss": 1.0872,
+      "step": 9148
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5949538551869844e-05,
+      "loss": 1.0721,
+      "step": 9149
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5948632913539e-05,
+      "loss": 1.0058,
+      "step": 9150
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594772719969234e-05,
+      "loss": 1.0744,
+      "step": 9151
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594682141034136e-05,
+      "loss": 1.0827,
+      "step": 9152
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945915545497563e-05,
+      "loss": 1.1122,
+      "step": 9153
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5945009605172442e-05,
+      "loss": 1.021,
+      "step": 9154
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594410358937751e-05,
+      "loss": 0.9958,
+      "step": 9155
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5943197498124252e-05,
+      "loss": 1.1371,
+      "step": 9156
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5942291331424185e-05,
+      "loss": 1.0118,
+      "step": 9157
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5941385089288804e-05,
+      "loss": 1.1176,
+      "step": 9158
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.594047877172962e-05,
+      "loss": 1.0907,
+      "step": 9159
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593957237875813e-05,
+      "loss": 1.0741,
+      "step": 9160
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5938665910385846e-05,
+      "loss": 1.1499,
+      "step": 9161
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5937759366624275e-05,
+      "loss": 1.0504,
+      "step": 9162
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5936852747484924e-05,
+      "loss": 0.8893,
+      "step": 9163
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935946052979302e-05,
+      "loss": 0.9349,
+      "step": 9164
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5935039283118922e-05,
+      "loss": 1.0092,
+      "step": 9165
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593413243791529e-05,
+      "loss": 1.1809,
+      "step": 9166
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.593322551737992e-05,
+      "loss": 1.0261,
+      "step": 9167
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5932318521524327e-05,
+      "loss": 1.0532,
+      "step": 9168
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5931411450360027e-05,
+      "loss": 1.0713,
+      "step": 9169
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5930504303898525e-05,
+      "loss": 0.8447,
+      "step": 9170
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592959708215135e-05,
+      "loss": 1.0546,
+      "step": 9171
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5928689785130013e-05,
+      "loss": 1.1138,
+      "step": 9172
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5927782412846028e-05,
+      "loss": 0.9655,
+      "step": 9173
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5926874965310917e-05,
+      "loss": 1.1043,
+      "step": 9174
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.59259674425362e-05,
+      "loss": 1.0596,
+      "step": 9175
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5925059844533403e-05,
+      "loss": 0.9127,
+      "step": 9176
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5924152171314034e-05,
+      "loss": 1.0227,
+      "step": 9177
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592324442288963e-05,
+      "loss": 1.0742,
+      "step": 9178
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592233659927171e-05,
+      "loss": 1.0516,
+      "step": 9179
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.592142870047179e-05,
+      "loss": 1.1213,
+      "step": 9180
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5920520726501408e-05,
+      "loss": 1.103,
+      "step": 9181
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591961267737208e-05,
+      "loss": 1.1169,
+      "step": 9182
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591870455309534e-05,
+      "loss": 0.9255,
+      "step": 9183
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5917796353682718e-05,
+      "loss": 1.102,
+      "step": 9184
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591688807914574e-05,
+      "loss": 1.0498,
+      "step": 9185
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5915979729495932e-05,
+      "loss": 1.1386,
+      "step": 9186
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591507130474483e-05,
+      "loss": 1.0427,
+      "step": 9187
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5914162804903965e-05,
+      "loss": 1.0156,
+      "step": 9188
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5913254229984874e-05,
+      "loss": 0.9396,
+      "step": 9189
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591234557999908e-05,
+      "loss": 0.9982,
+      "step": 9190
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5911436854958133e-05,
+      "loss": 0.9975,
+      "step": 9191
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.591052805487356e-05,
+      "loss": 1.0779,
+      "step": 9192
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5909619179756898e-05,
+      "loss": 1.1945,
+      "step": 9193
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5908710229619685e-05,
+      "loss": 1.0035,
+      "step": 9194
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5907801204473462e-05,
+      "loss": 1.005,
+      "step": 9195
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5906892104329766e-05,
+      "loss": 0.9859,
+      "step": 9196
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590598292920014e-05,
+      "loss": 1.1912,
+      "step": 9197
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5905073679096127e-05,
+      "loss": 1.0055,
+      "step": 9198
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590416435402927e-05,
+      "loss": 0.9541,
+      "step": 9199
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5903254954011106e-05,
+      "loss": 1.0437,
+      "step": 9200
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5902345479053183e-05,
+      "loss": 0.9316,
+      "step": 9201
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590143592916705e-05,
+      "loss": 1.0541,
+      "step": 9202
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.590052630436425e-05,
+      "loss": 1.0435,
+      "step": 9203
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5899616604656333e-05,
+      "loss": 0.9213,
+      "step": 9204
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589870683005484e-05,
+      "loss": 1.0197,
+      "step": 9205
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5897796980571328e-05,
+      "loss": 0.9381,
+      "step": 9206
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5896887056217348e-05,
+      "loss": 1.1609,
+      "step": 9207
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895977057004445e-05,
+      "loss": 1.0303,
+      "step": 9208
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5895066982944174e-05,
+      "loss": 1.0263,
+      "step": 9209
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5894156834048085e-05,
+      "loss": 1.1466,
+      "step": 9210
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5893246610327745e-05,
+      "loss": 1.1604,
+      "step": 9211
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.589233631179469e-05,
+      "loss": 1.1811,
+      "step": 9212
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5891425938460488e-05,
+      "loss": 0.9753,
+      "step": 9213
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5890515490336695e-05,
+      "loss": 1.0722,
+      "step": 9214
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5889604967434864e-05,
+      "loss": 1.0405,
+      "step": 9215
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5888694369766557e-05,
+      "loss": 1.0259,
+      "step": 9216
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5887783697343334e-05,
+      "loss": 1.0138,
+      "step": 9217
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5886872950176754e-05,
+      "loss": 1.0935,
+      "step": 9218
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588596212827838e-05,
+      "loss": 0.9359,
+      "step": 9219
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5885051231659775e-05,
+      "loss": 1.0868,
+      "step": 9220
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5884140260332502e-05,
+      "loss": 1.0494,
+      "step": 9221
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5883229214308124e-05,
+      "loss": 1.055,
+      "step": 9222
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588231809359821e-05,
+      "loss": 1.0277,
+      "step": 9223
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.588140689821432e-05,
+      "loss": 1.0038,
+      "step": 9224
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5880495628168027e-05,
+      "loss": 1.0399,
+      "step": 9225
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5879584283470898e-05,
+      "loss": 0.9377,
+      "step": 9226
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5878672864134503e-05,
+      "loss": 1.2226,
+      "step": 9227
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587776137017041e-05,
+      "loss": 0.9385,
+      "step": 9228
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5876849801590186e-05,
+      "loss": 1.1296,
+      "step": 9229
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875938158405412e-05,
+      "loss": 1.0039,
+      "step": 9230
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5875026440627658e-05,
+      "loss": 1.0096,
+      "step": 9231
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5874114648268493e-05,
+      "loss": 1.2018,
+      "step": 9232
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58732027813395e-05,
+      "loss": 1.091,
+      "step": 9233
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587229083985225e-05,
+      "loss": 0.992,
+      "step": 9234
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.587137882381832e-05,
+      "loss": 1.0063,
+      "step": 9235
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5870466733249284e-05,
+      "loss": 1.0752,
+      "step": 9236
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586955456815673e-05,
+      "loss": 1.106,
+      "step": 9237
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586864232855223e-05,
+      "loss": 1.0601,
+      "step": 9238
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586773001444737e-05,
+      "loss": 1.06,
+      "step": 9239
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5866817625853724e-05,
+      "loss": 1.1601,
+      "step": 9240
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586590516278288e-05,
+      "loss": 1.1267,
+      "step": 9241
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586499262524642e-05,
+      "loss": 1.1151,
+      "step": 9242
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586408001325593e-05,
+      "loss": 1.1294,
+      "step": 9243
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5863167326822994e-05,
+      "loss": 1.1224,
+      "step": 9244
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5862254565959195e-05,
+      "loss": 1.0038,
+      "step": 9245
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.586134173067613e-05,
+      "loss": 1.0084,
+      "step": 9246
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5860428820985374e-05,
+      "loss": 0.9798,
+      "step": 9247
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5859515836898527e-05,
+      "loss": 1.0777,
+      "step": 9248
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585860277842717e-05,
+      "loss": 1.1164,
+      "step": 9249
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5857689645582907e-05,
+      "loss": 0.9411,
+      "step": 9250
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5856776438377312e-05,
+      "loss": 0.9958,
+      "step": 9251
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5855863156821995e-05,
+      "loss": 1.0659,
+      "step": 9252
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854949800928534e-05,
+      "loss": 1.0744,
+      "step": 9253
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5854036370708533e-05,
+      "loss": 1.1924,
+      "step": 9254
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5853122866173594e-05,
+      "loss": 1.0441,
+      "step": 9255
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5852209287335302e-05,
+      "loss": 0.992,
+      "step": 9256
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585129563420526e-05,
+      "loss": 0.9193,
+      "step": 9257
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.585038190679506e-05,
+      "loss": 0.9751,
+      "step": 9258
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5849468105116308e-05,
+      "loss": 1.0207,
+      "step": 9259
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5848554229180607e-05,
+      "loss": 1.0422,
+      "step": 9260
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584764027899955e-05,
+      "loss": 1.1994,
+      "step": 9261
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5846726254584747e-05,
+      "loss": 0.9889,
+      "step": 9262
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.58458121559478e-05,
+      "loss": 1.005,
+      "step": 9263
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5844897983100306e-05,
+      "loss": 0.9595,
+      "step": 9264
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843983736053877e-05,
+      "loss": 1.0369,
+      "step": 9265
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5843069414820114e-05,
+      "loss": 0.9753,
+      "step": 9266
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5842155019410633e-05,
+      "loss": 1.2022,
+      "step": 9267
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.584124054983703e-05,
+      "loss": 0.9543,
+      "step": 9268
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5840326006110925e-05,
+      "loss": 1.1145,
+      "step": 9269
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583941138824392e-05,
+      "loss": 1.0419,
+      "step": 9270
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583849669624763e-05,
+      "loss": 0.9787,
+      "step": 9271
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5837581930133666e-05,
+      "loss": 0.9599,
+      "step": 9272
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5836667089913635e-05,
+      "loss": 0.904,
+      "step": 9273
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5835752175599163e-05,
+      "loss": 1.0645,
+      "step": 9274
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.583483718720185e-05,
+      "loss": 0.9937,
+      "step": 9275
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833922124733324e-05,
+      "loss": 1.0352,
+      "step": 9276
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5833006988205193e-05,
+      "loss": 1.0384,
+      "step": 9277
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5832091777629076e-05,
+      "loss": 1.091,
+      "step": 9278
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5831176493016594e-05,
+      "loss": 1.1564,
+      "step": 9279
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5830261134379365e-05,
+      "loss": 1.0253,
+      "step": 9280
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5829345701729006e-05,
+      "loss": 1.0812,
+      "step": 9281
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5828430195077144e-05,
+      "loss": 0.9725,
+      "step": 9282
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5827514614435395e-05,
+      "loss": 1.0174,
+      "step": 9283
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.582659895981539e-05,
+      "loss": 1.085,
+      "step": 9284
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5825683231228743e-05,
+      "loss": 0.9527,
+      "step": 9285
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5824767428687086e-05,
+      "loss": 1.0054,
+      "step": 9286
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5823851552202038e-05,
+      "loss": 1.1437,
+      "step": 9287
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822935601785236e-05,
+      "loss": 1.0667,
+      "step": 9288
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5822019577448297e-05,
+      "loss": 0.9982,
+      "step": 9289
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5821103479202857e-05,
+      "loss": 1.1726,
+      "step": 9290
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5820187307060545e-05,
+      "loss": 1.0337,
+      "step": 9291
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581927106103299e-05,
+      "loss": 0.9586,
+      "step": 9292
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581835474113182e-05,
+      "loss": 1.052,
+      "step": 9293
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581743834736867e-05,
+      "loss": 1.2972,
+      "step": 9294
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5816521879755173e-05,
+      "loss": 1.0433,
+      "step": 9295
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5815605338302965e-05,
+      "loss": 1.0073,
+      "step": 9296
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581468872302368e-05,
+      "loss": 0.9347,
+      "step": 9297
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5813772033928953e-05,
+      "loss": 1.0211,
+      "step": 9298
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5812855271030424e-05,
+      "loss": 0.9794,
+      "step": 9299
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.581193843433973e-05,
+      "loss": 1.0137,
+      "step": 9300
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5811021523868508e-05,
+      "loss": 1.1017,
+      "step": 9301
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5810104539628398e-05,
+      "loss": 1.1123,
+      "step": 9302
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5809187481631044e-05,
+      "loss": 1.0138,
+      "step": 9303
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580827034988808e-05,
+      "loss": 0.9918,
+      "step": 9304
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580735314441116e-05,
+      "loss": 1.085,
+      "step": 9305
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5806435865211918e-05,
+      "loss": 1.0643,
+      "step": 9306
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580551851230201e-05,
+      "loss": 1.124,
+      "step": 9307
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5804601085693067e-05,
+      "loss": 1.0912,
+      "step": 9308
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.580368358539674e-05,
+      "loss": 1.0073,
+      "step": 9309
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5802766011424683e-05,
+      "loss": 0.8704,
+      "step": 9310
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5801848363788538e-05,
+      "loss": 0.9507,
+      "step": 9311
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800930642499954e-05,
+      "loss": 1.0267,
+      "step": 9312
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5800012847570587e-05,
+      "loss": 1.0977,
+      "step": 9313
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579909497901208e-05,
+      "loss": 0.9528,
+      "step": 9314
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5798177036836092e-05,
+      "loss": 0.9501,
+      "step": 9315
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579725902105427e-05,
+      "loss": 0.9395,
+      "step": 9316
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5796340931678277e-05,
+      "loss": 1.1372,
+      "step": 9317
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579542276871976e-05,
+      "loss": 0.9218,
+      "step": 9318
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5794504532190374e-05,
+      "loss": 1.0594,
+      "step": 9319
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579358622210178e-05,
+      "loss": 1.0039,
+      "step": 9320
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5792667838465634e-05,
+      "loss": 0.9973,
+      "step": 9321
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5791749381293593e-05,
+      "loss": 1.0681,
+      "step": 9322
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.579083085059732e-05,
+      "loss": 0.999,
+      "step": 9323
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5789912246388473e-05,
+      "loss": 1.1013,
+      "step": 9324
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5788993568678715e-05,
+      "loss": 1.0478,
+      "step": 9325
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5788074817479705e-05,
+      "loss": 0.89,
+      "step": 9326
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5787155992803106e-05,
+      "loss": 0.9976,
+      "step": 9327
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578623709466059e-05,
+      "loss": 1.0265,
+      "step": 9328
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5785318123063813e-05,
+      "loss": 1.1407,
+      "step": 9329
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5784399078024448e-05,
+      "loss": 0.9563,
+      "step": 9330
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5783479959554154e-05,
+      "loss": 0.9771,
+      "step": 9331
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.578256076766461e-05,
+      "loss": 0.9555,
+      "step": 9332
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5781641502367472e-05,
+      "loss": 0.9696,
+      "step": 9333
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5780722163674416e-05,
+      "loss": 1.0597,
+      "step": 9334
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5779802751597118e-05,
+      "loss": 1.2041,
+      "step": 9335
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577888326614724e-05,
+      "loss": 1.0709,
+      "step": 9336
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5777963707336458e-05,
+      "loss": 1.188,
+      "step": 9337
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577704407517645e-05,
+      "loss": 1.0724,
+      "step": 9338
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5776124369678885e-05,
+      "loss": 0.9472,
+      "step": 9339
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5775204590855437e-05,
+      "loss": 0.9289,
+      "step": 9340
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.577428473871779e-05,
+      "loss": 1.0737,
+      "step": 9341
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 1.5773364813277616e-05,
+      "loss": 1.0141,
+      "step": 9342
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5772444814546593e-05,
+      "loss": 1.1524,
+      "step": 9343
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57715247425364e-05,
+      "loss": 1.1273,
+      "step": 9344
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5770604597258715e-05,
+      "loss": 0.8577,
+      "step": 9345
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5769684378725228e-05,
+      "loss": 1.0134,
+      "step": 9346
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5768764086947612e-05,
+      "loss": 1.0753,
+      "step": 9347
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576784372193755e-05,
+      "loss": 0.9254,
+      "step": 9348
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576692328370673e-05,
+      "loss": 1.0444,
+      "step": 9349
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5766002772266836e-05,
+      "loss": 0.9592,
+      "step": 9350
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576508218762955e-05,
+      "loss": 1.0979,
+      "step": 9351
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.576416152980656e-05,
+      "loss": 1.0786,
+      "step": 9352
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5763240798809557e-05,
+      "loss": 1.0542,
+      "step": 9353
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5762319994650222e-05,
+      "loss": 1.1174,
+      "step": 9354
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5761399117340255e-05,
+      "loss": 1.0371,
+      "step": 9355
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5760478166891338e-05,
+      "loss": 1.0433,
+      "step": 9356
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5759557143315162e-05,
+      "loss": 1.0558,
+      "step": 9357
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5758636046623425e-05,
+      "loss": 1.0309,
+      "step": 9358
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575771487682781e-05,
+      "loss": 1.0405,
+      "step": 9359
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5756793633940022e-05,
+      "loss": 1.067,
+      "step": 9360
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575587231797175e-05,
+      "loss": 1.0498,
+      "step": 9361
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575495092893469e-05,
+      "loss": 0.9939,
+      "step": 9362
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575402946684054e-05,
+      "loss": 1.0048,
+      "step": 9363
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5753107931700998e-05,
+      "loss": 1.053,
+      "step": 9364
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575218632352776e-05,
+      "loss": 0.985,
+      "step": 9365
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.575126464233253e-05,
+      "loss": 0.9791,
+      "step": 9366
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5750342888127e-05,
+      "loss": 1.0557,
+      "step": 9367
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574942106092288e-05,
+      "loss": 1.0295,
+      "step": 9368
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5748499160731875e-05,
+      "loss": 1.1169,
+      "step": 9369
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5747577187565677e-05,
+      "loss": 1.0225,
+      "step": 9370
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5746655141435996e-05,
+      "loss": 1.0775,
+      "step": 9371
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5745733022354535e-05,
+      "loss": 1.0803,
+      "step": 9372
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5744810830333004e-05,
+      "loss": 1.0085,
+      "step": 9373
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574388856538311e-05,
+      "loss": 1.0352,
+      "step": 9374
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574296622751655e-05,
+      "loss": 1.0114,
+      "step": 9375
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.574204381674505e-05,
+      "loss": 1.0856,
+      "step": 9376
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5741121333080306e-05,
+      "loss": 0.973,
+      "step": 9377
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5740198776534036e-05,
+      "loss": 1.0805,
+      "step": 9378
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573927614711795e-05,
+      "loss": 1.0418,
+      "step": 9379
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5738353444843753e-05,
+      "loss": 1.002,
+      "step": 9380
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573743066972317e-05,
+      "loss": 1.0604,
+      "step": 9381
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573650782176791e-05,
+      "loss": 0.9544,
+      "step": 9382
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5735584900989683e-05,
+      "loss": 0.9624,
+      "step": 9383
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573466190740022e-05,
+      "loss": 0.9754,
+      "step": 9384
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573373884101122e-05,
+      "loss": 1.1574,
+      "step": 9385
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5732815701834414e-05,
+      "loss": 1.0176,
+      "step": 9386
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5731892489881514e-05,
+      "loss": 1.0208,
+      "step": 9387
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5730969205164245e-05,
+      "loss": 1.1176,
+      "step": 9388
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.573004584769432e-05,
+      "loss": 1.0396,
+      "step": 9389
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5729122417483473e-05,
+      "loss": 1.0372,
+      "step": 9390
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5728198914543415e-05,
+      "loss": 0.9839,
+      "step": 9391
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5727275338885874e-05,
+      "loss": 0.9276,
+      "step": 9392
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5726351690522575e-05,
+      "loss": 1.033,
+      "step": 9393
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5725427969465245e-05,
+      "loss": 0.953,
+      "step": 9394
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5724504175725607e-05,
+      "loss": 1.0061,
+      "step": 9395
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5723580309315387e-05,
+      "loss": 1.0592,
+      "step": 9396
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5722656370246318e-05,
+      "loss": 1.0743,
+      "step": 9397
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5721732358530127e-05,
+      "loss": 1.1074,
+      "step": 9398
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5720808274178543e-05,
+      "loss": 0.9506,
+      "step": 9399
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.57198841172033e-05,
+      "loss": 1.0404,
+      "step": 9400
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5718959887616123e-05,
+      "loss": 1.0481,
+      "step": 9401
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571803558542875e-05,
+      "loss": 1.083,
+      "step": 9402
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5717111210652912e-05,
+      "loss": 1.0467,
+      "step": 9403
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.571618676330035e-05,
+      "loss": 1.0637,
+      "step": 9404
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5715262243382796e-05,
+      "loss": 1.0043,
+      "step": 9405
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5714337650911983e-05,
+      "loss": 1.0846,
+      "step": 9406
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5713412985899653e-05,
+      "loss": 0.9399,
+      "step": 9407
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5712488248357537e-05,
+      "loss": 0.9807,
+      "step": 9408
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5711563438297384e-05,
+      "loss": 0.9119,
+      "step": 9409
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5710638555730926e-05,
+      "loss": 1.1463,
+      "step": 9410
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570971360066991e-05,
+      "loss": 0.9145,
+      "step": 9411
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570878857312608e-05,
+      "loss": 1.1153,
+      "step": 9412
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5707863473111165e-05,
+      "loss": 0.9846,
+      "step": 9413
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706938300636924e-05,
+      "loss": 0.9101,
+      "step": 9414
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5706013055715094e-05,
+      "loss": 1.0971,
+      "step": 9415
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5705087738357425e-05,
+      "loss": 1.0209,
+      "step": 9416
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570416234857566e-05,
+      "loss": 1.0709,
+      "step": 9417
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.570323688638155e-05,
+      "loss": 1.1216,
+      "step": 9418
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5702311351786837e-05,
+      "loss": 1.0252,
+      "step": 9419
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5701385744803278e-05,
+      "loss": 0.9339,
+      "step": 9420
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5700460065442616e-05,
+      "loss": 1.0278,
+      "step": 9421
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569953431371661e-05,
+      "loss": 1.0639,
+      "step": 9422
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5698608489637005e-05,
+      "loss": 0.9541,
+      "step": 9423
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569768259321556e-05,
+      "loss": 1.0596,
+      "step": 9424
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5696756624464024e-05,
+      "loss": 0.9814,
+      "step": 9425
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5695830583394158e-05,
+      "loss": 1.0001,
+      "step": 9426
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.569490447001771e-05,
+      "loss": 1.0268,
+      "step": 9427
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693978284346442e-05,
+      "loss": 1.2026,
+      "step": 9428
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5693052026392103e-05,
+      "loss": 1.0434,
+      "step": 9429
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5692125696166466e-05,
+      "loss": 1.0276,
+      "step": 9430
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5691199293681282e-05,
+      "loss": 1.119,
+      "step": 9431
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5690272818948307e-05,
+      "loss": 1.1841,
+      "step": 9432
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5689346271979316e-05,
+      "loss": 1.0901,
+      "step": 9433
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568841965278606e-05,
+      "loss": 0.9697,
+      "step": 9434
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56874929613803e-05,
+      "loss": 1.0516,
+      "step": 9435
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5686566197773808e-05,
+      "loss": 1.0244,
+      "step": 9436
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5685639361978345e-05,
+      "loss": 0.9929,
+      "step": 9437
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568471245400568e-05,
+      "loss": 1.0467,
+      "step": 9438
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5683785473867574e-05,
+      "loss": 1.2673,
+      "step": 9439
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5682858421575803e-05,
+      "loss": 1.0605,
+      "step": 9440
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5681931297142127e-05,
+      "loss": 1.0322,
+      "step": 9441
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568100410057832e-05,
+      "loss": 1.0704,
+      "step": 9442
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.568007683189615e-05,
+      "loss": 1.1677,
+      "step": 9443
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5679149491107393e-05,
+      "loss": 1.0842,
+      "step": 9444
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5678222078223817e-05,
+      "loss": 1.0227,
+      "step": 9445
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5677294593257194e-05,
+      "loss": 1.1108,
+      "step": 9446
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5676367036219306e-05,
+      "loss": 1.0204,
+      "step": 9447
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5675439407121916e-05,
+      "loss": 1.1548,
+      "step": 9448
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5674511705976812e-05,
+      "loss": 1.0075,
+      "step": 9449
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567358393279576e-05,
+      "loss": 1.049,
+      "step": 9450
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567265608759055e-05,
+      "loss": 1.057,
+      "step": 9451
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.567172817037295e-05,
+      "loss": 1.1663,
+      "step": 9452
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5670800181154742e-05,
+      "loss": 0.9892,
+      "step": 9453
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566987211994771e-05,
+      "loss": 1.0286,
+      "step": 9454
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566894398676363e-05,
+      "loss": 1.009,
+      "step": 9455
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5668015781614293e-05,
+      "loss": 1.0533,
+      "step": 9456
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5667087504511474e-05,
+      "loss": 1.072,
+      "step": 9457
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566615915546696e-05,
+      "loss": 1.039,
+      "step": 9458
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5665230734492537e-05,
+      "loss": 0.9713,
+      "step": 9459
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5664302241599987e-05,
+      "loss": 1.1125,
+      "step": 9460
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5663373676801103e-05,
+      "loss": 0.9645,
+      "step": 9461
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5662445040107668e-05,
+      "loss": 1.0899,
+      "step": 9462
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5661516331531473e-05,
+      "loss": 1.0518,
+      "step": 9463
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.566058755108431e-05,
+      "loss": 1.0463,
+      "step": 9464
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5659658698777963e-05,
+      "loss": 1.1426,
+      "step": 9465
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565872977462423e-05,
+      "loss": 1.0673,
+      "step": 9466
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56578007786349e-05,
+      "loss": 1.0973,
+      "step": 9467
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5656871710821762e-05,
+      "loss": 1.0471,
+      "step": 9468
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5655942571196616e-05,
+      "loss": 0.984,
+      "step": 9469
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565501335977126e-05,
+      "loss": 0.9561,
+      "step": 9470
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5654084076557483e-05,
+      "loss": 0.9712,
+      "step": 9471
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5653154721567085e-05,
+      "loss": 1.1255,
+      "step": 9472
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5652225294811864e-05,
+      "loss": 1.1132,
+      "step": 9473
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5651295796303622e-05,
+      "loss": 1.1284,
+      "step": 9474
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.565036622605415e-05,
+      "loss": 1.1113,
+      "step": 9475
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564943658407526e-05,
+      "loss": 1.0303,
+      "step": 9476
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564850687037874e-05,
+      "loss": 1.0783,
+      "step": 9477
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5647577084976405e-05,
+      "loss": 0.9581,
+      "step": 9478
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564664722788005e-05,
+      "loss": 0.9657,
+      "step": 9479
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5645717299101484e-05,
+      "loss": 1.0923,
+      "step": 9480
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5644787298652512e-05,
+      "loss": 1.1758,
+      "step": 9481
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5643857226544935e-05,
+      "loss": 1.0525,
+      "step": 9482
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5642927082790565e-05,
+      "loss": 1.0088,
+      "step": 9483
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.564199686740121e-05,
+      "loss": 0.9825,
+      "step": 9484
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5641066580388668e-05,
+      "loss": 1.1814,
+      "step": 9485
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5640136221764767e-05,
+      "loss": 1.162,
+      "step": 9486
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.56392057915413e-05,
+      "loss": 1.1137,
+      "step": 9487
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5638275289730094e-05,
+      "loss": 1.0871,
+      "step": 9488
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563734471634295e-05,
+      "loss": 1.0136,
+      "step": 9489
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5636414071391687e-05,
+      "loss": 1.1597,
+      "step": 9490
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563548335488812e-05,
+      "loss": 0.9888,
+      "step": 9491
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5634552566844057e-05,
+      "loss": 1.1185,
+      "step": 9492
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563362170727132e-05,
+      "loss": 0.9815,
+      "step": 9493
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5632690776181724e-05,
+      "loss": 1.0613,
+      "step": 9494
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563175977358709e-05,
+      "loss": 1.0145,
+      "step": 9495
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.563082869949923e-05,
+      "loss": 1.0419,
+      "step": 9496
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5629897553929972e-05,
+      "loss": 1.0661,
+      "step": 9497
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562896633689113e-05,
+      "loss": 1.1058,
+      "step": 9498
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562803504839453e-05,
+      "loss": 0.8703,
+      "step": 9499
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562710368845199e-05,
+      "loss": 0.9801,
+      "step": 9500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5626172257075336e-05,
+      "loss": 0.9557,
+      "step": 9501
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5625240754276395e-05,
+      "loss": 1.0357,
+      "step": 9502
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5624309180066984e-05,
+      "loss": 1.0227,
+      "step": 9503
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5623377534458935e-05,
+      "loss": 1.0852,
+      "step": 9504
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562244581746408e-05,
+      "loss": 1.0462,
+      "step": 9505
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5621514029094228e-05,
+      "loss": 1.2155,
+      "step": 9506
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.562058216936123e-05,
+      "loss": 1.0496,
+      "step": 9507
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5619650238276906e-05,
+      "loss": 0.9983,
+      "step": 9508
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561871823585308e-05,
+      "loss": 1.0112,
+      "step": 9509
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5617786162101594e-05,
+      "loss": 1.0916,
+      "step": 9510
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5616854017034277e-05,
+      "loss": 1.0831,
+      "step": 9511
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5615921800662964e-05,
+      "loss": 1.1463,
+      "step": 9512
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5614989512999482e-05,
+      "loss": 1.0541,
+      "step": 9513
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561405715405567e-05,
+      "loss": 1.0257,
+      "step": 9514
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561312472384337e-05,
+      "loss": 0.8729,
+      "step": 9515
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.561219222237441e-05,
+      "loss": 1.1009,
+      "step": 9516
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5611259649660632e-05,
+      "loss": 1.1328,
+      "step": 9517
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5610327005713873e-05,
+      "loss": 0.9097,
+      "step": 9518
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5609394290545976e-05,
+      "loss": 1.0942,
+      "step": 9519
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560846150416878e-05,
+      "loss": 0.9584,
+      "step": 9520
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5607528646594122e-05,
+      "loss": 0.9089,
+      "step": 9521
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560659571783385e-05,
+      "loss": 1.0058,
+      "step": 9522
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5605662717899804e-05,
+      "loss": 0.9428,
+      "step": 9523
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5604729646803828e-05,
+      "loss": 1.0415,
+      "step": 9524
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560379650455777e-05,
+      "loss": 1.0706,
+      "step": 9525
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.560286329117348e-05,
+      "loss": 0.9437,
+      "step": 9526
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5601930006662787e-05,
+      "loss": 0.9306,
+      "step": 9527
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600996651037557e-05,
+      "loss": 1.0263,
+      "step": 9528
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5600063224309634e-05,
+      "loss": 1.0998,
+      "step": 9529
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5599129726490862e-05,
+      "loss": 1.1802,
+      "step": 9530
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5598196157593097e-05,
+      "loss": 1.0587,
+      "step": 9531
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559726251762819e-05,
+      "loss": 1.029,
+      "step": 9532
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559632880660799e-05,
+      "loss": 1.009,
+      "step": 9533
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559539502454435e-05,
+      "loss": 0.9734,
+      "step": 9534
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559446117144913e-05,
+      "loss": 0.981,
+      "step": 9535
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559352724733418e-05,
+      "loss": 1.0252,
+      "step": 9536
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5592593252211356e-05,
+      "loss": 1.1275,
+      "step": 9537
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.559165918609251e-05,
+      "loss": 0.9271,
+      "step": 9538
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5590725048989514e-05,
+      "loss": 1.0049,
+      "step": 9539
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5589790840914215e-05,
+      "loss": 1.0014,
+      "step": 9540
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5588856561878474e-05,
+      "loss": 1.0927,
+      "step": 9541
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5587922211894152e-05,
+      "loss": 1.0704,
+      "step": 9542
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586987790973108e-05,
+      "loss": 1.0765,
+      "step": 9543
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5586053299127213e-05,
+      "loss": 1.1083,
+      "step": 9544
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.558511873636832e-05,
+      "loss": 0.9717,
+      "step": 9545
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55841841027083e-05,
+      "loss": 0.9602,
+      "step": 9546
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5583249398159012e-05,
+      "loss": 0.9247,
+      "step": 9547
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5582314622732324e-05,
+      "loss": 0.9048,
+      "step": 9548
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5581379776440106e-05,
+      "loss": 1.1022,
+      "step": 9549
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5580444859294218e-05,
+      "loss": 1.0502,
+      "step": 9550
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5579509871306537e-05,
+      "loss": 1.0127,
+      "step": 9551
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5578574812488927e-05,
+      "loss": 1.042,
+      "step": 9552
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557763968285326e-05,
+      "loss": 0.9779,
+      "step": 9553
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5576704482411407e-05,
+      "loss": 1.0446,
+      "step": 9554
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5575769211175237e-05,
+      "loss": 1.0081,
+      "step": 9555
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.557483386915663e-05,
+      "loss": 0.9263,
+      "step": 9556
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5573898456367452e-05,
+      "loss": 1.1437,
+      "step": 9557
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5572962972819587e-05,
+      "loss": 1.1337,
+      "step": 9558
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.55720274185249e-05,
+      "loss": 1.0691,
+      "step": 9559
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5571091793495274e-05,
+      "loss": 1.1299,
+      "step": 9560
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5570156097742586e-05,
+      "loss": 1.0172,
+      "step": 9561
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556922033127871e-05,
+      "loss": 0.9536,
+      "step": 9562
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5568284494115532e-05,
+      "loss": 1.0776,
+      "step": 9563
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5567348586264927e-05,
+      "loss": 0.9981,
+      "step": 9564
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5566412607738777e-05,
+      "loss": 1.1634,
+      "step": 9565
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5565476558548968e-05,
+      "loss": 1.0882,
+      "step": 9566
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556454043870738e-05,
+      "loss": 1.1041,
+      "step": 9567
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556360424822589e-05,
+      "loss": 1.1333,
+      "step": 9568
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556266798711639e-05,
+      "loss": 1.1144,
+      "step": 9569
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.556173165539077e-05,
+      "loss": 1.073,
+      "step": 9570
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5560795253060907e-05,
+      "loss": 1.0614,
+      "step": 9571
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555985878013869e-05,
+      "loss": 0.9144,
+      "step": 9572
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5558922236636012e-05,
+      "loss": 1.0094,
+      "step": 9573
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555798562256476e-05,
+      "loss": 0.9761,
+      "step": 9574
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555704893793682e-05,
+      "loss": 1.0084,
+      "step": 9575
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5556112182764087e-05,
+      "loss": 0.9807,
+      "step": 9576
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5555175357058453e-05,
+      "loss": 0.9621,
+      "step": 9577
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5554238460831807e-05,
+      "loss": 1.0373,
+      "step": 9578
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5553301494096048e-05,
+      "loss": 1.0337,
+      "step": 9579
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555236445686307e-05,
+      "loss": 1.2215,
+      "step": 9580
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.555142734914476e-05,
+      "loss": 0.9814,
+      "step": 9581
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5550490170953023e-05,
+      "loss": 1.0511,
+      "step": 9582
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5549552922299756e-05,
+      "loss": 1.1162,
+      "step": 9583
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5548615603196854e-05,
+      "loss": 1.0452,
+      "step": 9584
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554767821365621e-05,
+      "loss": 1.0674,
+      "step": 9585
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5546740753689738e-05,
+      "loss": 1.0464,
+      "step": 9586
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5545803223309324e-05,
+      "loss": 0.9999,
+      "step": 9587
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554486562252688e-05,
+      "loss": 0.9124,
+      "step": 9588
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5543927951354304e-05,
+      "loss": 1.2531,
+      "step": 9589
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542990209803504e-05,
+      "loss": 1.0392,
+      "step": 9590
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5542052397886378e-05,
+      "loss": 0.8854,
+      "step": 9591
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554111451561483e-05,
+      "loss": 0.9119,
+      "step": 9592
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.554017656300078e-05,
+      "loss": 0.9895,
+      "step": 9593
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5539238540056117e-05,
+      "loss": 1.0739,
+      "step": 9594
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553830044679276e-05,
+      "loss": 0.9032,
+      "step": 9595
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5537362283222614e-05,
+      "loss": 1.0154,
+      "step": 9596
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5536424049357587e-05,
+      "loss": 1.105,
+      "step": 9597
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5535485745209595e-05,
+      "loss": 1.0028,
+      "step": 9598
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5534547370790547e-05,
+      "loss": 1.0278,
+      "step": 9599
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5533608926112355e-05,
+      "loss": 0.9745,
+      "step": 9600
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5532670411186926e-05,
+      "loss": 1.105,
+      "step": 9601
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5531731826026185e-05,
+      "loss": 1.0777,
+      "step": 9602
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.553079317064204e-05,
+      "loss": 1.0801,
+      "step": 9603
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552985444504641e-05,
+      "loss": 1.0466,
+      "step": 9604
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5528915649251212e-05,
+      "loss": 1.0878,
+      "step": 9605
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552797678326836e-05,
+      "loss": 1.0918,
+      "step": 9606
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5527037847109777e-05,
+      "loss": 0.9551,
+      "step": 9607
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552609884078738e-05,
+      "loss": 1.1382,
+      "step": 9608
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5525159764313088e-05,
+      "loss": 1.0472,
+      "step": 9609
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5524220617698822e-05,
+      "loss": 1.0245,
+      "step": 9610
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5523281400956512e-05,
+      "loss": 1.091,
+      "step": 9611
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5522342114098076e-05,
+      "loss": 1.0593,
+      "step": 9612
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552140275713543e-05,
+      "loss": 1.0248,
+      "step": 9613
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.552046333008051e-05,
+      "loss": 1.0153,
+      "step": 9614
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5519523832945237e-05,
+      "loss": 0.8878,
+      "step": 9615
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5518584265741537e-05,
+      "loss": 1.106,
+      "step": 9616
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5517644628481343e-05,
+      "loss": 1.0635,
+      "step": 9617
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5516704921176575e-05,
+      "loss": 1.0495,
+      "step": 9618
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5515765143839167e-05,
+      "loss": 1.0627,
+      "step": 9619
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5514825296481048e-05,
+      "loss": 1.0526,
+      "step": 9620
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5513885379114152e-05,
+      "loss": 1.0262,
+      "step": 9621
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512945391750405e-05,
+      "loss": 1.1087,
+      "step": 9622
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5512005334401747e-05,
+      "loss": 1.1266,
+      "step": 9623
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5511065207080103e-05,
+      "loss": 1.0696,
+      "step": 9624
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5510125009797418e-05,
+      "loss": 1.041,
+      "step": 9625
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.550918474256562e-05,
+      "loss": 1.0011,
+      "step": 9626
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5508244405396645e-05,
+      "loss": 0.9901,
+      "step": 9627
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5507303998302435e-05,
+      "loss": 1.1354,
+      "step": 9628
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5506363521294925e-05,
+      "loss": 1.0664,
+      "step": 9629
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 1.5505422974386054e-05,
+      "loss": 1.1327,
+      "step": 9630
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5504482357587764e-05,
+      "loss": 0.9932,
+      "step": 9631
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5503541670911996e-05,
+      "loss": 0.9696,
+      "step": 9632
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.550260091437069e-05,
+      "loss": 1.0575,
+      "step": 9633
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5501660087975784e-05,
+      "loss": 1.0015,
+      "step": 9634
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5500719191739227e-05,
+      "loss": 1.0212,
+      "step": 9635
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549977822567297e-05,
+      "loss": 1.0125,
+      "step": 9636
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5498837189788942e-05,
+      "loss": 1.0866,
+      "step": 9637
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54978960840991e-05,
+      "loss": 1.0528,
+      "step": 9638
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5496954908615395e-05,
+      "loss": 1.0481,
+      "step": 9639
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5496013663349764e-05,
+      "loss": 1.0499,
+      "step": 9640
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549507234831416e-05,
+      "loss": 0.994,
+      "step": 9641
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5494130963520537e-05,
+      "loss": 0.8864,
+      "step": 9642
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5493189508980833e-05,
+      "loss": 1.202,
+      "step": 9643
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.549224798470702e-05,
+      "loss": 1.1366,
+      "step": 9644
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5491306390711028e-05,
+      "loss": 0.9651,
+      "step": 9645
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5490364727004826e-05,
+      "loss": 0.8831,
+      "step": 9646
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5489422993600365e-05,
+      "loss": 0.9979,
+      "step": 9647
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5488481190509596e-05,
+      "loss": 0.979,
+      "step": 9648
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5487539317744475e-05,
+      "loss": 1.018,
+      "step": 9649
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5486597375316962e-05,
+      "loss": 1.0378,
+      "step": 9650
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548565536323901e-05,
+      "loss": 0.9906,
+      "step": 9651
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5484713281522584e-05,
+      "loss": 0.9998,
+      "step": 9652
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548377113017964e-05,
+      "loss": 1.1457,
+      "step": 9653
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5482828909222137e-05,
+      "loss": 1.0076,
+      "step": 9654
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.548188661866204e-05,
+      "loss": 0.9745,
+      "step": 9655
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5480944258511304e-05,
+      "loss": 1.0113,
+      "step": 9656
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54800018287819e-05,
+      "loss": 1.024,
+      "step": 9657
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5479059329485788e-05,
+      "loss": 1.0801,
+      "step": 9658
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547811676063493e-05,
+      "loss": 1.2232,
+      "step": 9659
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5477174122241295e-05,
+      "loss": 1.0472,
+      "step": 9660
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5476231414316853e-05,
+      "loss": 0.9196,
+      "step": 9661
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5475288636873568e-05,
+      "loss": 1.1287,
+      "step": 9662
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5474345789923402e-05,
+      "loss": 1.0529,
+      "step": 9663
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547340287347833e-05,
+      "loss": 0.9822,
+      "step": 9664
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5472459887550327e-05,
+      "loss": 1.0422,
+      "step": 9665
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5471516832151354e-05,
+      "loss": 1.0698,
+      "step": 9666
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.547057370729339e-05,
+      "loss": 1.0946,
+      "step": 9667
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5469630512988402e-05,
+      "loss": 1.0432,
+      "step": 9668
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546868724924837e-05,
+      "loss": 0.9454,
+      "step": 9669
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5467743916085264e-05,
+      "loss": 1.038,
+      "step": 9670
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5466800513511057e-05,
+      "loss": 0.9734,
+      "step": 9671
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546585704153773e-05,
+      "loss": 1.0571,
+      "step": 9672
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546491350017726e-05,
+      "loss": 1.0046,
+      "step": 9673
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463969889441622e-05,
+      "loss": 1.033,
+      "step": 9674
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5463026209342796e-05,
+      "loss": 1.0783,
+      "step": 9675
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.546208245989276e-05,
+      "loss": 1.0438,
+      "step": 9676
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54611386411035e-05,
+      "loss": 1.0115,
+      "step": 9677
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5460194752986993e-05,
+      "loss": 1.1325,
+      "step": 9678
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545925079555522e-05,
+      "loss": 0.992,
+      "step": 9679
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545830676882017e-05,
+      "loss": 1.1356,
+      "step": 9680
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545736267279382e-05,
+      "loss": 1.0812,
+      "step": 9681
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545641850748816e-05,
+      "loss": 1.0642,
+      "step": 9682
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5455474272915175e-05,
+      "loss": 1.0752,
+      "step": 9683
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5454529969086846e-05,
+      "loss": 1.0338,
+      "step": 9684
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545358559601517e-05,
+      "loss": 1.0755,
+      "step": 9685
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545264115371213e-05,
+      "loss": 1.0066,
+      "step": 9686
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545169664218972e-05,
+      "loss": 0.9334,
+      "step": 9687
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.545075206145992e-05,
+      "loss": 1.1156,
+      "step": 9688
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5449807411534737e-05,
+      "loss": 1.0579,
+      "step": 9689
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5448862692426144e-05,
+      "loss": 1.0656,
+      "step": 9690
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544791790414615e-05,
+      "loss": 0.9558,
+      "step": 9691
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5446973046706738e-05,
+      "loss": 0.984,
+      "step": 9692
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544602812011991e-05,
+      "loss": 0.982,
+      "step": 9693
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544508312439766e-05,
+      "loss": 1.0194,
+      "step": 9694
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544413805955198e-05,
+      "loss": 1.0648,
+      "step": 9695
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5443192925594873e-05,
+      "loss": 1.0638,
+      "step": 9696
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5442247722538333e-05,
+      "loss": 1.0543,
+      "step": 9697
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.544130245039436e-05,
+      "loss": 1.1107,
+      "step": 9698
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5440357109174952e-05,
+      "loss": 1.0306,
+      "step": 9699
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5439411698892116e-05,
+      "loss": 1.1194,
+      "step": 9700
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5438466219557847e-05,
+      "loss": 1.0433,
+      "step": 9701
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5437520671184154e-05,
+      "loss": 1.0805,
+      "step": 9702
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5436575053783033e-05,
+      "loss": 1.057,
+      "step": 9703
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5435629367366495e-05,
+      "loss": 1.0408,
+      "step": 9704
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5434683611946542e-05,
+      "loss": 0.9808,
+      "step": 9705
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5433737787535177e-05,
+      "loss": 1.0993,
+      "step": 9706
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5432791894144413e-05,
+      "loss": 1.1061,
+      "step": 9707
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.543184593178625e-05,
+      "loss": 0.9864,
+      "step": 9708
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5430899900472705e-05,
+      "loss": 1.0558,
+      "step": 9709
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5429953800215787e-05,
+      "loss": 1.045,
+      "step": 9710
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54290076310275e-05,
+      "loss": 1.0094,
+      "step": 9711
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542806139291986e-05,
+      "loss": 1.1426,
+      "step": 9712
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5427115085904876e-05,
+      "loss": 1.073,
+      "step": 9713
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5426168709994564e-05,
+      "loss": 1.0873,
+      "step": 9714
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542522226520094e-05,
+      "loss": 1.1346,
+      "step": 9715
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5424275751536012e-05,
+      "loss": 1.1068,
+      "step": 9716
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54233291690118e-05,
+      "loss": 1.0411,
+      "step": 9717
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542238251764032e-05,
+      "loss": 1.0089,
+      "step": 9718
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.542143579743359e-05,
+      "loss": 1.0402,
+      "step": 9719
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5420489008403626e-05,
+      "loss": 1.0613,
+      "step": 9720
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541954215056245e-05,
+      "loss": 0.9547,
+      "step": 9721
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541859522392208e-05,
+      "loss": 0.9854,
+      "step": 9722
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5417648228494538e-05,
+      "loss": 0.9575,
+      "step": 9723
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5416701164291842e-05,
+      "loss": 1.136,
+      "step": 9724
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5415754031326026e-05,
+      "loss": 1.0408,
+      "step": 9725
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54148068296091e-05,
+      "loss": 0.9417,
+      "step": 9726
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5413859559153092e-05,
+      "loss": 0.9818,
+      "step": 9727
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5412912219970034e-05,
+      "loss": 0.9574,
+      "step": 9728
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411964812071946e-05,
+      "loss": 1.1073,
+      "step": 9729
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5411017335470857e-05,
+      "loss": 0.9476,
+      "step": 9730
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.541006979017879e-05,
+      "loss": 1.0305,
+      "step": 9731
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540912217620778e-05,
+      "loss": 1.0602,
+      "step": 9732
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5408174493569857e-05,
+      "loss": 1.0371,
+      "step": 9733
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5407226742277047e-05,
+      "loss": 1.2433,
+      "step": 9734
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5406278922341388e-05,
+      "loss": 0.9257,
+      "step": 9735
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.54053310337749e-05,
+      "loss": 0.9633,
+      "step": 9736
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5404383076589632e-05,
+      "loss": 0.9864,
+      "step": 9737
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5403435050797604e-05,
+      "loss": 0.9169,
+      "step": 9738
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.540248695641086e-05,
+      "loss": 1.0778,
+      "step": 9739
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5401538793441432e-05,
+      "loss": 0.9743,
+      "step": 9740
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5400590561901358e-05,
+      "loss": 0.9635,
+      "step": 9741
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539964226180267e-05,
+      "loss": 1.1377,
+      "step": 9742
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539869389315741e-05,
+      "loss": 1.0555,
+      "step": 9743
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5397745455977625e-05,
+      "loss": 1.1482,
+      "step": 9744
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5396796950275344e-05,
+      "loss": 0.951,
+      "step": 9745
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5395848376062615e-05,
+      "loss": 1.0852,
+      "step": 9746
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5394899733351474e-05,
+      "loss": 1.142,
+      "step": 9747
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539395102215397e-05,
+      "loss": 1.0666,
+      "step": 9748
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5393002242482136e-05,
+      "loss": 1.062,
+      "step": 9749
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539205339434803e-05,
+      "loss": 1.069,
+      "step": 9750
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539110447776369e-05,
+      "loss": 0.9495,
+      "step": 9751
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.539015549274116e-05,
+      "loss": 1.1314,
+      "step": 9752
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538920643929249e-05,
+      "loss": 1.0663,
+      "step": 9753
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5388257317429736e-05,
+      "loss": 1.0708,
+      "step": 9754
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538730812716493e-05,
+      "loss": 1.0716,
+      "step": 9755
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5386358868510132e-05,
+      "loss": 1.0834,
+      "step": 9756
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5385409541477394e-05,
+      "loss": 1.0059,
+      "step": 9757
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5384460146078762e-05,
+      "loss": 1.0312,
+      "step": 9758
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538351068232629e-05,
+      "loss": 0.9597,
+      "step": 9759
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5382561150232038e-05,
+      "loss": 1.0347,
+      "step": 9760
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5381611549808048e-05,
+      "loss": 0.9789,
+      "step": 9761
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.538066188106638e-05,
+      "loss": 1.0724,
+      "step": 9762
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537971214401909e-05,
+      "loss": 1.0123,
+      "step": 9763
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537876233867823e-05,
+      "loss": 1.1451,
+      "step": 9764
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5377812465055874e-05,
+      "loss": 1.0049,
+      "step": 9765
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537686252316406e-05,
+      "loss": 1.1801,
+      "step": 9766
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5375912513014858e-05,
+      "loss": 1.1082,
+      "step": 9767
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5374962434620327e-05,
+      "loss": 1.031,
+      "step": 9768
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5374012287992525e-05,
+      "loss": 0.9233,
+      "step": 9769
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5373062073143514e-05,
+      "loss": 0.9731,
+      "step": 9770
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537211179008536e-05,
+      "loss": 1.1281,
+      "step": 9771
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5371161438830124e-05,
+      "loss": 0.9914,
+      "step": 9772
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.537021101938987e-05,
+      "loss": 1.0303,
+      "step": 9773
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5369260531776665e-05,
+      "loss": 1.0264,
+      "step": 9774
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5368309976002578e-05,
+      "loss": 0.8723,
+      "step": 9775
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5367359352079666e-05,
+      "loss": 0.9166,
+      "step": 9776
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536640866002001e-05,
+      "loss": 1.0471,
+      "step": 9777
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5365457899835666e-05,
+      "loss": 1.0657,
+      "step": 9778
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5364507071538713e-05,
+      "loss": 1.1385,
+      "step": 9779
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5363556175141212e-05,
+      "loss": 0.9716,
+      "step": 9780
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5362605210655245e-05,
+      "loss": 0.8751,
+      "step": 9781
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536165417809288e-05,
+      "loss": 0.9693,
+      "step": 9782
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.536070307746619e-05,
+      "loss": 1.1303,
+      "step": 9783
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5359751908787247e-05,
+      "loss": 0.9372,
+      "step": 9784
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5358800672068124e-05,
+      "loss": 1.1296,
+      "step": 9785
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5357849367320904e-05,
+      "loss": 1.0097,
+      "step": 9786
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5356897994557654e-05,
+      "loss": 1.0529,
+      "step": 9787
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535594655379046e-05,
+      "loss": 1.1039,
+      "step": 9788
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5354995045031396e-05,
+      "loss": 0.9825,
+      "step": 9789
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535404346829254e-05,
+      "loss": 0.8714,
+      "step": 9790
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5353091823585973e-05,
+      "loss": 1.0901,
+      "step": 9791
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.535214011092378e-05,
+      "loss": 1.074,
+      "step": 9792
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5351188330318035e-05,
+      "loss": 1.0479,
+      "step": 9793
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5350236481780827e-05,
+      "loss": 0.9329,
+      "step": 9794
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5349284565324238e-05,
+      "loss": 1.0364,
+      "step": 9795
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5348332580960353e-05,
+      "loss": 1.0269,
+      "step": 9796
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534738052870125e-05,
+      "loss": 1.0671,
+      "step": 9797
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5346428408559023e-05,
+      "loss": 1.0463,
+      "step": 9798
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5345476220545755e-05,
+      "loss": 1.0253,
+      "step": 9799
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5344523964673536e-05,
+      "loss": 0.9581,
+      "step": 9800
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5343571640954453e-05,
+      "loss": 1.0006,
+      "step": 9801
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5342619249400598e-05,
+      "loss": 0.9768,
+      "step": 9802
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5341666790024058e-05,
+      "loss": 1.0811,
+      "step": 9803
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.534071426283692e-05,
+      "loss": 1.044,
+      "step": 9804
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533976166785129e-05,
+      "loss": 1.0387,
+      "step": 9805
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5338809005079244e-05,
+      "loss": 1.0895,
+      "step": 9806
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5337856274532886e-05,
+      "loss": 1.04,
+      "step": 9807
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533690347622431e-05,
+      "loss": 1.128,
+      "step": 9808
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.533595061016561e-05,
+      "loss": 1.0196,
+      "step": 9809
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334997676368883e-05,
+      "loss": 1.0088,
+      "step": 9810
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5334044674846222e-05,
+      "loss": 1.0611,
+      "step": 9811
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5333091605609732e-05,
+      "loss": 0.9441,
+      "step": 9812
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5332138468671508e-05,
+      "loss": 1.052,
+      "step": 9813
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5331185264043645e-05,
+      "loss": 1.1754,
+      "step": 9814
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5330231991738253e-05,
+      "loss": 0.9932,
+      "step": 9815
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5329278651767427e-05,
+      "loss": 1.0975,
+      "step": 9816
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5328325244143273e-05,
+      "loss": 1.0944,
+      "step": 9817
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532737176887789e-05,
+      "loss": 1.0493,
+      "step": 9818
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5326418225983388e-05,
+      "loss": 1.0799,
+      "step": 9819
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5325464615471862e-05,
+      "loss": 0.9943,
+      "step": 9820
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532451093735543e-05,
+      "loss": 1.1,
+      "step": 9821
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532355719164619e-05,
+      "loss": 1.0569,
+      "step": 9822
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5322603378356254e-05,
+      "loss": 0.9976,
+      "step": 9823
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532164949749773e-05,
+      "loss": 1.0021,
+      "step": 9824
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.532069554908272e-05,
+      "loss": 1.0396,
+      "step": 9825
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5319741533123347e-05,
+      "loss": 1.0254,
+      "step": 9826
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531878744963171e-05,
+      "loss": 1.0102,
+      "step": 9827
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531783329861993e-05,
+      "loss": 0.9313,
+      "step": 9828
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531687908010011e-05,
+      "loss": 1.1762,
+      "step": 9829
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5315924794084368e-05,
+      "loss": 1.1,
+      "step": 9830
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531497044058482e-05,
+      "loss": 1.0955,
+      "step": 9831
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5314016019613583e-05,
+      "loss": 1.0075,
+      "step": 9832
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5313061531182768e-05,
+      "loss": 0.9222,
+      "step": 9833
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5312106975304495e-05,
+      "loss": 1.0001,
+      "step": 9834
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.531115235199088e-05,
+      "loss": 1.0549,
+      "step": 9835
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5310197661254044e-05,
+      "loss": 1.0724,
+      "step": 9836
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5309242903106102e-05,
+      "loss": 0.9654,
+      "step": 9837
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5308288077559176e-05,
+      "loss": 1.1223,
+      "step": 9838
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530733318462539e-05,
+      "loss": 0.9593,
+      "step": 9839
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5306378224316865e-05,
+      "loss": 1.0437,
+      "step": 9840
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5305423196645723e-05,
+      "loss": 0.987,
+      "step": 9841
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5304468101624087e-05,
+      "loss": 1.0086,
+      "step": 9842
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5303512939264084e-05,
+      "loss": 1.1175,
+      "step": 9843
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.530255770957784e-05,
+      "loss": 1.005,
+      "step": 9844
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5301602412577475e-05,
+      "loss": 1.0079,
+      "step": 9845
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5300647048275124e-05,
+      "loss": 1.0385,
+      "step": 9846
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5299691616682907e-05,
+      "loss": 1.0888,
+      "step": 9847
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5298736117812964e-05,
+      "loss": 1.0099,
+      "step": 9848
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529778055167741e-05,
+      "loss": 0.9177,
+      "step": 9849
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529682491828839e-05,
+      "loss": 1.108,
+      "step": 9850
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5295869217658027e-05,
+      "loss": 1.0578,
+      "step": 9851
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5294913449798453e-05,
+      "loss": 1.0323,
+      "step": 9852
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5293957614721807e-05,
+      "loss": 1.0771,
+      "step": 9853
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529300171244022e-05,
+      "loss": 0.9814,
+      "step": 9854
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5292045742965824e-05,
+      "loss": 1.1045,
+      "step": 9855
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.529108970631076e-05,
+      "loss": 1.1429,
+      "step": 9856
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5290133602487158e-05,
+      "loss": 1.0151,
+      "step": 9857
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5289177431507162e-05,
+      "loss": 0.9487,
+      "step": 9858
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5288221193382905e-05,
+      "loss": 1.0043,
+      "step": 9859
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5287264888126527e-05,
+      "loss": 1.0383,
+      "step": 9860
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528630851575017e-05,
+      "loss": 0.9868,
+      "step": 9861
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.528535207626598e-05,
+      "loss": 0.9604,
+      "step": 9862
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5284395569686085e-05,
+      "loss": 1.184,
+      "step": 9863
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5283438996022638e-05,
+      "loss": 1.0717,
+      "step": 9864
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5282482355287782e-05,
+      "loss": 0.9704,
+      "step": 9865
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5281525647493656e-05,
+      "loss": 1.0518,
+      "step": 9866
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5280568872652407e-05,
+      "loss": 1.0186,
+      "step": 9867
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5279612030776188e-05,
+      "loss": 1.1486,
+      "step": 9868
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5278655121877135e-05,
+      "loss": 1.0467,
+      "step": 9869
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52776981459674e-05,
+      "loss": 1.03,
+      "step": 9870
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5276741103059133e-05,
+      "loss": 0.8722,
+      "step": 9871
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5275783993164485e-05,
+      "loss": 1.0807,
+      "step": 9872
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5274826816295595e-05,
+      "loss": 1.0152,
+      "step": 9873
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.527386957246463e-05,
+      "loss": 0.962,
+      "step": 9874
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5272912261683735e-05,
+      "loss": 1.0571,
+      "step": 9875
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5271954883965057e-05,
+      "loss": 1.0999,
+      "step": 9876
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5270997439320755e-05,
+      "loss": 1.0843,
+      "step": 9877
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5270039927762984e-05,
+      "loss": 1.0389,
+      "step": 9878
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5269082349303902e-05,
+      "loss": 1.036,
+      "step": 9879
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5268124703955655e-05,
+      "loss": 1.0711,
+      "step": 9880
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526716699173041e-05,
+      "loss": 0.9404,
+      "step": 9881
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526620921264032e-05,
+      "loss": 1.0936,
+      "step": 9882
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526525136669754e-05,
+      "loss": 1.0751,
+      "step": 9883
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5264293453914243e-05,
+      "loss": 1.1586,
+      "step": 9884
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.526333547430258e-05,
+      "loss": 1.1109,
+      "step": 9885
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5262377427874707e-05,
+      "loss": 1.0447,
+      "step": 9886
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5261419314642793e-05,
+      "loss": 0.8957,
+      "step": 9887
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5260461134619005e-05,
+      "loss": 1.0976,
+      "step": 9888
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5259502887815497e-05,
+      "loss": 1.0827,
+      "step": 9889
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5258544574244438e-05,
+      "loss": 1.0224,
+      "step": 9890
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5257586193917994e-05,
+      "loss": 1.0881,
+      "step": 9891
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5256627746848334e-05,
+      "loss": 1.0385,
+      "step": 9892
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5255669233047621e-05,
+      "loss": 1.0048,
+      "step": 9893
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525471065252802e-05,
+      "loss": 1.1715,
+      "step": 9894
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.525375200530171e-05,
+      "loss": 0.872,
+      "step": 9895
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5252793291380852e-05,
+      "loss": 1.1371,
+      "step": 9896
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5251834510777616e-05,
+      "loss": 1.0054,
+      "step": 9897
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5250875663504184e-05,
+      "loss": 0.943,
+      "step": 9898
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5249916749572715e-05,
+      "loss": 0.9492,
+      "step": 9899
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524895776899539e-05,
+      "loss": 0.9579,
+      "step": 9900
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247998721784379e-05,
+      "loss": 0.9527,
+      "step": 9901
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5247039607951862e-05,
+      "loss": 1.0,
+      "step": 9902
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5246080427510009e-05,
+      "loss": 0.9991,
+      "step": 9903
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5245121180470996e-05,
+      "loss": 1.0628,
+      "step": 9904
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5244161866847007e-05,
+      "loss": 1.0984,
+      "step": 9905
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5243202486650215e-05,
+      "loss": 0.95,
+      "step": 9906
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.52422430398928e-05,
+      "loss": 1.0306,
+      "step": 9907
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.524128352658694e-05,
+      "loss": 1.1077,
+      "step": 9908
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5240323946744822e-05,
+      "loss": 0.9816,
+      "step": 9909
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5239364300378624e-05,
+      "loss": 1.0301,
+      "step": 9910
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5238404587500521e-05,
+      "loss": 1.0706,
+      "step": 9911
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5237444808122708e-05,
+      "loss": 0.9597,
+      "step": 9912
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5236484962257358e-05,
+      "loss": 0.9899,
+      "step": 9913
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5235525049916668e-05,
+      "loss": 0.9193,
+      "step": 9914
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5234565071112814e-05,
+      "loss": 1.0411,
+      "step": 9915
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5233605025857988e-05,
+      "loss": 1.1049,
+      "step": 9916
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 1.5232644914164377e-05,
+      "loss": 1.0608,
+      "step": 9917
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5231684736044163e-05,
+      "loss": 0.9725,
+      "step": 9918
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5230724491509543e-05,
+      "loss": 0.9886,
+      "step": 9919
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5229764180572704e-05,
+      "loss": 1.0198,
+      "step": 9920
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5228803803245835e-05,
+      "loss": 0.996,
+      "step": 9921
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522784335954113e-05,
+      "loss": 1.0153,
+      "step": 9922
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5226882849470783e-05,
+      "loss": 0.9795,
+      "step": 9923
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5225922273046982e-05,
+      "loss": 0.9743,
+      "step": 9924
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224961630281926e-05,
+      "loss": 1.009,
+      "step": 9925
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5224000921187809e-05,
+      "loss": 0.9716,
+      "step": 9926
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522304014577683e-05,
+      "loss": 0.8887,
+      "step": 9927
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5222079304061174e-05,
+      "loss": 1.073,
+      "step": 9928
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.522111839605305e-05,
+      "loss": 1.0684,
+      "step": 9929
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5220157421764655e-05,
+      "loss": 0.9741,
+      "step": 9930
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5219196381208183e-05,
+      "loss": 0.9151,
+      "step": 9931
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5218235274395839e-05,
+      "loss": 1.1333,
+      "step": 9932
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5217274101339823e-05,
+      "loss": 0.9673,
+      "step": 9933
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5216312862052334e-05,
+      "loss": 1.0159,
+      "step": 9934
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5215351556545577e-05,
+      "loss": 1.0159,
+      "step": 9935
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5214390184831756e-05,
+      "loss": 0.9699,
+      "step": 9936
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5213428746923074e-05,
+      "loss": 1.0174,
+      "step": 9937
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5212467242831734e-05,
+      "loss": 1.0738,
+      "step": 9938
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5211505672569944e-05,
+      "loss": 1.068,
+      "step": 9939
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5210544036149918e-05,
+      "loss": 1.1549,
+      "step": 9940
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.520958233358385e-05,
+      "loss": 0.9763,
+      "step": 9941
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5208620564883955e-05,
+      "loss": 0.935,
+      "step": 9942
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5207658730062444e-05,
+      "loss": 1.1773,
+      "step": 9943
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5206696829131528e-05,
+      "loss": 1.0681,
+      "step": 9944
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5205734862103414e-05,
+      "loss": 1.039,
+      "step": 9945
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5204772828990316e-05,
+      "loss": 0.9918,
+      "step": 9946
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5203810729804445e-05,
+      "loss": 1.0495,
+      "step": 9947
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5202848564558018e-05,
+      "loss": 1.025,
+      "step": 9948
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5201886333263246e-05,
+      "loss": 0.9188,
+      "step": 9949
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5200924035932342e-05,
+      "loss": 1.0292,
+      "step": 9950
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519996167257753e-05,
+      "loss": 1.008,
+      "step": 9951
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198999243211018e-05,
+      "loss": 1.0734,
+      "step": 9952
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5198036747845031e-05,
+      "loss": 1.1462,
+      "step": 9953
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5197074186491786e-05,
+      "loss": 1.0108,
+      "step": 9954
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5196111559163496e-05,
+      "loss": 1.0336,
+      "step": 9955
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5195148865872387e-05,
+      "loss": 1.1238,
+      "step": 9956
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519418610663068e-05,
+      "loss": 1.0328,
+      "step": 9957
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5193223281450598e-05,
+      "loss": 1.1633,
+      "step": 9958
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.519226039034436e-05,
+      "loss": 1.1105,
+      "step": 9959
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5191297433324188e-05,
+      "loss": 1.0425,
+      "step": 9960
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5190334410402314e-05,
+      "loss": 1.0109,
+      "step": 9961
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5189371321590957e-05,
+      "loss": 1.0121,
+      "step": 9962
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5188408166902347e-05,
+      "loss": 1.1575,
+      "step": 9963
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5187444946348704e-05,
+      "loss": 0.9539,
+      "step": 9964
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5186481659942267e-05,
+      "loss": 1.0787,
+      "step": 9965
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5185518307695253e-05,
+      "loss": 1.0473,
+      "step": 9966
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51845548896199e-05,
+      "loss": 0.9557,
+      "step": 9967
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5183591405728435e-05,
+      "loss": 1.079,
+      "step": 9968
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5182627856033087e-05,
+      "loss": 0.9621,
+      "step": 9969
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5181664240546089e-05,
+      "loss": 1.0021,
+      "step": 9970
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5180700559279675e-05,
+      "loss": 1.2113,
+      "step": 9971
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5179736812246082e-05,
+      "loss": 1.0123,
+      "step": 9972
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5178772999457537e-05,
+      "loss": 1.0569,
+      "step": 9973
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.517780912092628e-05,
+      "loss": 1.0874,
+      "step": 9974
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5176845176664547e-05,
+      "loss": 1.1057,
+      "step": 9975
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5175881166684572e-05,
+      "loss": 1.0486,
+      "step": 9976
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5174917090998595e-05,
+      "loss": 1.0217,
+      "step": 9977
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5173952949618854e-05,
+      "loss": 1.0451,
+      "step": 9978
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172988742557594e-05,
+      "loss": 1.0152,
+      "step": 9979
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5172024469827045e-05,
+      "loss": 1.0588,
+      "step": 9980
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5171060131439452e-05,
+      "loss": 1.0657,
+      "step": 9981
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5170095727407062e-05,
+      "loss": 0.9292,
+      "step": 9982
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516913125774211e-05,
+      "loss": 0.9916,
+      "step": 9983
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5168166722456846e-05,
+      "loss": 1.048,
+      "step": 9984
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516720212156351e-05,
+      "loss": 1.0157,
+      "step": 9985
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5166237455074353e-05,
+      "loss": 0.9415,
+      "step": 9986
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5165272723001613e-05,
+      "loss": 1.0844,
+      "step": 9987
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5164307925357544e-05,
+      "loss": 0.9759,
+      "step": 9988
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.516334306215439e-05,
+      "loss": 1.0653,
+      "step": 9989
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5162378133404403e-05,
+      "loss": 0.9485,
+      "step": 9990
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5161413139119824e-05,
+      "loss": 1.0277,
+      "step": 9991
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5160448079312915e-05,
+      "loss": 0.987,
+      "step": 9992
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5159482953995919e-05,
+      "loss": 1.1081,
+      "step": 9993
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515851776318109e-05,
+      "loss": 1.014,
+      "step": 9994
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5157552506880681e-05,
+      "loss": 0.9672,
+      "step": 9995
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5156587185106951e-05,
+      "loss": 0.957,
+      "step": 9996
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5155621797872143e-05,
+      "loss": 1.0953,
+      "step": 9997
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515465634518852e-05,
+      "loss": 1.121,
+      "step": 9998
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5153690827068336e-05,
+      "loss": 0.9672,
+      "step": 9999
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.515272524352385e-05,
+      "loss": 0.9908,
+      "step": 10000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5151759594567314e-05,
+      "loss": 0.9918,
+      "step": 10001
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5150793880210994e-05,
+      "loss": 1.0081,
+      "step": 10002
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.514982810046715e-05,
+      "loss": 1.1165,
+      "step": 10003
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5148862255348032e-05,
+      "loss": 1.0367,
+      "step": 10004
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5147896344865909e-05,
+      "loss": 0.9489,
+      "step": 10005
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5146930369033039e-05,
+      "loss": 0.9411,
+      "step": 10006
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5145964327861694e-05,
+      "loss": 0.9629,
+      "step": 10007
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144998221364122e-05,
+      "loss": 0.9123,
+      "step": 10008
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5144032049552602e-05,
+      "loss": 1.0274,
+      "step": 10009
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5143065812439393e-05,
+      "loss": 0.9531,
+      "step": 10010
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5142099510036757e-05,
+      "loss": 1.1029,
+      "step": 10011
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5141133142356967e-05,
+      "loss": 0.9601,
+      "step": 10012
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5140166709412292e-05,
+      "loss": 1.0767,
+      "step": 10013
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5139200211214993e-05,
+      "loss": 1.038,
+      "step": 10014
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5138233647777346e-05,
+      "loss": 0.9698,
+      "step": 10015
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513726701911162e-05,
+      "loss": 0.9581,
+      "step": 10016
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5136300325230082e-05,
+      "loss": 0.9986,
+      "step": 10017
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5135333566145005e-05,
+      "loss": 0.9895,
+      "step": 10018
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513436674186867e-05,
+      "loss": 0.9943,
+      "step": 10019
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5133399852413338e-05,
+      "loss": 1.05,
+      "step": 10020
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.513243289779129e-05,
+      "loss": 1.1509,
+      "step": 10021
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5131465878014802e-05,
+      "loss": 0.9878,
+      "step": 10022
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5130498793096149e-05,
+      "loss": 1.058,
+      "step": 10023
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5129531643047606e-05,
+      "loss": 1.033,
+      "step": 10024
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5128564427881452e-05,
+      "loss": 1.0393,
+      "step": 10025
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5127597147609964e-05,
+      "loss": 1.0863,
+      "step": 10026
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5126629802245422e-05,
+      "loss": 1.1465,
+      "step": 10027
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5125662391800109e-05,
+      "loss": 0.9593,
+      "step": 10028
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5124694916286301e-05,
+      "loss": 1.044,
+      "step": 10029
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5123727375716286e-05,
+      "loss": 0.9793,
+      "step": 10030
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5122759770102342e-05,
+      "loss": 1.0153,
+      "step": 10031
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.512179209945675e-05,
+      "loss": 1.0931,
+      "step": 10032
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51208243637918e-05,
+      "loss": 1.0185,
+      "step": 10033
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5119856563119775e-05,
+      "loss": 1.0889,
+      "step": 10034
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5118888697452959e-05,
+      "loss": 1.0173,
+      "step": 10035
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.511792076680364e-05,
+      "loss": 1.0892,
+      "step": 10036
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5116952771184108e-05,
+      "loss": 1.1336,
+      "step": 10037
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115984710606646e-05,
+      "loss": 1.0523,
+      "step": 10038
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5115016585083548e-05,
+      "loss": 1.0928,
+      "step": 10039
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5114048394627103e-05,
+      "loss": 1.1162,
+      "step": 10040
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.51130801392496e-05,
+      "loss": 1.0254,
+      "step": 10041
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5112111818963333e-05,
+      "loss": 0.9418,
+      "step": 10042
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5111143433780593e-05,
+      "loss": 1.0462,
+      "step": 10043
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5110174983713675e-05,
+      "loss": 0.9241,
+      "step": 10044
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5109206468774868e-05,
+      "loss": 1.0782,
+      "step": 10045
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5108237888976474e-05,
+      "loss": 1.0045,
+      "step": 10046
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5107269244330786e-05,
+      "loss": 1.0184,
+      "step": 10047
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5106300534850098e-05,
+      "loss": 0.8738,
+      "step": 10048
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5105331760546708e-05,
+      "loss": 1.0561,
+      "step": 10049
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5104362921432918e-05,
+      "loss": 0.9937,
+      "step": 10050
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5103394017521028e-05,
+      "loss": 1.0042,
+      "step": 10051
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5102425048823328e-05,
+      "loss": 1.0183,
+      "step": 10052
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5101456015352129e-05,
+      "loss": 0.9686,
+      "step": 10053
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5100486917119732e-05,
+      "loss": 0.9842,
+      "step": 10054
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5099517754138434e-05,
+      "loss": 0.9607,
+      "step": 10055
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5098548526420537e-05,
+      "loss": 1.0084,
+      "step": 10056
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5097579233978352e-05,
+      "loss": 0.9524,
+      "step": 10057
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5096609876824182e-05,
+      "loss": 1.0418,
+      "step": 10058
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5095640454970328e-05,
+      "loss": 1.1166,
+      "step": 10059
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5094670968429102e-05,
+      "loss": 1.0876,
+      "step": 10060
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5093701417212807e-05,
+      "loss": 1.1382,
+      "step": 10061
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5092731801333753e-05,
+      "loss": 1.0536,
+      "step": 10062
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.509176212080425e-05,
+      "loss": 1.0412,
+      "step": 10063
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5090792375636604e-05,
+      "loss": 1.0017,
+      "step": 10064
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5089822565843131e-05,
+      "loss": 1.2047,
+      "step": 10065
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508885269143614e-05,
+      "loss": 1.1131,
+      "step": 10066
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508788275242794e-05,
+      "loss": 1.1342,
+      "step": 10067
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5086912748830852e-05,
+      "loss": 0.8971,
+      "step": 10068
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5085942680657178e-05,
+      "loss": 0.9936,
+      "step": 10069
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5084972547919243e-05,
+      "loss": 0.9392,
+      "step": 10070
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5084002350629358e-05,
+      "loss": 1.1011,
+      "step": 10071
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5083032088799843e-05,
+      "loss": 1.0389,
+      "step": 10072
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508206176244301e-05,
+      "loss": 0.9783,
+      "step": 10073
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5081091371571178e-05,
+      "loss": 0.923,
+      "step": 10074
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.508012091619667e-05,
+      "loss": 1.0165,
+      "step": 10075
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.50791503963318e-05,
+      "loss": 1.021,
+      "step": 10076
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5078179811988892e-05,
+      "loss": 1.0421,
+      "step": 10077
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5077209163180264e-05,
+      "loss": 0.9274,
+      "step": 10078
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5076238449918244e-05,
+      "loss": 1.0888,
+      "step": 10079
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507526767221515e-05,
+      "loss": 1.0301,
+      "step": 10080
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5074296830083306e-05,
+      "loss": 0.9959,
+      "step": 10081
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507332592353504e-05,
+      "loss": 1.0559,
+      "step": 10082
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507235495258267e-05,
+      "loss": 1.0107,
+      "step": 10083
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.507138391723853e-05,
+      "loss": 1.1564,
+      "step": 10084
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5070412817514943e-05,
+      "loss": 1.1047,
+      "step": 10085
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5069441653424242e-05,
+      "loss": 0.9542,
+      "step": 10086
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5068470424978747e-05,
+      "loss": 1.126,
+      "step": 10087
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506749913219079e-05,
+      "loss": 1.0216,
+      "step": 10088
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5066527775072705e-05,
+      "loss": 1.0004,
+      "step": 10089
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5065556353636821e-05,
+      "loss": 1.0219,
+      "step": 10090
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5064584867895468e-05,
+      "loss": 1.0216,
+      "step": 10091
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5063613317860984e-05,
+      "loss": 0.9719,
+      "step": 10092
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5062641703545696e-05,
+      "loss": 1.0558,
+      "step": 10093
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.506167002496194e-05,
+      "loss": 1.0026,
+      "step": 10094
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5060698282122052e-05,
+      "loss": 1.1464,
+      "step": 10095
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5059726475038373e-05,
+      "loss": 0.9603,
+      "step": 10096
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5058754603723231e-05,
+      "loss": 1.0765,
+      "step": 10097
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5057782668188967e-05,
+      "loss": 1.0557,
+      "step": 10098
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5056810668447921e-05,
+      "loss": 1.0254,
+      "step": 10099
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5055838604512434e-05,
+      "loss": 0.9732,
+      "step": 10100
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.505486647639484e-05,
+      "loss": 1.0649,
+      "step": 10101
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5053894284107483e-05,
+      "loss": 0.954,
+      "step": 10102
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5052922027662708e-05,
+      "loss": 0.9344,
+      "step": 10103
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5051949707072848e-05,
+      "loss": 1.0768,
+      "step": 10104
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050977322350258e-05,
+      "loss": 0.9952,
+      "step": 10105
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5050004873507273e-05,
+      "loss": 0.9896,
+      "step": 10106
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5049032360556242e-05,
+      "loss": 1.0771,
+      "step": 10107
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504805978350951e-05,
+      "loss": 1.0199,
+      "step": 10108
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5047087142379425e-05,
+      "loss": 0.9704,
+      "step": 10109
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5046114437178332e-05,
+      "loss": 1.2165,
+      "step": 10110
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5045141667918577e-05,
+      "loss": 1.0887,
+      "step": 10111
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5044168834612515e-05,
+      "loss": 0.8945,
+      "step": 10112
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5043195937272492e-05,
+      "loss": 0.9882,
+      "step": 10113
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.504222297591086e-05,
+      "loss": 1.1158,
+      "step": 10114
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5041249950539972e-05,
+      "loss": 1.1713,
+      "step": 10115
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5040276861172173e-05,
+      "loss": 0.9489,
+      "step": 10116
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5039303707819826e-05,
+      "loss": 0.9524,
+      "step": 10117
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5038330490495274e-05,
+      "loss": 1.0556,
+      "step": 10118
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5037357209210882e-05,
+      "loss": 1.0133,
+      "step": 10119
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5036383863979e-05,
+      "loss": 0.9954,
+      "step": 10120
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5035410454811987e-05,
+      "loss": 0.9947,
+      "step": 10121
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5034436981722195e-05,
+      "loss": 0.9267,
+      "step": 10122
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5033463444721986e-05,
+      "loss": 1.0816,
+      "step": 10123
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.503248984382372e-05,
+      "loss": 1.0733,
+      "step": 10124
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5031516179039752e-05,
+      "loss": 0.9193,
+      "step": 10125
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5030542450382445e-05,
+      "loss": 0.9434,
+      "step": 10126
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5029568657864163e-05,
+      "loss": 0.9112,
+      "step": 10127
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502859480149726e-05,
+      "loss": 1.0243,
+      "step": 10128
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5027620881294107e-05,
+      "loss": 1.2147,
+      "step": 10129
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5026646897267063e-05,
+      "loss": 1.0865,
+      "step": 10130
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5025672849428496e-05,
+      "loss": 0.9671,
+      "step": 10131
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5024698737790764e-05,
+      "loss": 1.2069,
+      "step": 10132
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5023724562366245e-05,
+      "loss": 1.1946,
+      "step": 10133
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5022750323167292e-05,
+      "loss": 1.0731,
+      "step": 10134
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5021776020206283e-05,
+      "loss": 0.8466,
+      "step": 10135
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.502080165349558e-05,
+      "loss": 1.1299,
+      "step": 10136
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5019827223047558e-05,
+      "loss": 1.0558,
+      "step": 10137
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5018852728874582e-05,
+      "loss": 1.1259,
+      "step": 10138
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5017878170989027e-05,
+      "loss": 1.0573,
+      "step": 10139
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501690354940326e-05,
+      "loss": 1.0566,
+      "step": 10140
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5015928864129656e-05,
+      "loss": 1.0533,
+      "step": 10141
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.501495411518059e-05,
+      "loss": 1.006,
+      "step": 10142
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013979302568434e-05,
+      "loss": 1.2177,
+      "step": 10143
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5013004426305564e-05,
+      "loss": 1.1035,
+      "step": 10144
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5012029486404354e-05,
+      "loss": 1.0435,
+      "step": 10145
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5011054482877181e-05,
+      "loss": 1.0417,
+      "step": 10146
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5010079415736423e-05,
+      "loss": 1.145,
+      "step": 10147
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.500910428499446e-05,
+      "loss": 1.0326,
+      "step": 10148
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5008129090663666e-05,
+      "loss": 1.1457,
+      "step": 10149
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5007153832756424e-05,
+      "loss": 1.0055,
+      "step": 10150
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5006178511285118e-05,
+      "loss": 1.0547,
+      "step": 10151
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5005203126262122e-05,
+      "loss": 1.0134,
+      "step": 10152
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5004227677699823e-05,
+      "loss": 0.9251,
+      "step": 10153
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5003252165610604e-05,
+      "loss": 0.9524,
+      "step": 10154
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5002276590006845e-05,
+      "loss": 0.9597,
+      "step": 10155
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5001300950900935e-05,
+      "loss": 0.9812,
+      "step": 10156
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.5000325248305257e-05,
+      "loss": 1.0723,
+      "step": 10157
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4999349482232197e-05,
+      "loss": 1.0479,
+      "step": 10158
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4998373652694144e-05,
+      "loss": 1.0093,
+      "step": 10159
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4997397759703484e-05,
+      "loss": 1.1552,
+      "step": 10160
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499642180327261e-05,
+      "loss": 1.0269,
+      "step": 10161
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4995445783413902e-05,
+      "loss": 1.0075,
+      "step": 10162
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4994469700139757e-05,
+      "loss": 1.087,
+      "step": 10163
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4993493553462565e-05,
+      "loss": 0.9483,
+      "step": 10164
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4992517343394722e-05,
+      "loss": 0.9809,
+      "step": 10165
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.499154106994861e-05,
+      "loss": 0.9606,
+      "step": 10166
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4990564733136631e-05,
+      "loss": 1.0344,
+      "step": 10167
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4989588332971181e-05,
+      "loss": 0.941,
+      "step": 10168
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4988611869464647e-05,
+      "loss": 1.0039,
+      "step": 10169
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.498763534262943e-05,
+      "loss": 1.0722,
+      "step": 10170
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4986658752477924e-05,
+      "loss": 0.9573,
+      "step": 10171
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4985682099022531e-05,
+      "loss": 1.0534,
+      "step": 10172
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4984705382275647e-05,
+      "loss": 0.9866,
+      "step": 10173
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4983728602249667e-05,
+      "loss": 1.1664,
+      "step": 10174
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4982751758956999e-05,
+      "loss": 0.9785,
+      "step": 10175
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4981774852410036e-05,
+      "loss": 0.986,
+      "step": 10176
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4980797882621181e-05,
+      "loss": 1.0028,
+      "step": 10177
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497982084960284e-05,
+      "loss": 1.0567,
+      "step": 10178
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4978843753367414e-05,
+      "loss": 1.0482,
+      "step": 10179
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4977866593927307e-05,
+      "loss": 1.0614,
+      "step": 10180
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4976889371294923e-05,
+      "loss": 1.0175,
+      "step": 10181
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.497591208548267e-05,
+      "loss": 1.0341,
+      "step": 10182
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4974934736502948e-05,
+      "loss": 1.1306,
+      "step": 10183
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4973957324368172e-05,
+      "loss": 0.9847,
+      "step": 10184
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972979849090742e-05,
+      "loss": 0.9146,
+      "step": 10185
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4972002310683078e-05,
+      "loss": 1.1157,
+      "step": 10186
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4971024709157578e-05,
+      "loss": 0.9397,
+      "step": 10187
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4970047044526656e-05,
+      "loss": 1.1092,
+      "step": 10188
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4969069316802727e-05,
+      "loss": 1.0295,
+      "step": 10189
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4968091525998199e-05,
+      "loss": 1.1419,
+      "step": 10190
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4967113672125484e-05,
+      "loss": 0.9927,
+      "step": 10191
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4966135755197e-05,
+      "loss": 0.983,
+      "step": 10192
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4965157775225156e-05,
+      "loss": 1.048,
+      "step": 10193
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.496417973222237e-05,
+      "loss": 0.9799,
+      "step": 10194
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4963201626201058e-05,
+      "loss": 0.9466,
+      "step": 10195
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4962223457173636e-05,
+      "loss": 0.9394,
+      "step": 10196
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4961245225152521e-05,
+      "loss": 1.0873,
+      "step": 10197
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4960266930150134e-05,
+      "loss": 0.9287,
+      "step": 10198
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4959288572178889e-05,
+      "loss": 0.9209,
+      "step": 10199
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4958310151251212e-05,
+      "loss": 0.9936,
+      "step": 10200
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495733166737952e-05,
+      "loss": 0.9845,
+      "step": 10201
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4956353120576236e-05,
+      "loss": 1.0383,
+      "step": 10202
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.495537451085378e-05,
+      "loss": 1.1491,
+      "step": 10203
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4954395838224575e-05,
+      "loss": 1.1324,
+      "step": 10204
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 1.4953417102701049e-05,
+      "loss": 1.0551,
+      "step": 10205
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4952438304295624e-05,
+      "loss": 1.0882,
+      "step": 10206
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4951459443020725e-05,
+      "loss": 1.0041,
+      "step": 10207
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.495048051888878e-05,
+      "loss": 1.0441,
+      "step": 10208
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4949501531912214e-05,
+      "loss": 1.0584,
+      "step": 10209
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494852248210346e-05,
+      "loss": 1.1301,
+      "step": 10210
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.494754336947494e-05,
+      "loss": 1.0544,
+      "step": 10211
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4946564194039083e-05,
+      "loss": 0.9773,
+      "step": 10212
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4945584955808325e-05,
+      "loss": 0.9189,
+      "step": 10213
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4944605654795098e-05,
+      "loss": 1.0025,
+      "step": 10214
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4943626291011828e-05,
+      "loss": 0.9581,
+      "step": 10215
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4942646864470949e-05,
+      "loss": 1.0913,
+      "step": 10216
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4941667375184898e-05,
+      "loss": 1.068,
+      "step": 10217
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4940687823166102e-05,
+      "loss": 1.0391,
+      "step": 10218
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4939708208427006e-05,
+      "loss": 1.0249,
+      "step": 10219
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493872853098004e-05,
+      "loss": 1.0389,
+      "step": 10220
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4937748790837643e-05,
+      "loss": 0.9471,
+      "step": 10221
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4936768988012249e-05,
+      "loss": 0.9199,
+      "step": 10222
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4935789122516297e-05,
+      "loss": 0.9457,
+      "step": 10223
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4934809194362233e-05,
+      "loss": 0.9227,
+      "step": 10224
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4933829203562482e-05,
+      "loss": 1.0143,
+      "step": 10225
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.49328491501295e-05,
+      "loss": 1.0151,
+      "step": 10226
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493186903407572e-05,
+      "loss": 1.0585,
+      "step": 10227
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.493088885541359e-05,
+      "loss": 0.8931,
+      "step": 10228
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4929908614155546e-05,
+      "loss": 0.9781,
+      "step": 10229
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4928928310314037e-05,
+      "loss": 0.8889,
+      "step": 10230
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4927947943901505e-05,
+      "loss": 1.1415,
+      "step": 10231
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4926967514930395e-05,
+      "loss": 1.0693,
+      "step": 10232
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925987023413158e-05,
+      "loss": 1.0759,
+      "step": 10233
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4925006469362235e-05,
+      "loss": 0.8997,
+      "step": 10234
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4924025852790076e-05,
+      "loss": 0.99,
+      "step": 10235
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4923045173709131e-05,
+      "loss": 1.0401,
+      "step": 10236
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4922064432131848e-05,
+      "loss": 1.1294,
+      "step": 10237
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.492108362807068e-05,
+      "loss": 0.9671,
+      "step": 10238
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4920102761538072e-05,
+      "loss": 1.0266,
+      "step": 10239
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4919121832546482e-05,
+      "loss": 0.9239,
+      "step": 10240
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4918140841108356e-05,
+      "loss": 0.9943,
+      "step": 10241
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4917159787236153e-05,
+      "loss": 1.1547,
+      "step": 10242
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4916178670942327e-05,
+      "loss": 0.8202,
+      "step": 10243
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491519749223933e-05,
+      "loss": 1.1162,
+      "step": 10244
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491421625113962e-05,
+      "loss": 1.0495,
+      "step": 10245
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4913234947655648e-05,
+      "loss": 0.9702,
+      "step": 10246
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.491225358179988e-05,
+      "loss": 1.0739,
+      "step": 10247
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4911272153584769e-05,
+      "loss": 1.0319,
+      "step": 10248
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4910290663022772e-05,
+      "loss": 0.9682,
+      "step": 10249
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4909309110126356e-05,
+      "loss": 1.0293,
+      "step": 10250
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4908327494907973e-05,
+      "loss": 0.9671,
+      "step": 10251
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490734581738009e-05,
+      "loss": 0.9475,
+      "step": 10252
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4906364077555165e-05,
+      "loss": 1.0657,
+      "step": 10253
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4905382275445665e-05,
+      "loss": 1.1404,
+      "step": 10254
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490440041106405e-05,
+      "loss": 0.9824,
+      "step": 10255
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4903418484422787e-05,
+      "loss": 0.9777,
+      "step": 10256
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4902436495534342e-05,
+      "loss": 1.0074,
+      "step": 10257
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490145444441118e-05,
+      "loss": 1.0249,
+      "step": 10258
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.490047233106576e-05,
+      "loss": 1.029,
+      "step": 10259
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4899490155510563e-05,
+      "loss": 0.9454,
+      "step": 10260
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.489850791775805e-05,
+      "loss": 0.993,
+      "step": 10261
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4897525617820691e-05,
+      "loss": 0.9792,
+      "step": 10262
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4896543255710956e-05,
+      "loss": 1.0933,
+      "step": 10263
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4895560831441316e-05,
+      "loss": 0.9654,
+      "step": 10264
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4894578345024242e-05,
+      "loss": 1.0794,
+      "step": 10265
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4893595796472206e-05,
+      "loss": 1.1509,
+      "step": 10266
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4892613185797682e-05,
+      "loss": 1.0327,
+      "step": 10267
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4891630513013147e-05,
+      "loss": 1.1332,
+      "step": 10268
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4890647778131069e-05,
+      "loss": 1.0209,
+      "step": 10269
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4889664981163927e-05,
+      "loss": 1.027,
+      "step": 10270
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4888682122124198e-05,
+      "loss": 1.063,
+      "step": 10271
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488769920102436e-05,
+      "loss": 1.0403,
+      "step": 10272
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4886716217876886e-05,
+      "loss": 0.9733,
+      "step": 10273
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4885733172694257e-05,
+      "loss": 1.1455,
+      "step": 10274
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4884750065488957e-05,
+      "loss": 1.0384,
+      "step": 10275
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.488376689627346e-05,
+      "loss": 1.1189,
+      "step": 10276
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4882783665060249e-05,
+      "loss": 1.0113,
+      "step": 10277
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4881800371861808e-05,
+      "loss": 1.1061,
+      "step": 10278
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4880817016690616e-05,
+      "loss": 1.0414,
+      "step": 10279
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4879833599559158e-05,
+      "loss": 1.0478,
+      "step": 10280
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4878850120479919e-05,
+      "loss": 1.0946,
+      "step": 10281
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4877866579465384e-05,
+      "loss": 1.0002,
+      "step": 10282
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4876882976528036e-05,
+      "loss": 1.1217,
+      "step": 10283
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4875899311680362e-05,
+      "loss": 1.0674,
+      "step": 10284
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4874915584934853e-05,
+      "loss": 0.987,
+      "step": 10285
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4873931796303995e-05,
+      "loss": 0.9679,
+      "step": 10286
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4872947945800273e-05,
+      "loss": 1.1338,
+      "step": 10287
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.487196403343618e-05,
+      "loss": 1.0759,
+      "step": 10288
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4870980059224214e-05,
+      "loss": 0.9486,
+      "step": 10289
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869996023176852e-05,
+      "loss": 1.0208,
+      "step": 10290
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4869011925306594e-05,
+      "loss": 1.0289,
+      "step": 10291
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4868027765625932e-05,
+      "loss": 1.0599,
+      "step": 10292
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486704354414736e-05,
+      "loss": 1.058,
+      "step": 10293
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486605926088337e-05,
+      "loss": 0.8844,
+      "step": 10294
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4865074915846458e-05,
+      "loss": 1.0682,
+      "step": 10295
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4864090509049124e-05,
+      "loss": 1.0443,
+      "step": 10296
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4863106040503857e-05,
+      "loss": 1.0851,
+      "step": 10297
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.486212151022316e-05,
+      "loss": 0.9965,
+      "step": 10298
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4861136918219528e-05,
+      "loss": 0.9427,
+      "step": 10299
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4860152264505464e-05,
+      "loss": 1.0515,
+      "step": 10300
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4859167549093467e-05,
+      "loss": 0.9985,
+      "step": 10301
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4858182771996035e-05,
+      "loss": 1.0407,
+      "step": 10302
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4857197933225674e-05,
+      "loss": 1.0522,
+      "step": 10303
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.485621303279488e-05,
+      "loss": 1.037,
+      "step": 10304
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4855228070716156e-05,
+      "loss": 1.0624,
+      "step": 10305
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4854243047002013e-05,
+      "loss": 0.976,
+      "step": 10306
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4853257961664951e-05,
+      "loss": 1.1172,
+      "step": 10307
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4852272814717475e-05,
+      "loss": 1.0916,
+      "step": 10308
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4851287606172089e-05,
+      "loss": 0.9123,
+      "step": 10309
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4850302336041307e-05,
+      "loss": 1.1631,
+      "step": 10310
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4849317004337629e-05,
+      "loss": 1.1743,
+      "step": 10311
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4848331611073568e-05,
+      "loss": 1.0429,
+      "step": 10312
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4847346156261631e-05,
+      "loss": 1.0097,
+      "step": 10313
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.484636063991433e-05,
+      "loss": 1.0334,
+      "step": 10314
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4845375062044174e-05,
+      "loss": 1.0048,
+      "step": 10315
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4844389422663674e-05,
+      "loss": 1.0571,
+      "step": 10316
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4843403721785347e-05,
+      "loss": 1.0527,
+      "step": 10317
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.48424179594217e-05,
+      "loss": 1.047,
+      "step": 10318
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4841432135585249e-05,
+      "loss": 1.187,
+      "step": 10319
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4840446250288509e-05,
+      "loss": 1.0118,
+      "step": 10320
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4839460303544e-05,
+      "loss": 0.9942,
+      "step": 10321
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4838474295364232e-05,
+      "loss": 1.0673,
+      "step": 10322
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.483748822576172e-05,
+      "loss": 1.0774,
+      "step": 10323
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4836502094748992e-05,
+      "loss": 0.9617,
+      "step": 10324
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4835515902338557e-05,
+      "loss": 1.1426,
+      "step": 10325
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4834529648542937e-05,
+      "loss": 1.1051,
+      "step": 10326
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4833543333374655e-05,
+      "loss": 0.8862,
+      "step": 10327
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4832556956846231e-05,
+      "loss": 1.0978,
+      "step": 10328
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4831570518970182e-05,
+      "loss": 1.0546,
+      "step": 10329
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4830584019759035e-05,
+      "loss": 1.059,
+      "step": 10330
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4829597459225315e-05,
+      "loss": 1.1462,
+      "step": 10331
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.482861083738154e-05,
+      "loss": 1.1233,
+      "step": 10332
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4827624154240242e-05,
+      "loss": 0.9311,
+      "step": 10333
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4826637409813938e-05,
+      "loss": 1.0344,
+      "step": 10334
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4825650604115164e-05,
+      "loss": 1.13,
+      "step": 10335
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4824663737156439e-05,
+      "loss": 1.0788,
+      "step": 10336
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4823676808950296e-05,
+      "loss": 1.0359,
+      "step": 10337
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4822689819509262e-05,
+      "loss": 1.156,
+      "step": 10338
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4821702768845863e-05,
+      "loss": 0.9429,
+      "step": 10339
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4820715656972638e-05,
+      "loss": 1.0118,
+      "step": 10340
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481972848390211e-05,
+      "loss": 1.0443,
+      "step": 10341
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4818741249646813e-05,
+      "loss": 1.0669,
+      "step": 10342
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.481775395421928e-05,
+      "loss": 1.1073,
+      "step": 10343
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4816766597632047e-05,
+      "loss": 0.999,
+      "step": 10344
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4815779179897645e-05,
+      "loss": 0.9279,
+      "step": 10345
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4814791701028605e-05,
+      "loss": 1.0752,
+      "step": 10346
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4813804161037473e-05,
+      "loss": 1.0173,
+      "step": 10347
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4812816559936778e-05,
+      "loss": 1.0686,
+      "step": 10348
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4811828897739058e-05,
+      "loss": 1.1104,
+      "step": 10349
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4810841174456853e-05,
+      "loss": 1.0854,
+      "step": 10350
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4809853390102703e-05,
+      "loss": 1.1587,
+      "step": 10351
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4808865544689144e-05,
+      "loss": 1.0374,
+      "step": 10352
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4807877638228714e-05,
+      "loss": 1.0592,
+      "step": 10353
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4806889670733961e-05,
+      "loss": 1.1305,
+      "step": 10354
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4805901642217424e-05,
+      "loss": 1.0581,
+      "step": 10355
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4804913552691644e-05,
+      "loss": 1.1227,
+      "step": 10356
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4803925402169169e-05,
+      "loss": 1.0483,
+      "step": 10357
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4802937190662539e-05,
+      "loss": 1.0244,
+      "step": 10358
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4801948918184299e-05,
+      "loss": 0.9946,
+      "step": 10359
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4800960584746995e-05,
+      "loss": 1.0407,
+      "step": 10360
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4799972190363176e-05,
+      "loss": 1.003,
+      "step": 10361
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4798983735045388e-05,
+      "loss": 1.1071,
+      "step": 10362
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797995218806179e-05,
+      "loss": 1.0611,
+      "step": 10363
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4797006641658099e-05,
+      "loss": 0.9713,
+      "step": 10364
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4796018003613693e-05,
+      "loss": 1.0948,
+      "step": 10365
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4795029304685516e-05,
+      "loss": 1.0228,
+      "step": 10366
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4794040544886117e-05,
+      "loss": 0.9034,
+      "step": 10367
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4793051724228054e-05,
+      "loss": 0.996,
+      "step": 10368
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4792062842723872e-05,
+      "loss": 1.0016,
+      "step": 10369
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4791073900386124e-05,
+      "loss": 1.0685,
+      "step": 10370
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.479008489722737e-05,
+      "loss": 1.084,
+      "step": 10371
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4789095833260163e-05,
+      "loss": 0.9155,
+      "step": 10372
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478810670849706e-05,
+      "loss": 1.0298,
+      "step": 10373
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.478711752295061e-05,
+      "loss": 1.0431,
+      "step": 10374
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4786128276633382e-05,
+      "loss": 1.0634,
+      "step": 10375
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4785138969557926e-05,
+      "loss": 1.1192,
+      "step": 10376
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4784149601736804e-05,
+      "loss": 1.1252,
+      "step": 10377
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4783160173182576e-05,
+      "loss": 1.0436,
+      "step": 10378
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4782170683907802e-05,
+      "loss": 0.9438,
+      "step": 10379
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4781181133925038e-05,
+      "loss": 1.0719,
+      "step": 10380
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4780191523246854e-05,
+      "loss": 1.0857,
+      "step": 10381
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4779201851885811e-05,
+      "loss": 1.02,
+      "step": 10382
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4778212119854469e-05,
+      "loss": 1.0113,
+      "step": 10383
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4777222327165395e-05,
+      "loss": 1.0157,
+      "step": 10384
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4776232473831154e-05,
+      "loss": 1.0554,
+      "step": 10385
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477524255986431e-05,
+      "loss": 1.0421,
+      "step": 10386
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477425258527743e-05,
+      "loss": 0.9706,
+      "step": 10387
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4773262550083083e-05,
+      "loss": 0.9674,
+      "step": 10388
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4772272454293837e-05,
+      "loss": 1.0917,
+      "step": 10389
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.477128229792226e-05,
+      "loss": 0.8652,
+      "step": 10390
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4770292080980922e-05,
+      "loss": 1.0179,
+      "step": 10391
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4769301803482395e-05,
+      "loss": 0.9366,
+      "step": 10392
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4768311465439245e-05,
+      "loss": 1.0772,
+      "step": 10393
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4767321066864048e-05,
+      "loss": 1.1896,
+      "step": 10394
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4766330607769379e-05,
+      "loss": 1.0919,
+      "step": 10395
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.476534008816781e-05,
+      "loss": 0.9933,
+      "step": 10396
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4764349508071911e-05,
+      "loss": 1.0779,
+      "step": 10397
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4763358867494261e-05,
+      "loss": 0.9874,
+      "step": 10398
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4762368166447436e-05,
+      "loss": 1.0605,
+      "step": 10399
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4761377404944014e-05,
+      "loss": 1.2262,
+      "step": 10400
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4760386582996567e-05,
+      "loss": 0.9484,
+      "step": 10401
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4759395700617674e-05,
+      "loss": 0.7984,
+      "step": 10402
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.475840475781992e-05,
+      "loss": 0.9562,
+      "step": 10403
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4757413754615879e-05,
+      "loss": 0.9653,
+      "step": 10404
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4756422691018135e-05,
+      "loss": 1.1421,
+      "step": 10405
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4755431567039267e-05,
+      "loss": 1.1491,
+      "step": 10406
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4754440382691857e-05,
+      "loss": 1.1094,
+      "step": 10407
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4753449137988487e-05,
+      "loss": 0.9228,
+      "step": 10408
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4752457832941742e-05,
+      "loss": 1.0561,
+      "step": 10409
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4751466467564208e-05,
+      "loss": 1.0725,
+      "step": 10410
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4750475041868466e-05,
+      "loss": 1.0224,
+      "step": 10411
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4749483555867107e-05,
+      "loss": 0.9817,
+      "step": 10412
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4748492009572712e-05,
+      "loss": 1.1051,
+      "step": 10413
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474750040299787e-05,
+      "loss": 0.9442,
+      "step": 10414
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.474650873615517e-05,
+      "loss": 1.1026,
+      "step": 10415
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4745517009057203e-05,
+      "loss": 1.0695,
+      "step": 10416
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4744525221716552e-05,
+      "loss": 1.1515,
+      "step": 10417
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4743533374145816e-05,
+      "loss": 1.071,
+      "step": 10418
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4742541466357578e-05,
+      "loss": 1.0063,
+      "step": 10419
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4741549498364435e-05,
+      "loss": 1.0927,
+      "step": 10420
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4740557470178981e-05,
+      "loss": 1.0384,
+      "step": 10421
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4739565381813806e-05,
+      "loss": 0.9996,
+      "step": 10422
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4738573233281504e-05,
+      "loss": 1.0635,
+      "step": 10423
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4737581024594673e-05,
+      "loss": 1.0765,
+      "step": 10424
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4736588755765904e-05,
+      "loss": 1.0993,
+      "step": 10425
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4735596426807797e-05,
+      "loss": 1.0045,
+      "step": 10426
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473460403773295e-05,
+      "loss": 1.0024,
+      "step": 10427
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473361158855396e-05,
+      "loss": 0.9567,
+      "step": 10428
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4732619079283423e-05,
+      "loss": 1.0854,
+      "step": 10429
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4731626509933944e-05,
+      "loss": 1.1066,
+      "step": 10430
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.473063388051812e-05,
+      "loss": 0.979,
+      "step": 10431
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472964119104855e-05,
+      "loss": 0.9699,
+      "step": 10432
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4728648441537838e-05,
+      "loss": 0.9264,
+      "step": 10433
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4727655631998587e-05,
+      "loss": 0.9615,
+      "step": 10434
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4726662762443402e-05,
+      "loss": 1.141,
+      "step": 10435
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4725669832884885e-05,
+      "loss": 1.0039,
+      "step": 10436
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472467684333564e-05,
+      "loss": 1.0307,
+      "step": 10437
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4723683793808276e-05,
+      "loss": 0.935,
+      "step": 10438
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4722690684315395e-05,
+      "loss": 1.0341,
+      "step": 10439
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.472169751486961e-05,
+      "loss": 1.021,
+      "step": 10440
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4720704285483523e-05,
+      "loss": 1.0305,
+      "step": 10441
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4719710996169747e-05,
+      "loss": 1.1069,
+      "step": 10442
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4718717646940888e-05,
+      "loss": 1.0407,
+      "step": 10443
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4717724237809557e-05,
+      "loss": 1.1207,
+      "step": 10444
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4716730768788368e-05,
+      "loss": 0.9289,
+      "step": 10445
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4715737239889927e-05,
+      "loss": 1.0674,
+      "step": 10446
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4714743651126855e-05,
+      "loss": 0.9444,
+      "step": 10447
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4713750002511757e-05,
+      "loss": 1.1706,
+      "step": 10448
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4712756294057253e-05,
+      "loss": 1.143,
+      "step": 10449
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4711762525775952e-05,
+      "loss": 0.942,
+      "step": 10450
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4710768697680473e-05,
+      "loss": 0.9981,
+      "step": 10451
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4709774809783435e-05,
+      "loss": 1.0333,
+      "step": 10452
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4708780862097448e-05,
+      "loss": 1.0573,
+      "step": 10453
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4707786854635136e-05,
+      "loss": 1.0423,
+      "step": 10454
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4706792787409115e-05,
+      "loss": 0.9875,
+      "step": 10455
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4705798660432008e-05,
+      "loss": 1.0564,
+      "step": 10456
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4704804473716425e-05,
+      "loss": 1.0552,
+      "step": 10457
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4703810227274997e-05,
+      "loss": 1.0118,
+      "step": 10458
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4702815921120346e-05,
+      "loss": 0.9914,
+      "step": 10459
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4701821555265086e-05,
+      "loss": 0.9491,
+      "step": 10460
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4700827129721846e-05,
+      "loss": 1.129,
+      "step": 10461
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469983264450325e-05,
+      "loss": 1.0358,
+      "step": 10462
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469883809962192e-05,
+      "loss": 0.9255,
+      "step": 10463
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4697843495090483e-05,
+      "loss": 1.1371,
+      "step": 10464
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4696848830921563e-05,
+      "loss": 1.0674,
+      "step": 10465
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4695854107127794e-05,
+      "loss": 0.9963,
+      "step": 10466
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4694859323721793e-05,
+      "loss": 0.9766,
+      "step": 10467
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4693864480716195e-05,
+      "loss": 1.0823,
+      "step": 10468
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.469286957812363e-05,
+      "loss": 0.9405,
+      "step": 10469
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4691874615956724e-05,
+      "loss": 1.0343,
+      "step": 10470
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4690879594228113e-05,
+      "loss": 1.0588,
+      "step": 10471
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4689884512950423e-05,
+      "loss": 1.0845,
+      "step": 10472
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468888937213629e-05,
+      "loss": 0.9112,
+      "step": 10473
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4687894171798342e-05,
+      "loss": 1.0577,
+      "step": 10474
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4686898911949218e-05,
+      "loss": 1.0431,
+      "step": 10475
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468590359260155e-05,
+      "loss": 0.9662,
+      "step": 10476
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4684908213767974e-05,
+      "loss": 0.9424,
+      "step": 10477
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.468391277546113e-05,
+      "loss": 1.1785,
+      "step": 10478
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4682917277693644e-05,
+      "loss": 1.0386,
+      "step": 10479
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4681921720478163e-05,
+      "loss": 1.1585,
+      "step": 10480
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4680926103827321e-05,
+      "loss": 1.0751,
+      "step": 10481
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4679930427753761e-05,
+      "loss": 0.9978,
+      "step": 10482
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4678934692270117e-05,
+      "loss": 0.948,
+      "step": 10483
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4677938897389035e-05,
+      "loss": 1.111,
+      "step": 10484
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4676943043123153e-05,
+      "loss": 0.8838,
+      "step": 10485
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4675947129485113e-05,
+      "loss": 1.056,
+      "step": 10486
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4674951156487558e-05,
+      "loss": 1.0028,
+      "step": 10487
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4673955124143133e-05,
+      "loss": 0.9174,
+      "step": 10488
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4672959032464484e-05,
+      "loss": 1.0603,
+      "step": 10489
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467196288146425e-05,
+      "loss": 0.9677,
+      "step": 10490
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.467096667115508e-05,
+      "loss": 1.0232,
+      "step": 10491
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 1.4669970401549624e-05,
+      "loss": 0.9569,
+      "step": 10492
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4668974072660524e-05,
+      "loss": 0.9729,
+      "step": 10493
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466797768450043e-05,
+      "loss": 0.9852,
+      "step": 10494
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.466698123708199e-05,
+      "loss": 0.8844,
+      "step": 10495
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4665984730417857e-05,
+      "loss": 0.999,
+      "step": 10496
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4664988164520676e-05,
+      "loss": 0.9782,
+      "step": 10497
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4663991539403104e-05,
+      "loss": 1.0256,
+      "step": 10498
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4662994855077789e-05,
+      "loss": 1.0134,
+      "step": 10499
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661998111557382e-05,
+      "loss": 1.067,
+      "step": 10500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4661001308854538e-05,
+      "loss": 0.9673,
+      "step": 10501
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4660004446981915e-05,
+      "loss": 1.069,
+      "step": 10502
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4659007525952163e-05,
+      "loss": 1.1067,
+      "step": 10503
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.465801054577794e-05,
+      "loss": 1.0052,
+      "step": 10504
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4657013506471898e-05,
+      "loss": 0.9713,
+      "step": 10505
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4656016408046698e-05,
+      "loss": 0.9889,
+      "step": 10506
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4655019250515e-05,
+      "loss": 1.0128,
+      "step": 10507
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4654022033889457e-05,
+      "loss": 0.9052,
+      "step": 10508
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4653024758182732e-05,
+      "loss": 0.9961,
+      "step": 10509
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4652027423407484e-05,
+      "loss": 1.1341,
+      "step": 10510
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4651030029576374e-05,
+      "loss": 0.9665,
+      "step": 10511
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4650032576702065e-05,
+      "loss": 1.0426,
+      "step": 10512
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4649035064797215e-05,
+      "loss": 1.1273,
+      "step": 10513
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4648037493874492e-05,
+      "loss": 0.9594,
+      "step": 10514
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4647039863946554e-05,
+      "loss": 0.9277,
+      "step": 10515
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4646042175026074e-05,
+      "loss": 1.119,
+      "step": 10516
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4645044427125711e-05,
+      "loss": 1.1275,
+      "step": 10517
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4644046620258132e-05,
+      "loss": 0.994,
+      "step": 10518
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4643048754436005e-05,
+      "loss": 0.9636,
+      "step": 10519
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4642050829671997e-05,
+      "loss": 1.0021,
+      "step": 10520
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4641052845978776e-05,
+      "loss": 0.9391,
+      "step": 10521
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.464005480336901e-05,
+      "loss": 1.0357,
+      "step": 10522
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4639056701855373e-05,
+      "loss": 0.9547,
+      "step": 10523
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.463805854145053e-05,
+      "loss": 0.911,
+      "step": 10524
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4637060322167157e-05,
+      "loss": 0.9356,
+      "step": 10525
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4636062044017922e-05,
+      "loss": 1.0029,
+      "step": 10526
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4635063707015503e-05,
+      "loss": 1.0097,
+      "step": 10527
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4634065311172568e-05,
+      "loss": 1.0886,
+      "step": 10528
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4633066856501792e-05,
+      "loss": 1.1224,
+      "step": 10529
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4632068343015852e-05,
+      "loss": 1.0707,
+      "step": 10530
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4631069770727428e-05,
+      "loss": 1.1128,
+      "step": 10531
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4630071139649188e-05,
+      "loss": 1.0384,
+      "step": 10532
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4629072449793814e-05,
+      "loss": 1.0177,
+      "step": 10533
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4628073701173986e-05,
+      "loss": 1.0327,
+      "step": 10534
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4627074893802378e-05,
+      "loss": 0.9584,
+      "step": 10535
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4626076027691668e-05,
+      "loss": 0.9074,
+      "step": 10536
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4625077102854544e-05,
+      "loss": 1.0126,
+      "step": 10537
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4624078119303682e-05,
+      "loss": 0.9772,
+      "step": 10538
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4623079077051765e-05,
+      "loss": 1.079,
+      "step": 10539
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4622079976111474e-05,
+      "loss": 1.1748,
+      "step": 10540
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4621080816495493e-05,
+      "loss": 0.9831,
+      "step": 10541
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.462008159821651e-05,
+      "loss": 0.9793,
+      "step": 10542
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4619082321287201e-05,
+      "loss": 1.0667,
+      "step": 10543
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461808298572026e-05,
+      "loss": 0.9847,
+      "step": 10544
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4617083591528367e-05,
+      "loss": 1.0498,
+      "step": 10545
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4616084138724214e-05,
+      "loss": 1.063,
+      "step": 10546
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4615084627320485e-05,
+      "loss": 0.9407,
+      "step": 10547
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.461408505732987e-05,
+      "loss": 1.1394,
+      "step": 10548
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4613085428765057e-05,
+      "loss": 1.0261,
+      "step": 10549
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4612085741638737e-05,
+      "loss": 1.0233,
+      "step": 10550
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4611085995963601e-05,
+      "loss": 1.0472,
+      "step": 10551
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4610086191752341e-05,
+      "loss": 1.1562,
+      "step": 10552
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4609086329017648e-05,
+      "loss": 1.0835,
+      "step": 10553
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4608086407772215e-05,
+      "loss": 1.1347,
+      "step": 10554
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4607086428028736e-05,
+      "loss": 1.1505,
+      "step": 10555
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4606086389799904e-05,
+      "loss": 1.1149,
+      "step": 10556
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4605086293098416e-05,
+      "loss": 1.074,
+      "step": 10557
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4604086137936965e-05,
+      "loss": 1.0279,
+      "step": 10558
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4603085924328256e-05,
+      "loss": 1.1014,
+      "step": 10559
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4602085652284971e-05,
+      "loss": 1.0162,
+      "step": 10560
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4601085321819825e-05,
+      "loss": 0.9614,
+      "step": 10561
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4600084932945506e-05,
+      "loss": 0.9958,
+      "step": 10562
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459908448567472e-05,
+      "loss": 1.0804,
+      "step": 10563
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459808398002016e-05,
+      "loss": 0.9113,
+      "step": 10564
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4597083415994531e-05,
+      "loss": 1.0566,
+      "step": 10565
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4596082793610538e-05,
+      "loss": 0.9242,
+      "step": 10566
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4595082112880877e-05,
+      "loss": 1.0612,
+      "step": 10567
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4594081373818255e-05,
+      "loss": 0.9535,
+      "step": 10568
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4593080576435377e-05,
+      "loss": 1.0232,
+      "step": 10569
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4592079720744946e-05,
+      "loss": 0.9192,
+      "step": 10570
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.459107880675967e-05,
+      "loss": 1.0146,
+      "step": 10571
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4590077834492252e-05,
+      "loss": 1.0579,
+      "step": 10572
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.45890768039554e-05,
+      "loss": 1.0422,
+      "step": 10573
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.458807571516182e-05,
+      "loss": 0.9813,
+      "step": 10574
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4587074568124227e-05,
+      "loss": 1.1073,
+      "step": 10575
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4586073362855322e-05,
+      "loss": 1.0237,
+      "step": 10576
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4585072099367819e-05,
+      "loss": 1.0029,
+      "step": 10577
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4584070777674431e-05,
+      "loss": 1.1035,
+      "step": 10578
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4583069397787863e-05,
+      "loss": 1.0445,
+      "step": 10579
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4582067959720835e-05,
+      "loss": 1.1208,
+      "step": 10580
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4581066463486053e-05,
+      "loss": 1.0436,
+      "step": 10581
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4580064909096235e-05,
+      "loss": 0.8899,
+      "step": 10582
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4579063296564093e-05,
+      "loss": 1.1463,
+      "step": 10583
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4578061625902342e-05,
+      "loss": 1.0219,
+      "step": 10584
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4577059897123702e-05,
+      "loss": 1.0537,
+      "step": 10585
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4576058110240885e-05,
+      "loss": 1.0035,
+      "step": 10586
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4575056265266608e-05,
+      "loss": 0.9383,
+      "step": 10587
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4574054362213591e-05,
+      "loss": 0.965,
+      "step": 10588
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4573052401094554e-05,
+      "loss": 1.0292,
+      "step": 10589
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4572050381922217e-05,
+      "loss": 1.0484,
+      "step": 10590
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4571048304709294e-05,
+      "loss": 1.0561,
+      "step": 10591
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4570046169468515e-05,
+      "loss": 1.031,
+      "step": 10592
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4569043976212595e-05,
+      "loss": 0.9776,
+      "step": 10593
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4568041724954258e-05,
+      "loss": 1.1559,
+      "step": 10594
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456703941570623e-05,
+      "loss": 1.18,
+      "step": 10595
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4566037048481234e-05,
+      "loss": 1.0644,
+      "step": 10596
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4565034623291991e-05,
+      "loss": 1.0925,
+      "step": 10597
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456403214015123e-05,
+      "loss": 0.8869,
+      "step": 10598
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4563029599071676e-05,
+      "loss": 1.0052,
+      "step": 10599
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4562027000066056e-05,
+      "loss": 0.9911,
+      "step": 10600
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4561024343147096e-05,
+      "loss": 1.0672,
+      "step": 10601
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.456002162832753e-05,
+      "loss": 1.0253,
+      "step": 10602
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4559018855620083e-05,
+      "loss": 0.9628,
+      "step": 10603
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4558016025037483e-05,
+      "loss": 1.0309,
+      "step": 10604
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4557013136592464e-05,
+      "loss": 1.0483,
+      "step": 10605
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4556010190297756e-05,
+      "loss": 1.0433,
+      "step": 10606
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.455500718616609e-05,
+      "loss": 1.0232,
+      "step": 10607
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4554004124210201e-05,
+      "loss": 1.0486,
+      "step": 10608
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4553001004442822e-05,
+      "loss": 0.9776,
+      "step": 10609
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4551997826876684e-05,
+      "loss": 1.1364,
+      "step": 10610
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4550994591524527e-05,
+      "loss": 1.0026,
+      "step": 10611
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4549991298399084e-05,
+      "loss": 1.2314,
+      "step": 10612
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4548987947513093e-05,
+      "loss": 1.0229,
+      "step": 10613
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454798453887929e-05,
+      "loss": 1.014,
+      "step": 10614
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454698107251041e-05,
+      "loss": 1.0621,
+      "step": 10615
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4545977548419195e-05,
+      "loss": 1.0424,
+      "step": 10616
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4544973966618387e-05,
+      "loss": 0.9583,
+      "step": 10617
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4543970327120722e-05,
+      "loss": 0.9481,
+      "step": 10618
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454296662993894e-05,
+      "loss": 1.0679,
+      "step": 10619
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.454196287508579e-05,
+      "loss": 0.9014,
+      "step": 10620
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4540959062574003e-05,
+      "loss": 1.0963,
+      "step": 10621
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453995519241633e-05,
+      "loss": 1.0966,
+      "step": 10622
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4538951264625512e-05,
+      "loss": 0.9206,
+      "step": 10623
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4537947279214295e-05,
+      "loss": 1.161,
+      "step": 10624
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4536943236195424e-05,
+      "loss": 1.1074,
+      "step": 10625
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4535939135581642e-05,
+      "loss": 1.0012,
+      "step": 10626
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4534934977385702e-05,
+      "loss": 1.1,
+      "step": 10627
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4533930761620346e-05,
+      "loss": 1.0837,
+      "step": 10628
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453292648829832e-05,
+      "loss": 0.9191,
+      "step": 10629
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.453192215743238e-05,
+      "loss": 0.9029,
+      "step": 10630
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4530917769035275e-05,
+      "loss": 0.9557,
+      "step": 10631
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4529913323119749e-05,
+      "loss": 0.9701,
+      "step": 10632
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4528908819698558e-05,
+      "loss": 0.9675,
+      "step": 10633
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4527904258784454e-05,
+      "loss": 1.0578,
+      "step": 10634
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4526899640390183e-05,
+      "loss": 0.9522,
+      "step": 10635
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4525894964528507e-05,
+      "loss": 1.0872,
+      "step": 10636
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4524890231212174e-05,
+      "loss": 1.0116,
+      "step": 10637
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4523885440453947e-05,
+      "loss": 1.0658,
+      "step": 10638
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.452288059226657e-05,
+      "loss": 0.9459,
+      "step": 10639
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4521875686662806e-05,
+      "loss": 1.026,
+      "step": 10640
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4520870723655414e-05,
+      "loss": 1.1158,
+      "step": 10641
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4519865703257143e-05,
+      "loss": 1.0622,
+      "step": 10642
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.451886062548076e-05,
+      "loss": 1.037,
+      "step": 10643
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4517855490339019e-05,
+      "loss": 0.9708,
+      "step": 10644
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4516850297844685e-05,
+      "loss": 0.9715,
+      "step": 10645
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4515845048010512e-05,
+      "loss": 1.0346,
+      "step": 10646
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4514839740849265e-05,
+      "loss": 1.0489,
+      "step": 10647
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4513834376373708e-05,
+      "loss": 1.0962,
+      "step": 10648
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4512828954596598e-05,
+      "loss": 0.9881,
+      "step": 10649
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4511823475530704e-05,
+      "loss": 0.9312,
+      "step": 10650
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4510817939188788e-05,
+      "loss": 1.0381,
+      "step": 10651
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4509812345583616e-05,
+      "loss": 1.1996,
+      "step": 10652
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450880669472795e-05,
+      "loss": 1.1031,
+      "step": 10653
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450780098663456e-05,
+      "loss": 1.0562,
+      "step": 10654
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4506795221316212e-05,
+      "loss": 0.9792,
+      "step": 10655
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4505789398785674e-05,
+      "loss": 0.8943,
+      "step": 10656
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4504783519055714e-05,
+      "loss": 0.9846,
+      "step": 10657
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4503777582139102e-05,
+      "loss": 0.9602,
+      "step": 10658
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4502771588048612e-05,
+      "loss": 0.9823,
+      "step": 10659
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4501765536797003e-05,
+      "loss": 1.0943,
+      "step": 10660
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.450075942839706e-05,
+      "loss": 1.0187,
+      "step": 10661
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4499753262861548e-05,
+      "loss": 1.0428,
+      "step": 10662
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449874704020324e-05,
+      "loss": 1.013,
+      "step": 10663
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449774076043491e-05,
+      "loss": 1.0792,
+      "step": 10664
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4496734423569335e-05,
+      "loss": 1.0224,
+      "step": 10665
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.449572802961929e-05,
+      "loss": 0.8995,
+      "step": 10666
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4494721578597543e-05,
+      "loss": 1.0483,
+      "step": 10667
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4493715070516881e-05,
+      "loss": 0.8233,
+      "step": 10668
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4492708505390078e-05,
+      "loss": 1.074,
+      "step": 10669
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4491701883229909e-05,
+      "loss": 1.031,
+      "step": 10670
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4490695204049154e-05,
+      "loss": 1.1502,
+      "step": 10671
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4489688467860595e-05,
+      "loss": 1.0731,
+      "step": 10672
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4488681674677008e-05,
+      "loss": 1.1196,
+      "step": 10673
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4487674824511175e-05,
+      "loss": 0.9042,
+      "step": 10674
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4486667917375882e-05,
+      "loss": 1.0892,
+      "step": 10675
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448566095328391e-05,
+      "loss": 1.0002,
+      "step": 10676
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4484653932248036e-05,
+      "loss": 1.1394,
+      "step": 10677
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4483646854281048e-05,
+      "loss": 0.9249,
+      "step": 10678
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.448263971939573e-05,
+      "loss": 1.0636,
+      "step": 10679
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4481632527604869e-05,
+      "loss": 1.0496,
+      "step": 10680
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4480625278921248e-05,
+      "loss": 1.1154,
+      "step": 10681
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4479617973357658e-05,
+      "loss": 0.9327,
+      "step": 10682
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4478610610926883e-05,
+      "loss": 1.046,
+      "step": 10683
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4477603191641709e-05,
+      "loss": 1.0525,
+      "step": 10684
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4476595715514932e-05,
+      "loss": 1.1398,
+      "step": 10685
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4475588182559332e-05,
+      "loss": 1.0697,
+      "step": 10686
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.447458059278771e-05,
+      "loss": 1.1499,
+      "step": 10687
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4473572946212846e-05,
+      "loss": 1.0933,
+      "step": 10688
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4472565242847541e-05,
+      "loss": 1.0057,
+      "step": 10689
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4471557482704584e-05,
+      "loss": 1.1002,
+      "step": 10690
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4470549665796762e-05,
+      "loss": 0.8643,
+      "step": 10691
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4469541792136879e-05,
+      "loss": 1.188,
+      "step": 10692
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4468533861737726e-05,
+      "loss": 0.9909,
+      "step": 10693
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4467525874612096e-05,
+      "loss": 1.0377,
+      "step": 10694
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4466517830772784e-05,
+      "loss": 1.1382,
+      "step": 10695
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4465509730232595e-05,
+      "loss": 1.0205,
+      "step": 10696
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4464501573004319e-05,
+      "loss": 0.9533,
+      "step": 10697
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4463493359100754e-05,
+      "loss": 1.0379,
+      "step": 10698
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44624850885347e-05,
+      "loss": 0.9669,
+      "step": 10699
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4461476761318962e-05,
+      "loss": 1.199,
+      "step": 10700
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4460468377466332e-05,
+      "loss": 1.0413,
+      "step": 10701
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4459459936989615e-05,
+      "loss": 0.9825,
+      "step": 10702
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4458451439901618e-05,
+      "loss": 0.9709,
+      "step": 10703
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4457442886215134e-05,
+      "loss": 1.1454,
+      "step": 10704
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445643427594297e-05,
+      "loss": 0.9825,
+      "step": 10705
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4455425609097932e-05,
+      "loss": 0.9571,
+      "step": 10706
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4454416885692824e-05,
+      "loss": 0.9052,
+      "step": 10707
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.445340810574045e-05,
+      "loss": 1.0946,
+      "step": 10708
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4452399269253615e-05,
+      "loss": 0.9847,
+      "step": 10709
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4451390376245131e-05,
+      "loss": 1.0289,
+      "step": 10710
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.44503814267278e-05,
+      "loss": 1.1278,
+      "step": 10711
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4449372420714432e-05,
+      "loss": 1.0043,
+      "step": 10712
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4448363358217839e-05,
+      "loss": 1.0959,
+      "step": 10713
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4447354239250823e-05,
+      "loss": 0.9512,
+      "step": 10714
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4446345063826202e-05,
+      "loss": 0.9785,
+      "step": 10715
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4445335831956783e-05,
+      "loss": 1.0522,
+      "step": 10716
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4444326543655384e-05,
+      "loss": 0.8518,
+      "step": 10717
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.444331719893481e-05,
+      "loss": 1.14,
+      "step": 10718
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4442307797807878e-05,
+      "loss": 1.0179,
+      "step": 10719
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4441298340287402e-05,
+      "loss": 0.9428,
+      "step": 10720
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4440288826386196e-05,
+      "loss": 0.9452,
+      "step": 10721
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4439279256117071e-05,
+      "loss": 1.0835,
+      "step": 10722
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4438269629492852e-05,
+      "loss": 0.9975,
+      "step": 10723
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443725994652635e-05,
+      "loss": 1.1115,
+      "step": 10724
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4436250207230384e-05,
+      "loss": 0.9701,
+      "step": 10725
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4435240411617773e-05,
+      "loss": 0.9938,
+      "step": 10726
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4434230559701335e-05,
+      "loss": 1.0625,
+      "step": 10727
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443322065149389e-05,
+      "loss": 0.9502,
+      "step": 10728
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4432210687008257e-05,
+      "loss": 0.945,
+      "step": 10729
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.443120066625726e-05,
+      "loss": 0.9089,
+      "step": 10730
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4430190589253719e-05,
+      "loss": 1.1744,
+      "step": 10731
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4429180456010456e-05,
+      "loss": 0.9521,
+      "step": 10732
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4428170266540296e-05,
+      "loss": 0.9437,
+      "step": 10733
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4427160020856064e-05,
+      "loss": 0.9513,
+      "step": 10734
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442614971897058e-05,
+      "loss": 0.995,
+      "step": 10735
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4425139360896673e-05,
+      "loss": 1.0344,
+      "step": 10736
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4424128946647169e-05,
+      "loss": 1.0992,
+      "step": 10737
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4423118476234896e-05,
+      "loss": 1.1005,
+      "step": 10738
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4422107949672679e-05,
+      "loss": 0.9235,
+      "step": 10739
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4421097366973347e-05,
+      "loss": 0.9076,
+      "step": 10740
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.442008672814973e-05,
+      "loss": 0.9979,
+      "step": 10741
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4419076033214655e-05,
+      "loss": 1.0239,
+      "step": 10742
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4418065282180955e-05,
+      "loss": 0.9877,
+      "step": 10743
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4417054475061463e-05,
+      "loss": 1.0598,
+      "step": 10744
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4416043611869006e-05,
+      "loss": 1.2144,
+      "step": 10745
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441503269261642e-05,
+      "loss": 0.9609,
+      "step": 10746
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4414021717316535e-05,
+      "loss": 0.9462,
+      "step": 10747
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.441301068598219e-05,
+      "loss": 1.1122,
+      "step": 10748
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4411999598626216e-05,
+      "loss": 1.0873,
+      "step": 10749
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4410988455261448e-05,
+      "loss": 1.0817,
+      "step": 10750
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4409977255900725e-05,
+      "loss": 1.0207,
+      "step": 10751
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4408966000556883e-05,
+      "loss": 1.0717,
+      "step": 10752
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4407954689242756e-05,
+      "loss": 1.0023,
+      "step": 10753
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4406943321971186e-05,
+      "loss": 1.1139,
+      "step": 10754
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4405931898755012e-05,
+      "loss": 1.0271,
+      "step": 10755
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.440492041960707e-05,
+      "loss": 1.1722,
+      "step": 10756
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4403908884540204e-05,
+      "loss": 1.0012,
+      "step": 10757
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4402897293567255e-05,
+      "loss": 1.0911,
+      "step": 10758
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4401885646701064e-05,
+      "loss": 1.0687,
+      "step": 10759
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4400873943954473e-05,
+      "loss": 0.9162,
+      "step": 10760
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4399862185340325e-05,
+      "loss": 1.1169,
+      "step": 10761
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4398850370871467e-05,
+      "loss": 1.0449,
+      "step": 10762
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4397838500560737e-05,
+      "loss": 1.1192,
+      "step": 10763
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4396826574420986e-05,
+      "loss": 1.1392,
+      "step": 10764
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4395814592465059e-05,
+      "loss": 1.1715,
+      "step": 10765
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4394802554705802e-05,
+      "loss": 1.0212,
+      "step": 10766
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4393790461156063e-05,
+      "loss": 1.0617,
+      "step": 10767
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.439277831182869e-05,
+      "loss": 1.0784,
+      "step": 10768
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4391766106736533e-05,
+      "loss": 1.1361,
+      "step": 10769
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4390753845892439e-05,
+      "loss": 0.9678,
+      "step": 10770
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4389741529309258e-05,
+      "loss": 1.0645,
+      "step": 10771
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4388729156999843e-05,
+      "loss": 1.2306,
+      "step": 10772
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4387716728977047e-05,
+      "loss": 1.0057,
+      "step": 10773
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.438670424525372e-05,
+      "loss": 1.1298,
+      "step": 10774
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4385691705842717e-05,
+      "loss": 1.0825,
+      "step": 10775
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4384679110756889e-05,
+      "loss": 1.0195,
+      "step": 10776
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4383666460009095e-05,
+      "loss": 1.0631,
+      "step": 10777
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4382653753612186e-05,
+      "loss": 1.0378,
+      "step": 10778
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 1.4381640991579017e-05,
+      "loss": 0.9667,
+      "step": 10779
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4380628173922452e-05,
+      "loss": 0.9279,
+      "step": 10780
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4379615300655339e-05,
+      "loss": 1.0448,
+      "step": 10781
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4378602371790542e-05,
+      "loss": 0.9346,
+      "step": 10782
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437758938734092e-05,
+      "loss": 0.982,
+      "step": 10783
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4376576347319331e-05,
+      "loss": 1.103,
+      "step": 10784
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4375563251738633e-05,
+      "loss": 1.1518,
+      "step": 10785
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4374550100611688e-05,
+      "loss": 1.0089,
+      "step": 10786
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4373536893951363e-05,
+      "loss": 1.0423,
+      "step": 10787
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.437252363177051e-05,
+      "loss": 1.0529,
+      "step": 10788
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4371510314082002e-05,
+      "loss": 0.9742,
+      "step": 10789
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4370496940898699e-05,
+      "loss": 1.0522,
+      "step": 10790
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4369483512233461e-05,
+      "loss": 0.9713,
+      "step": 10791
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4368470028099162e-05,
+      "loss": 0.9628,
+      "step": 10792
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.436745648850866e-05,
+      "loss": 1.1372,
+      "step": 10793
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4366442893474825e-05,
+      "loss": 1.0719,
+      "step": 10794
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4365429243010523e-05,
+      "loss": 1.0932,
+      "step": 10795
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4364415537128624e-05,
+      "loss": 0.9185,
+      "step": 10796
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4363401775841996e-05,
+      "loss": 1.2239,
+      "step": 10797
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4362387959163504e-05,
+      "loss": 1.0057,
+      "step": 10798
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4361374087106026e-05,
+      "loss": 0.9493,
+      "step": 10799
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4360360159682425e-05,
+      "loss": 1.0189,
+      "step": 10800
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435934617690558e-05,
+      "loss": 1.0884,
+      "step": 10801
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4358332138788353e-05,
+      "loss": 1.0628,
+      "step": 10802
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4357318045343628e-05,
+      "loss": 0.8576,
+      "step": 10803
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4356303896584271e-05,
+      "loss": 1.0863,
+      "step": 10804
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4355289692523158e-05,
+      "loss": 1.0161,
+      "step": 10805
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4354275433173168e-05,
+      "loss": 0.9416,
+      "step": 10806
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435326111854717e-05,
+      "loss": 1.1149,
+      "step": 10807
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4352246748658048e-05,
+      "loss": 0.9731,
+      "step": 10808
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4351232323518669e-05,
+      "loss": 1.0035,
+      "step": 10809
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.435021784314192e-05,
+      "loss": 0.9325,
+      "step": 10810
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4349203307540678e-05,
+      "loss": 0.9442,
+      "step": 10811
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4348188716727816e-05,
+      "loss": 1.1064,
+      "step": 10812
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4347174070716219e-05,
+      "loss": 1.0497,
+      "step": 10813
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4346159369518768e-05,
+      "loss": 1.0891,
+      "step": 10814
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4345144613148341e-05,
+      "loss": 1.1518,
+      "step": 10815
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4344129801617824e-05,
+      "loss": 1.0705,
+      "step": 10816
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4343114934940098e-05,
+      "loss": 1.0347,
+      "step": 10817
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4342100013128046e-05,
+      "loss": 1.0018,
+      "step": 10818
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.434108503619455e-05,
+      "loss": 1.0954,
+      "step": 10819
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4340070004152499e-05,
+      "loss": 1.0285,
+      "step": 10820
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4339054917014774e-05,
+      "loss": 1.193,
+      "step": 10821
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4338039774794263e-05,
+      "loss": 1.1273,
+      "step": 10822
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4337024577503856e-05,
+      "loss": 0.9356,
+      "step": 10823
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4336009325156437e-05,
+      "loss": 0.948,
+      "step": 10824
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4334994017764898e-05,
+      "loss": 1.045,
+      "step": 10825
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.433397865534212e-05,
+      "loss": 1.0349,
+      "step": 10826
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4332963237901001e-05,
+      "loss": 1.1133,
+      "step": 10827
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4331947765454426e-05,
+      "loss": 1.0337,
+      "step": 10828
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4330932238015289e-05,
+      "loss": 0.9501,
+      "step": 10829
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4329916655596483e-05,
+      "loss": 0.9729,
+      "step": 10830
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4328901018210899e-05,
+      "loss": 0.8875,
+      "step": 10831
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4327885325871426e-05,
+      "loss": 0.9083,
+      "step": 10832
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4326869578590963e-05,
+      "loss": 1.019,
+      "step": 10833
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4325853776382406e-05,
+      "loss": 1.1039,
+      "step": 10834
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4324837919258645e-05,
+      "loss": 0.9932,
+      "step": 10835
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4323822007232577e-05,
+      "loss": 1.032,
+      "step": 10836
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.43228060403171e-05,
+      "loss": 1.0754,
+      "step": 10837
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4321790018525115e-05,
+      "loss": 1.0933,
+      "step": 10838
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4320773941869512e-05,
+      "loss": 1.0214,
+      "step": 10839
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4319757810363196e-05,
+      "loss": 1.0916,
+      "step": 10840
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4318741624019065e-05,
+      "loss": 1.0116,
+      "step": 10841
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4317725382850019e-05,
+      "loss": 0.9777,
+      "step": 10842
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4316709086868956e-05,
+      "loss": 1.2775,
+      "step": 10843
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4315692736088779e-05,
+      "loss": 1.0791,
+      "step": 10844
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4314676330522396e-05,
+      "loss": 1.0232,
+      "step": 10845
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4313659870182703e-05,
+      "loss": 1.0296,
+      "step": 10846
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4312643355082605e-05,
+      "loss": 1.1589,
+      "step": 10847
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.431162678523501e-05,
+      "loss": 0.9674,
+      "step": 10848
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4310610160652818e-05,
+      "loss": 1.094,
+      "step": 10849
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4309593481348936e-05,
+      "loss": 1.085,
+      "step": 10850
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430857674733627e-05,
+      "loss": 1.0351,
+      "step": 10851
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.430755995862773e-05,
+      "loss": 1.0833,
+      "step": 10852
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4306543115236222e-05,
+      "loss": 1.0137,
+      "step": 10853
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4305526217174652e-05,
+      "loss": 0.9709,
+      "step": 10854
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4304509264455937e-05,
+      "loss": 1.0404,
+      "step": 10855
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4303492257092978e-05,
+      "loss": 1.091,
+      "step": 10856
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4302475195098688e-05,
+      "loss": 1.0221,
+      "step": 10857
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4301458078485978e-05,
+      "loss": 0.9918,
+      "step": 10858
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4300440907267768e-05,
+      "loss": 1.1067,
+      "step": 10859
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429942368145696e-05,
+      "loss": 1.1429,
+      "step": 10860
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4298406401066468e-05,
+      "loss": 1.1289,
+      "step": 10861
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4297389066109214e-05,
+      "loss": 1.1485,
+      "step": 10862
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4296371676598104e-05,
+      "loss": 0.9732,
+      "step": 10863
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.429535423254606e-05,
+      "loss": 1.0542,
+      "step": 10864
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4294336733965995e-05,
+      "loss": 1.1323,
+      "step": 10865
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4293319180870824e-05,
+      "loss": 1.0556,
+      "step": 10866
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4292301573273469e-05,
+      "loss": 0.9595,
+      "step": 10867
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4291283911186843e-05,
+      "loss": 0.9864,
+      "step": 10868
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4290266194623871e-05,
+      "loss": 1.1728,
+      "step": 10869
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.428924842359747e-05,
+      "loss": 1.0461,
+      "step": 10870
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4288230598120556e-05,
+      "loss": 0.96,
+      "step": 10871
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4287212718206057e-05,
+      "loss": 1.0489,
+      "step": 10872
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4286194783866889e-05,
+      "loss": 1.0926,
+      "step": 10873
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4285176795115976e-05,
+      "loss": 1.0302,
+      "step": 10874
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4284158751966241e-05,
+      "loss": 1.125,
+      "step": 10875
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4283140654430611e-05,
+      "loss": 1.0812,
+      "step": 10876
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4282122502522005e-05,
+      "loss": 1.0028,
+      "step": 10877
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4281104296253352e-05,
+      "loss": 1.0737,
+      "step": 10878
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4280086035637575e-05,
+      "loss": 1.0719,
+      "step": 10879
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4279067720687602e-05,
+      "loss": 1.0861,
+      "step": 10880
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4278049351416359e-05,
+      "loss": 1.0679,
+      "step": 10881
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4277030927836776e-05,
+      "loss": 0.9858,
+      "step": 10882
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4276012449961782e-05,
+      "loss": 1.1344,
+      "step": 10883
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4274993917804306e-05,
+      "loss": 1.0793,
+      "step": 10884
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427397533137727e-05,
+      "loss": 1.0208,
+      "step": 10885
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4272956690693616e-05,
+      "loss": 0.9795,
+      "step": 10886
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.427193799576627e-05,
+      "loss": 1.0393,
+      "step": 10887
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4270919246608158e-05,
+      "loss": 1.1178,
+      "step": 10888
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.426990044323223e-05,
+      "loss": 1.1084,
+      "step": 10889
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4268881585651402e-05,
+      "loss": 1.1258,
+      "step": 10890
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4267862673878617e-05,
+      "loss": 1.0244,
+      "step": 10891
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4266843707926803e-05,
+      "loss": 1.1525,
+      "step": 10892
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42658246878089e-05,
+      "loss": 1.0311,
+      "step": 10893
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4264805613537849e-05,
+      "loss": 1.0112,
+      "step": 10894
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4263786485126574e-05,
+      "loss": 1.1449,
+      "step": 10895
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4262767302588026e-05,
+      "loss": 1.0543,
+      "step": 10896
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4261748065935134e-05,
+      "loss": 1.0627,
+      "step": 10897
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4260728775180844e-05,
+      "loss": 0.9663,
+      "step": 10898
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4259709430338087e-05,
+      "loss": 0.918,
+      "step": 10899
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4258690031419808e-05,
+      "loss": 1.033,
+      "step": 10900
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4257670578438949e-05,
+      "loss": 0.9505,
+      "step": 10901
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4256651071408446e-05,
+      "loss": 0.926,
+      "step": 10902
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4255631510341253e-05,
+      "loss": 1.1483,
+      "step": 10903
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4254611895250304e-05,
+      "loss": 1.0174,
+      "step": 10904
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4253592226148539e-05,
+      "loss": 0.964,
+      "step": 10905
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4252572503048912e-05,
+      "loss": 1.1142,
+      "step": 10906
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4251552725964361e-05,
+      "loss": 1.0327,
+      "step": 10907
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4250532894907837e-05,
+      "loss": 1.1938,
+      "step": 10908
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424951300989228e-05,
+      "loss": 0.9927,
+      "step": 10909
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4248493070930642e-05,
+      "loss": 1.0314,
+      "step": 10910
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4247473078035871e-05,
+      "loss": 1.0595,
+      "step": 10911
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4246453031220912e-05,
+      "loss": 0.9314,
+      "step": 10912
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424543293049872e-05,
+      "loss": 1.0497,
+      "step": 10913
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4244412775882235e-05,
+      "loss": 0.8634,
+      "step": 10914
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4243392567384418e-05,
+      "loss": 0.9921,
+      "step": 10915
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.424237230501821e-05,
+      "loss": 0.9865,
+      "step": 10916
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4241351988796574e-05,
+      "loss": 1.0951,
+      "step": 10917
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4240331618732456e-05,
+      "loss": 1.0353,
+      "step": 10918
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4239311194838809e-05,
+      "loss": 1.1173,
+      "step": 10919
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.423829071712859e-05,
+      "loss": 1.0717,
+      "step": 10920
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4237270185614751e-05,
+      "loss": 0.9941,
+      "step": 10921
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4236249600310246e-05,
+      "loss": 1.0639,
+      "step": 10922
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4235228961228034e-05,
+      "loss": 1.1599,
+      "step": 10923
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4234208268381073e-05,
+      "loss": 0.9559,
+      "step": 10924
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4233187521782317e-05,
+      "loss": 1.0531,
+      "step": 10925
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4232166721444722e-05,
+      "loss": 1.131,
+      "step": 10926
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4231145867381253e-05,
+      "loss": 1.0039,
+      "step": 10927
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4230124959604864e-05,
+      "loss": 0.9597,
+      "step": 10928
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.422910399812852e-05,
+      "loss": 1.0829,
+      "step": 10929
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4228082982965177e-05,
+      "loss": 1.0315,
+      "step": 10930
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42270619141278e-05,
+      "loss": 1.0064,
+      "step": 10931
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4226040791629349e-05,
+      "loss": 1.0005,
+      "step": 10932
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4225019615482785e-05,
+      "loss": 1.2176,
+      "step": 10933
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4223998385701077e-05,
+      "loss": 1.0826,
+      "step": 10934
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4222977102297183e-05,
+      "loss": 1.1792,
+      "step": 10935
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4221955765284072e-05,
+      "loss": 1.0328,
+      "step": 10936
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4220934374674708e-05,
+      "loss": 1.0851,
+      "step": 10937
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4219912930482061e-05,
+      "loss": 0.9739,
+      "step": 10938
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4218891432719093e-05,
+      "loss": 1.1072,
+      "step": 10939
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4217869881398766e-05,
+      "loss": 1.0493,
+      "step": 10940
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4216848276534064e-05,
+      "loss": 1.0042,
+      "step": 10941
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4215826618137943e-05,
+      "loss": 0.9593,
+      "step": 10942
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4214804906223377e-05,
+      "loss": 1.0188,
+      "step": 10943
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4213783140803335e-05,
+      "loss": 1.094,
+      "step": 10944
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4212761321890791e-05,
+      "loss": 1.0485,
+      "step": 10945
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4211739449498714e-05,
+      "loss": 1.0532,
+      "step": 10946
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4210717523640077e-05,
+      "loss": 1.0642,
+      "step": 10947
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4209695544327851e-05,
+      "loss": 1.1069,
+      "step": 10948
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4208673511575013e-05,
+      "loss": 1.0168,
+      "step": 10949
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4207651425394536e-05,
+      "loss": 0.9451,
+      "step": 10950
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4206629285799394e-05,
+      "loss": 0.9835,
+      "step": 10951
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4205607092802565e-05,
+      "loss": 1.0681,
+      "step": 10952
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4204584846417023e-05,
+      "loss": 0.9808,
+      "step": 10953
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4203562546655745e-05,
+      "loss": 1.0908,
+      "step": 10954
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.420254019353171e-05,
+      "loss": 1.0041,
+      "step": 10955
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.42015177870579e-05,
+      "loss": 1.1403,
+      "step": 10956
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4200495327247285e-05,
+      "loss": 1.0436,
+      "step": 10957
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4199472814112851e-05,
+      "loss": 0.9749,
+      "step": 10958
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4198450247667582e-05,
+      "loss": 1.0311,
+      "step": 10959
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419742762792445e-05,
+      "loss": 0.9841,
+      "step": 10960
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4196404954896443e-05,
+      "loss": 0.9539,
+      "step": 10961
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.419538222859654e-05,
+      "loss": 1.0082,
+      "step": 10962
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4194359449037729e-05,
+      "loss": 0.9254,
+      "step": 10963
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4193336616232988e-05,
+      "loss": 0.9866,
+      "step": 10964
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4192313730195308e-05,
+      "loss": 0.9727,
+      "step": 10965
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4191290790937669e-05,
+      "loss": 0.9666,
+      "step": 10966
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4190267798473057e-05,
+      "loss": 1.1375,
+      "step": 10967
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4189244752814462e-05,
+      "loss": 1.1075,
+      "step": 10968
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418822165397487e-05,
+      "loss": 0.9851,
+      "step": 10969
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4187198501967262e-05,
+      "loss": 1.0804,
+      "step": 10970
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4186175296804639e-05,
+      "loss": 1.0412,
+      "step": 10971
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4185152038499982e-05,
+      "loss": 1.0201,
+      "step": 10972
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4184128727066282e-05,
+      "loss": 1.0288,
+      "step": 10973
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4183105362516529e-05,
+      "loss": 0.9511,
+      "step": 10974
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4182081944863719e-05,
+      "loss": 1.06,
+      "step": 10975
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.418105847412084e-05,
+      "loss": 1.023,
+      "step": 10976
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4180034950300883e-05,
+      "loss": 0.8664,
+      "step": 10977
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4179011373416844e-05,
+      "loss": 0.996,
+      "step": 10978
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4177987743481716e-05,
+      "loss": 1.1234,
+      "step": 10979
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4176964060508496e-05,
+      "loss": 1.067,
+      "step": 10980
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4175940324510174e-05,
+      "loss": 0.9176,
+      "step": 10981
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4174916535499753e-05,
+      "loss": 1.0404,
+      "step": 10982
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4173892693490222e-05,
+      "loss": 1.0222,
+      "step": 10983
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4172868798494585e-05,
+      "loss": 0.8511,
+      "step": 10984
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4171844850525836e-05,
+      "loss": 1.0848,
+      "step": 10985
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4170820849596973e-05,
+      "loss": 0.9125,
+      "step": 10986
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4169796795721001e-05,
+      "loss": 1.0375,
+      "step": 10987
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4168772688910914e-05,
+      "loss": 1.0402,
+      "step": 10988
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4167748529179715e-05,
+      "loss": 1.0846,
+      "step": 10989
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4166724316540409e-05,
+      "loss": 1.0715,
+      "step": 10990
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4165700051005989e-05,
+      "loss": 1.0446,
+      "step": 10991
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4164675732589464e-05,
+      "loss": 1.042,
+      "step": 10992
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4163651361303834e-05,
+      "loss": 1.0931,
+      "step": 10993
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4162626937162111e-05,
+      "loss": 1.0823,
+      "step": 10994
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4161602460177288e-05,
+      "loss": 0.8412,
+      "step": 10995
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.416057793036238e-05,
+      "loss": 0.9785,
+      "step": 10996
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4159553347730393e-05,
+      "loss": 1.1399,
+      "step": 10997
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4158528712294324e-05,
+      "loss": 0.8981,
+      "step": 10998
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4157504024067188e-05,
+      "loss": 0.9811,
+      "step": 10999
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4156479283061992e-05,
+      "loss": 1.0721,
+      "step": 11000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4155454489291746e-05,
+      "loss": 1.1943,
+      "step": 11001
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4154429642769457e-05,
+      "loss": 0.9521,
+      "step": 11002
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4153404743508135e-05,
+      "loss": 1.0071,
+      "step": 11003
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4152379791520794e-05,
+      "loss": 1.1227,
+      "step": 11004
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4151354786820438e-05,
+      "loss": 1.034,
+      "step": 11005
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4150329729420086e-05,
+      "loss": 0.9299,
+      "step": 11006
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414930461933275e-05,
+      "loss": 1.0155,
+      "step": 11007
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4148279456571444e-05,
+      "loss": 1.0811,
+      "step": 11008
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4147254241149176e-05,
+      "loss": 0.982,
+      "step": 11009
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4146228973078968e-05,
+      "loss": 1.154,
+      "step": 11010
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4145203652373832e-05,
+      "loss": 1.1421,
+      "step": 11011
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414417827904678e-05,
+      "loss": 1.001,
+      "step": 11012
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414315285311084e-05,
+      "loss": 1.104,
+      "step": 11013
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414212737457902e-05,
+      "loss": 1.0138,
+      "step": 11014
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4141101843464342e-05,
+      "loss": 0.9587,
+      "step": 11015
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.414007625977982e-05,
+      "loss": 1.0032,
+      "step": 11016
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4139050623538481e-05,
+      "loss": 0.9769,
+      "step": 11017
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4138024934753341e-05,
+      "loss": 1.0259,
+      "step": 11018
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4136999193437417e-05,
+      "loss": 0.9611,
+      "step": 11019
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4135973399603737e-05,
+      "loss": 0.9459,
+      "step": 11020
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4134947553265322e-05,
+      "loss": 1.1407,
+      "step": 11021
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.413392165443519e-05,
+      "loss": 1.0486,
+      "step": 11022
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4132895703126365e-05,
+      "loss": 0.895,
+      "step": 11023
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4131869699351882e-05,
+      "loss": 0.9902,
+      "step": 11024
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4130843643124754e-05,
+      "loss": 1.1141,
+      "step": 11025
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4129817534458006e-05,
+      "loss": 1.0538,
+      "step": 11026
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4128791373364669e-05,
+      "loss": 1.0056,
+      "step": 11027
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4127765159857772e-05,
+      "loss": 1.2242,
+      "step": 11028
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4126738893950338e-05,
+      "loss": 1.0506,
+      "step": 11029
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4125712575655392e-05,
+      "loss": 0.9851,
+      "step": 11030
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4124686204985973e-05,
+      "loss": 1.1759,
+      "step": 11031
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4123659781955104e-05,
+      "loss": 1.0951,
+      "step": 11032
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4122633306575811e-05,
+      "loss": 1.0383,
+      "step": 11033
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4121606778861132e-05,
+      "loss": 0.888,
+      "step": 11034
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4120580198824095e-05,
+      "loss": 0.9421,
+      "step": 11035
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411955356647773e-05,
+      "loss": 1.0552,
+      "step": 11036
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4118526881835075e-05,
+      "loss": 0.9782,
+      "step": 11037
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4117500144909162e-05,
+      "loss": 0.9565,
+      "step": 11038
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4116473355713025e-05,
+      "loss": 1.0354,
+      "step": 11039
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4115446514259692e-05,
+      "loss": 1.0067,
+      "step": 11040
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411441962056221e-05,
+      "loss": 1.0448,
+      "step": 11041
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4113392674633603e-05,
+      "loss": 0.9992,
+      "step": 11042
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4112365676486919e-05,
+      "loss": 0.9772,
+      "step": 11043
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4111338626135186e-05,
+      "loss": 1.066,
+      "step": 11044
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.411031152359145e-05,
+      "loss": 1.1786,
+      "step": 11045
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4109284368868747e-05,
+      "loss": 1.1529,
+      "step": 11046
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410825716198011e-05,
+      "loss": 1.0879,
+      "step": 11047
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410722990293859e-05,
+      "loss": 1.1104,
+      "step": 11048
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4106202591757216e-05,
+      "loss": 1.0456,
+      "step": 11049
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410517522844904e-05,
+      "loss": 1.0968,
+      "step": 11050
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4104147813027096e-05,
+      "loss": 1.0128,
+      "step": 11051
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4103120345504433e-05,
+      "loss": 0.9994,
+      "step": 11052
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.410209282589409e-05,
+      "loss": 0.9229,
+      "step": 11053
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4101065254209111e-05,
+      "loss": 1.0306,
+      "step": 11054
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4100037630462547e-05,
+      "loss": 1.0023,
+      "step": 11055
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4099009954667433e-05,
+      "loss": 1.0281,
+      "step": 11056
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4097982226836821e-05,
+      "loss": 1.1501,
+      "step": 11057
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409695444698376e-05,
+      "loss": 0.9157,
+      "step": 11058
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4095926615121294e-05,
+      "loss": 0.994,
+      "step": 11059
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.409489873126247e-05,
+      "loss": 1.0926,
+      "step": 11060
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4093870795420337e-05,
+      "loss": 0.9938,
+      "step": 11061
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4092842807607948e-05,
+      "loss": 1.0158,
+      "step": 11062
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4091814767838348e-05,
+      "loss": 0.918,
+      "step": 11063
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4090786676124588e-05,
+      "loss": 1.0967,
+      "step": 11064
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4089758532479722e-05,
+      "loss": 1.0288,
+      "step": 11065
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4088730336916804e-05,
+      "loss": 1.0846,
+      "step": 11066
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 1.4087702089448883e-05,
+      "loss": 0.9409,
+      "step": 11067
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4086673790089012e-05,
+      "loss": 0.9647,
+      "step": 11068
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4085645438850244e-05,
+      "loss": 0.9801,
+      "step": 11069
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4084617035745636e-05,
+      "loss": 0.9222,
+      "step": 11070
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4083588580788243e-05,
+      "loss": 0.9481,
+      "step": 11071
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.408256007399112e-05,
+      "loss": 1.0825,
+      "step": 11072
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4081531515367326e-05,
+      "loss": 1.1425,
+      "step": 11073
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4080502904929913e-05,
+      "loss": 0.9505,
+      "step": 11074
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4079474242691941e-05,
+      "loss": 1.003,
+      "step": 11075
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4078445528666473e-05,
+      "loss": 1.0118,
+      "step": 11076
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4077416762866562e-05,
+      "loss": 1.0033,
+      "step": 11077
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4076387945305272e-05,
+      "loss": 0.9592,
+      "step": 11078
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407535907599566e-05,
+      "loss": 0.8704,
+      "step": 11079
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4074330154950792e-05,
+      "loss": 1.087,
+      "step": 11080
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4073301182183725e-05,
+      "loss": 0.9792,
+      "step": 11081
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4072272157707522e-05,
+      "loss": 0.9827,
+      "step": 11082
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407124308153525e-05,
+      "loss": 1.0294,
+      "step": 11083
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.407021395367997e-05,
+      "loss": 1.1882,
+      "step": 11084
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4069184774154744e-05,
+      "loss": 1.0732,
+      "step": 11085
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4068155542972641e-05,
+      "loss": 0.8844,
+      "step": 11086
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4067126260146726e-05,
+      "loss": 0.9786,
+      "step": 11087
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4066096925690065e-05,
+      "loss": 1.0539,
+      "step": 11088
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4065067539615722e-05,
+      "loss": 1.2062,
+      "step": 11089
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4064038101936771e-05,
+      "loss": 1.0875,
+      "step": 11090
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4063008612666275e-05,
+      "loss": 1.0474,
+      "step": 11091
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4061979071817304e-05,
+      "loss": 1.1101,
+      "step": 11092
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.406094947940293e-05,
+      "loss": 0.9294,
+      "step": 11093
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4059919835436222e-05,
+      "loss": 1.1431,
+      "step": 11094
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405889013993025e-05,
+      "loss": 1.1866,
+      "step": 11095
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4057860392898086e-05,
+      "loss": 1.0505,
+      "step": 11096
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4056830594352803e-05,
+      "loss": 1.1484,
+      "step": 11097
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4055800744307475e-05,
+      "loss": 1.1945,
+      "step": 11098
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4054770842775171e-05,
+      "loss": 0.9469,
+      "step": 11099
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4053740889768971e-05,
+      "loss": 1.1821,
+      "step": 11100
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4052710885301947e-05,
+      "loss": 1.1259,
+      "step": 11101
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4051680829387176e-05,
+      "loss": 0.9331,
+      "step": 11102
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.405065072203773e-05,
+      "loss": 1.0884,
+      "step": 11103
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4049620563266693e-05,
+      "loss": 1.1363,
+      "step": 11104
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4048590353087135e-05,
+      "loss": 0.9845,
+      "step": 11105
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4047560091512138e-05,
+      "loss": 0.9793,
+      "step": 11106
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404652977855478e-05,
+      "loss": 0.9345,
+      "step": 11107
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4045499414228143e-05,
+      "loss": 0.9285,
+      "step": 11108
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4044468998545305e-05,
+      "loss": 1.0559,
+      "step": 11109
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4043438531519345e-05,
+      "loss": 0.9606,
+      "step": 11110
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.404240801316335e-05,
+      "loss": 1.007,
+      "step": 11111
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4041377443490394e-05,
+      "loss": 0.9408,
+      "step": 11112
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4040346822513562e-05,
+      "loss": 1.1594,
+      "step": 11113
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4039316150245944e-05,
+      "loss": 1.058,
+      "step": 11114
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4038285426700618e-05,
+      "loss": 0.9966,
+      "step": 11115
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403725465189067e-05,
+      "loss": 1.0726,
+      "step": 11116
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4036223825829182e-05,
+      "loss": 0.9845,
+      "step": 11117
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403519294852925e-05,
+      "loss": 0.8799,
+      "step": 11118
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.403416202000395e-05,
+      "loss": 1.1043,
+      "step": 11119
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4033131040266372e-05,
+      "loss": 1.134,
+      "step": 11120
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4032100009329608e-05,
+      "loss": 1.1123,
+      "step": 11121
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4031068927206743e-05,
+      "loss": 0.9758,
+      "step": 11122
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4030037793910867e-05,
+      "loss": 1.0655,
+      "step": 11123
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402900660945507e-05,
+      "loss": 1.0155,
+      "step": 11124
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402797537385244e-05,
+      "loss": 1.0568,
+      "step": 11125
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4026944087116073e-05,
+      "loss": 0.9597,
+      "step": 11126
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4025912749259058e-05,
+      "loss": 1.0943,
+      "step": 11127
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4024881360294487e-05,
+      "loss": 0.9111,
+      "step": 11128
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4023849920235457e-05,
+      "loss": 0.9154,
+      "step": 11129
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4022818429095053e-05,
+      "loss": 1.0269,
+      "step": 11130
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.402178688688638e-05,
+      "loss": 0.9932,
+      "step": 11131
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4020755293622528e-05,
+      "loss": 0.9164,
+      "step": 11132
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401972364931659e-05,
+      "loss": 0.9171,
+      "step": 11133
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4018691953981667e-05,
+      "loss": 1.0198,
+      "step": 11134
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4017660207630854e-05,
+      "loss": 1.039,
+      "step": 11135
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4016628410277251e-05,
+      "loss": 1.1191,
+      "step": 11136
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401559656193395e-05,
+      "loss": 1.0755,
+      "step": 11137
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4014564662614059e-05,
+      "loss": 0.9329,
+      "step": 11138
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4013532712330671e-05,
+      "loss": 0.9616,
+      "step": 11139
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4012500711096885e-05,
+      "loss": 1.0132,
+      "step": 11140
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401146865892581e-05,
+      "loss": 0.9909,
+      "step": 11141
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.401043655583054e-05,
+      "loss": 0.9683,
+      "step": 11142
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4009404401824181e-05,
+      "loss": 0.9928,
+      "step": 11143
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.400837219691983e-05,
+      "loss": 0.996,
+      "step": 11144
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4007339941130602e-05,
+      "loss": 1.0037,
+      "step": 11145
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4006307634469591e-05,
+      "loss": 1.132,
+      "step": 11146
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4005275276949903e-05,
+      "loss": 0.9066,
+      "step": 11147
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4004242868584647e-05,
+      "loss": 1.0495,
+      "step": 11148
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.400321040938693e-05,
+      "loss": 0.9743,
+      "step": 11149
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4002177899369851e-05,
+      "loss": 1.0979,
+      "step": 11150
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4001145338546525e-05,
+      "loss": 0.9961,
+      "step": 11151
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.4000112726930058e-05,
+      "loss": 1.0805,
+      "step": 11152
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399908006453356e-05,
+      "loss": 1.0223,
+      "step": 11153
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3998047351370131e-05,
+      "loss": 0.9483,
+      "step": 11154
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3997014587452897e-05,
+      "loss": 1.1201,
+      "step": 11155
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3995981772794956e-05,
+      "loss": 1.1065,
+      "step": 11156
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3994948907409422e-05,
+      "loss": 1.0557,
+      "step": 11157
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399391599130941e-05,
+      "loss": 1.0702,
+      "step": 11158
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399288302450803e-05,
+      "loss": 0.9769,
+      "step": 11159
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3991850007018397e-05,
+      "loss": 0.9804,
+      "step": 11160
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.399081693885362e-05,
+      "loss": 1.0357,
+      "step": 11161
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3989783820026821e-05,
+      "loss": 0.9796,
+      "step": 11162
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3988750650551108e-05,
+      "loss": 1.0702,
+      "step": 11163
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39877174304396e-05,
+      "loss": 0.958,
+      "step": 11164
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3986684159705415e-05,
+      "loss": 0.9811,
+      "step": 11165
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3985650838361667e-05,
+      "loss": 1.1092,
+      "step": 11166
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3984617466421475e-05,
+      "loss": 1.0422,
+      "step": 11167
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3983584043897956e-05,
+      "loss": 0.9987,
+      "step": 11168
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.398255057080423e-05,
+      "loss": 1.0332,
+      "step": 11169
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3981517047153418e-05,
+      "loss": 1.028,
+      "step": 11170
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3980483472958637e-05,
+      "loss": 1.0473,
+      "step": 11171
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3979449848233011e-05,
+      "loss": 1.0698,
+      "step": 11172
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397841617298966e-05,
+      "loss": 1.0885,
+      "step": 11173
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3977382447241706e-05,
+      "loss": 0.9829,
+      "step": 11174
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397634867100227e-05,
+      "loss": 1.0099,
+      "step": 11175
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3975314844284483e-05,
+      "loss": 0.9503,
+      "step": 11176
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.397428096710146e-05,
+      "loss": 1.1221,
+      "step": 11177
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3973247039466327e-05,
+      "loss": 0.9934,
+      "step": 11178
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3972213061392213e-05,
+      "loss": 0.982,
+      "step": 11179
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3971179032892246e-05,
+      "loss": 1.2042,
+      "step": 11180
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3970144953979545e-05,
+      "loss": 1.0858,
+      "step": 11181
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3969110824667244e-05,
+      "loss": 0.9439,
+      "step": 11182
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396807664496847e-05,
+      "loss": 1.0031,
+      "step": 11183
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3967042414896347e-05,
+      "loss": 1.0973,
+      "step": 11184
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3966008134464007e-05,
+      "loss": 0.9673,
+      "step": 11185
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3964973803684578e-05,
+      "loss": 0.9382,
+      "step": 11186
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3963939422571197e-05,
+      "loss": 1.0499,
+      "step": 11187
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.396290499113699e-05,
+      "loss": 1.0114,
+      "step": 11188
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3961870509395088e-05,
+      "loss": 1.0574,
+      "step": 11189
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3960835977358627e-05,
+      "loss": 1.035,
+      "step": 11190
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3959801395040734e-05,
+      "loss": 0.9857,
+      "step": 11191
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3958766762454547e-05,
+      "loss": 0.973,
+      "step": 11192
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.39577320796132e-05,
+      "loss": 1.1243,
+      "step": 11193
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395669734652983e-05,
+      "loss": 0.9491,
+      "step": 11194
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3955662563217569e-05,
+      "loss": 0.9749,
+      "step": 11195
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3954627729689554e-05,
+      "loss": 1.0456,
+      "step": 11196
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3953592845958923e-05,
+      "loss": 0.9813,
+      "step": 11197
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3952557912038811e-05,
+      "loss": 1.015,
+      "step": 11198
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.395152292794236e-05,
+      "loss": 0.9851,
+      "step": 11199
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3950487893682705e-05,
+      "loss": 1.0168,
+      "step": 11200
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3949452809272987e-05,
+      "loss": 0.9369,
+      "step": 11201
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3948417674726347e-05,
+      "loss": 1.1053,
+      "step": 11202
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3947382490055925e-05,
+      "loss": 1.1084,
+      "step": 11203
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3946347255274862e-05,
+      "loss": 1.1201,
+      "step": 11204
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3945311970396298e-05,
+      "loss": 0.9541,
+      "step": 11205
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3944276635433377e-05,
+      "loss": 0.8768,
+      "step": 11206
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3943241250399246e-05,
+      "loss": 0.9845,
+      "step": 11207
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3942205815307046e-05,
+      "loss": 0.9677,
+      "step": 11208
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.394117033016992e-05,
+      "loss": 1.1515,
+      "step": 11209
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3940134795001013e-05,
+      "loss": 0.9758,
+      "step": 11210
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3939099209813474e-05,
+      "loss": 1.1074,
+      "step": 11211
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3938063574620445e-05,
+      "loss": 1.0978,
+      "step": 11212
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3937027889435075e-05,
+      "loss": 1.13,
+      "step": 11213
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3935992154270514e-05,
+      "loss": 1.0776,
+      "step": 11214
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3934956369139909e-05,
+      "loss": 0.9364,
+      "step": 11215
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3933920534056406e-05,
+      "loss": 0.9924,
+      "step": 11216
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3932884649033158e-05,
+      "loss": 1.1062,
+      "step": 11217
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3931848714083315e-05,
+      "loss": 1.0874,
+      "step": 11218
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3930812729220027e-05,
+      "loss": 1.0728,
+      "step": 11219
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3929776694456443e-05,
+      "loss": 1.0343,
+      "step": 11220
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3928740609805718e-05,
+      "loss": 0.9882,
+      "step": 11221
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3927704475281008e-05,
+      "loss": 0.9013,
+      "step": 11222
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3926668290895457e-05,
+      "loss": 1.0797,
+      "step": 11223
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3925632056662229e-05,
+      "loss": 1.0411,
+      "step": 11224
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3924595772594474e-05,
+      "loss": 1.1379,
+      "step": 11225
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3923559438705346e-05,
+      "loss": 1.0405,
+      "step": 11226
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3922523055007999e-05,
+      "loss": 1.0479,
+      "step": 11227
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3921486621515596e-05,
+      "loss": 1.1786,
+      "step": 11228
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3920450138241294e-05,
+      "loss": 1.1118,
+      "step": 11229
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3919413605198244e-05,
+      "loss": 0.9632,
+      "step": 11230
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391837702239961e-05,
+      "loss": 1.1224,
+      "step": 11231
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.391734038985855e-05,
+      "loss": 0.9262,
+      "step": 11232
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3916303707588216e-05,
+      "loss": 1.022,
+      "step": 11233
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3915266975601784e-05,
+      "loss": 0.9987,
+      "step": 11234
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3914230193912403e-05,
+      "loss": 1.0344,
+      "step": 11235
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3913193362533239e-05,
+      "loss": 0.95,
+      "step": 11236
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3912156481477453e-05,
+      "loss": 1.0547,
+      "step": 11237
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3911119550758205e-05,
+      "loss": 1.0424,
+      "step": 11238
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3910082570388666e-05,
+      "loss": 0.9613,
+      "step": 11239
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3909045540381989e-05,
+      "loss": 0.9562,
+      "step": 11240
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3908008460751352e-05,
+      "loss": 0.9818,
+      "step": 11241
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3906971331509908e-05,
+      "loss": 1.1134,
+      "step": 11242
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3905934152670834e-05,
+      "loss": 0.9889,
+      "step": 11243
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3904896924247289e-05,
+      "loss": 1.1406,
+      "step": 11244
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390385964625244e-05,
+      "loss": 1.0535,
+      "step": 11245
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.390282231869946e-05,
+      "loss": 0.9757,
+      "step": 11246
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3901784941601512e-05,
+      "loss": 1.0085,
+      "step": 11247
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3900747514971774e-05,
+      "loss": 1.0449,
+      "step": 11248
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3899710038823404e-05,
+      "loss": 1.0445,
+      "step": 11249
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3898672513169583e-05,
+      "loss": 0.9978,
+      "step": 11250
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3897634938023471e-05,
+      "loss": 0.9876,
+      "step": 11251
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3896597313398249e-05,
+      "loss": 1.0189,
+      "step": 11252
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3895559639307088e-05,
+      "loss": 1.101,
+      "step": 11253
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3894521915763153e-05,
+      "loss": 1.1245,
+      "step": 11254
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3893484142779629e-05,
+      "loss": 1.0108,
+      "step": 11255
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3892446320369684e-05,
+      "loss": 0.9125,
+      "step": 11256
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3891408448546493e-05,
+      "loss": 1.0364,
+      "step": 11257
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.389037052732323e-05,
+      "loss": 0.9942,
+      "step": 11258
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3889332556713077e-05,
+      "loss": 1.1736,
+      "step": 11259
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3888294536729206e-05,
+      "loss": 0.9884,
+      "step": 11260
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388725646738479e-05,
+      "loss": 1.0002,
+      "step": 11261
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3886218348693017e-05,
+      "loss": 0.9913,
+      "step": 11262
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388518018066706e-05,
+      "loss": 0.9933,
+      "step": 11263
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3884141963320095e-05,
+      "loss": 1.0671,
+      "step": 11264
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.388310369666531e-05,
+      "loss": 1.0458,
+      "step": 11265
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3882065380715876e-05,
+      "loss": 1.058,
+      "step": 11266
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3881027015484984e-05,
+      "loss": 0.8463,
+      "step": 11267
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3879988600985805e-05,
+      "loss": 0.9598,
+      "step": 11268
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3878950137231533e-05,
+      "loss": 0.9497,
+      "step": 11269
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387791162423534e-05,
+      "loss": 1.108,
+      "step": 11270
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3876873062010417e-05,
+      "loss": 0.9858,
+      "step": 11271
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3875834450569945e-05,
+      "loss": 0.896,
+      "step": 11272
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3874795789927112e-05,
+      "loss": 1.0052,
+      "step": 11273
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3873757080095097e-05,
+      "loss": 0.8242,
+      "step": 11274
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387271832108709e-05,
+      "loss": 0.9786,
+      "step": 11275
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387167951291628e-05,
+      "loss": 0.9889,
+      "step": 11276
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.387064065559585e-05,
+      "loss": 0.9683,
+      "step": 11277
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3869601749138988e-05,
+      "loss": 1.0292,
+      "step": 11278
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3868562793558886e-05,
+      "loss": 1.0627,
+      "step": 11279
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3867523788868735e-05,
+      "loss": 0.95,
+      "step": 11280
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3866484735081717e-05,
+      "loss": 1.1265,
+      "step": 11281
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3865445632211027e-05,
+      "loss": 0.9684,
+      "step": 11282
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386440648026986e-05,
+      "loss": 0.9345,
+      "step": 11283
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3863367279271401e-05,
+      "loss": 0.8887,
+      "step": 11284
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3862328029228841e-05,
+      "loss": 0.9712,
+      "step": 11285
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386128873015538e-05,
+      "loss": 1.0733,
+      "step": 11286
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.386024938206421e-05,
+      "loss": 1.0235,
+      "step": 11287
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3859209984968523e-05,
+      "loss": 0.9954,
+      "step": 11288
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3858170538881512e-05,
+      "loss": 0.9916,
+      "step": 11289
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3857131043816377e-05,
+      "loss": 1.0556,
+      "step": 11290
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3856091499786308e-05,
+      "loss": 0.982,
+      "step": 11291
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3855051906804508e-05,
+      "loss": 0.9036,
+      "step": 11292
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385401226488417e-05,
+      "loss": 0.9456,
+      "step": 11293
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3852972574038495e-05,
+      "loss": 0.9798,
+      "step": 11294
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.385193283428068e-05,
+      "loss": 1.0759,
+      "step": 11295
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3850893045623921e-05,
+      "loss": 1.1465,
+      "step": 11296
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3849853208081425e-05,
+      "loss": 1.0806,
+      "step": 11297
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3848813321666385e-05,
+      "loss": 1.0728,
+      "step": 11298
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3847773386392006e-05,
+      "loss": 1.0893,
+      "step": 11299
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384673340227149e-05,
+      "loss": 0.9607,
+      "step": 11300
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3845693369318038e-05,
+      "loss": 1.0859,
+      "step": 11301
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.384465328754485e-05,
+      "loss": 1.0263,
+      "step": 11302
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3843613156965137e-05,
+      "loss": 1.0194,
+      "step": 11303
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3842572977592097e-05,
+      "loss": 0.9393,
+      "step": 11304
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3841532749438933e-05,
+      "loss": 1.0732,
+      "step": 11305
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3840492472518856e-05,
+      "loss": 1.0347,
+      "step": 11306
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383945214684507e-05,
+      "loss": 0.9398,
+      "step": 11307
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383841177243078e-05,
+      "loss": 0.9045,
+      "step": 11308
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3837371349289192e-05,
+      "loss": 1.0264,
+      "step": 11309
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.383633087743352e-05,
+      "loss": 0.9385,
+      "step": 11310
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3835290356876967e-05,
+      "loss": 0.978,
+      "step": 11311
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3834249787632744e-05,
+      "loss": 1.084,
+      "step": 11312
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3833209169714059e-05,
+      "loss": 1.1523,
+      "step": 11313
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3832168503134122e-05,
+      "loss": 1.0161,
+      "step": 11314
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3831127787906149e-05,
+      "loss": 1.027,
+      "step": 11315
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3830087024043346e-05,
+      "loss": 1.0458,
+      "step": 11316
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3829046211558925e-05,
+      "loss": 0.9712,
+      "step": 11317
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3828005350466105e-05,
+      "loss": 1.0365,
+      "step": 11318
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3826964440778092e-05,
+      "loss": 1.0435,
+      "step": 11319
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3825923482508104e-05,
+      "loss": 1.1272,
+      "step": 11320
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3824882475669354e-05,
+      "loss": 0.9805,
+      "step": 11321
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.382384142027506e-05,
+      "loss": 1.1166,
+      "step": 11322
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3822800316338434e-05,
+      "loss": 1.0541,
+      "step": 11323
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3821759163872694e-05,
+      "loss": 0.9753,
+      "step": 11324
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3820717962891059e-05,
+      "loss": 1.0214,
+      "step": 11325
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3819676713406742e-05,
+      "loss": 1.0179,
+      "step": 11326
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3818635415432966e-05,
+      "loss": 1.1764,
+      "step": 11327
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3817594068982948e-05,
+      "loss": 0.8676,
+      "step": 11328
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381655267406991e-05,
+      "loss": 1.0302,
+      "step": 11329
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3815511230707065e-05,
+      "loss": 1.0747,
+      "step": 11330
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3814469738907641e-05,
+      "loss": 0.9976,
+      "step": 11331
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.381342819868486e-05,
+      "loss": 1.0678,
+      "step": 11332
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3812386610051936e-05,
+      "loss": 1.0451,
+      "step": 11333
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3811344973022098e-05,
+      "loss": 1.1728,
+      "step": 11334
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3810303287608569e-05,
+      "loss": 1.0358,
+      "step": 11335
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3809261553824573e-05,
+      "loss": 1.0775,
+      "step": 11336
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3808219771683329e-05,
+      "loss": 0.9183,
+      "step": 11337
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3807177941198069e-05,
+      "loss": 1.0005,
+      "step": 11338
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3806136062382017e-05,
+      "loss": 1.052,
+      "step": 11339
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3805094135248393e-05,
+      "loss": 1.1083,
+      "step": 11340
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3804052159810432e-05,
+      "loss": 1.1619,
+      "step": 11341
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3803010136081358e-05,
+      "loss": 1.0818,
+      "step": 11342
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3801968064074402e-05,
+      "loss": 1.0612,
+      "step": 11343
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3800925943802787e-05,
+      "loss": 1.0418,
+      "step": 11344
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3799883775279748e-05,
+      "loss": 1.0445,
+      "step": 11345
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3798841558518512e-05,
+      "loss": 0.9959,
+      "step": 11346
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3797799293532308e-05,
+      "loss": 1.069,
+      "step": 11347
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3796756980334374e-05,
+      "loss": 0.9642,
+      "step": 11348
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3795714618937934e-05,
+      "loss": 1.0881,
+      "step": 11349
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3794672209356226e-05,
+      "loss": 1.0472,
+      "step": 11350
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3793629751602477e-05,
+      "loss": 0.933,
+      "step": 11351
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3792587245689926e-05,
+      "loss": 0.9472,
+      "step": 11352
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3791544691631809e-05,
+      "loss": 0.9746,
+      "step": 11353
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 1.3790502089441351e-05,
+      "loss": 1.0818,
+      "step": 11354
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3789459439131799e-05,
+      "loss": 1.1462,
+      "step": 11355
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378841674071638e-05,
+      "loss": 1.0637,
+      "step": 11356
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378737399420834e-05,
+      "loss": 0.9497,
+      "step": 11357
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3786331199620909e-05,
+      "loss": 1.096,
+      "step": 11358
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3785288356967323e-05,
+      "loss": 1.0541,
+      "step": 11359
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3784245466260828e-05,
+      "loss": 0.9405,
+      "step": 11360
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3783202527514656e-05,
+      "loss": 0.9805,
+      "step": 11361
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3782159540742055e-05,
+      "loss": 0.9841,
+      "step": 11362
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3781116505956257e-05,
+      "loss": 1.0962,
+      "step": 11363
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.378007342317051e-05,
+      "loss": 0.9563,
+      "step": 11364
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3779030292398052e-05,
+      "loss": 1.0143,
+      "step": 11365
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3777987113652123e-05,
+      "loss": 0.9567,
+      "step": 11366
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3776943886945971e-05,
+      "loss": 0.9135,
+      "step": 11367
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377590061229283e-05,
+      "loss": 0.9567,
+      "step": 11368
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3774857289705956e-05,
+      "loss": 1.0688,
+      "step": 11369
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3773813919198586e-05,
+      "loss": 0.9223,
+      "step": 11370
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377277050078397e-05,
+      "loss": 1.1321,
+      "step": 11371
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.377172703447535e-05,
+      "loss": 1.0751,
+      "step": 11372
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3770683520285972e-05,
+      "loss": 0.9593,
+      "step": 11373
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3769639958229089e-05,
+      "loss": 1.0271,
+      "step": 11374
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376859634831794e-05,
+      "loss": 0.9548,
+      "step": 11375
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376755269056578e-05,
+      "loss": 1.002,
+      "step": 11376
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3766508984985854e-05,
+      "loss": 1.0971,
+      "step": 11377
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3765465231591413e-05,
+      "loss": 1.0677,
+      "step": 11378
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3764421430395708e-05,
+      "loss": 1.0042,
+      "step": 11379
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3763377581411987e-05,
+      "loss": 1.0842,
+      "step": 11380
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3762333684653506e-05,
+      "loss": 0.9816,
+      "step": 11381
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376128974013351e-05,
+      "loss": 0.9931,
+      "step": 11382
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.376024574786526e-05,
+      "loss": 1.015,
+      "step": 11383
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3759201707862002e-05,
+      "loss": 1.0075,
+      "step": 11384
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3758157620136992e-05,
+      "loss": 0.9873,
+      "step": 11385
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3757113484703486e-05,
+      "loss": 0.9809,
+      "step": 11386
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3756069301574738e-05,
+      "loss": 1.0604,
+      "step": 11387
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3755025070763998e-05,
+      "loss": 1.0446,
+      "step": 11388
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3753980792284532e-05,
+      "loss": 1.18,
+      "step": 11389
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3752936466149592e-05,
+      "loss": 0.9062,
+      "step": 11390
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3751892092372434e-05,
+      "loss": 0.982,
+      "step": 11391
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3750847670966318e-05,
+      "loss": 0.9566,
+      "step": 11392
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3749803201944503e-05,
+      "loss": 0.9855,
+      "step": 11393
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3748758685320244e-05,
+      "loss": 0.9882,
+      "step": 11394
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3747714121106806e-05,
+      "loss": 1.0015,
+      "step": 11395
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3746669509317445e-05,
+      "loss": 1.0939,
+      "step": 11396
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374562484996543e-05,
+      "loss": 0.9576,
+      "step": 11397
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374458014306401e-05,
+      "loss": 1.1909,
+      "step": 11398
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3743535388626457e-05,
+      "loss": 1.1098,
+      "step": 11399
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3742490586666031e-05,
+      "loss": 1.0715,
+      "step": 11400
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3741445737195993e-05,
+      "loss": 1.0294,
+      "step": 11401
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.374040084022961e-05,
+      "loss": 0.9431,
+      "step": 11402
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3739355895780147e-05,
+      "loss": 0.98,
+      "step": 11403
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3738310903860868e-05,
+      "loss": 0.9789,
+      "step": 11404
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3737265864485039e-05,
+      "loss": 1.0875,
+      "step": 11405
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3736220777665925e-05,
+      "loss": 1.0,
+      "step": 11406
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3735175643416794e-05,
+      "loss": 1.0099,
+      "step": 11407
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3734130461750914e-05,
+      "loss": 0.9816,
+      "step": 11408
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3733085232681553e-05,
+      "loss": 1.0969,
+      "step": 11409
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3732039956221979e-05,
+      "loss": 1.2374,
+      "step": 11410
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3730994632385466e-05,
+      "loss": 1.0114,
+      "step": 11411
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3729949261185275e-05,
+      "loss": 1.0646,
+      "step": 11412
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3728903842634684e-05,
+      "loss": 1.0968,
+      "step": 11413
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3727858376746963e-05,
+      "loss": 0.9633,
+      "step": 11414
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3726812863535383e-05,
+      "loss": 1.0219,
+      "step": 11415
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3725767303013216e-05,
+      "loss": 1.0683,
+      "step": 11416
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3724721695193735e-05,
+      "loss": 1.0842,
+      "step": 11417
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3723676040090215e-05,
+      "loss": 1.1136,
+      "step": 11418
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3722630337715927e-05,
+      "loss": 1.0201,
+      "step": 11419
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.372158458808415e-05,
+      "loss": 1.0309,
+      "step": 11420
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3720538791208156e-05,
+      "loss": 0.934,
+      "step": 11421
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3719492947101224e-05,
+      "loss": 1.0214,
+      "step": 11422
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371844705577663e-05,
+      "loss": 0.9605,
+      "step": 11423
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3717401117247649e-05,
+      "loss": 1.0307,
+      "step": 11424
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.371635513152756e-05,
+      "loss": 0.8874,
+      "step": 11425
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3715309098629641e-05,
+      "loss": 0.9716,
+      "step": 11426
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3714263018567171e-05,
+      "loss": 1.0236,
+      "step": 11427
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3713216891353431e-05,
+      "loss": 0.9564,
+      "step": 11428
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3712170717001704e-05,
+      "loss": 0.8844,
+      "step": 11429
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3711124495525262e-05,
+      "loss": 0.9364,
+      "step": 11430
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3710078226937395e-05,
+      "loss": 0.9541,
+      "step": 11431
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3709031911251382e-05,
+      "loss": 1.07,
+      "step": 11432
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3707985548480505e-05,
+      "loss": 0.9863,
+      "step": 11433
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3706939138638044e-05,
+      "loss": 1.139,
+      "step": 11434
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370589268173729e-05,
+      "loss": 1.0016,
+      "step": 11435
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3704846177791525e-05,
+      "loss": 0.9583,
+      "step": 11436
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.370379962681403e-05,
+      "loss": 1.0176,
+      "step": 11437
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3702753028818092e-05,
+      "loss": 1.0689,
+      "step": 11438
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3701706383817001e-05,
+      "loss": 1.0261,
+      "step": 11439
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3700659691824041e-05,
+      "loss": 0.9574,
+      "step": 11440
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3699612952852498e-05,
+      "loss": 1.034,
+      "step": 11441
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3698566166915664e-05,
+      "loss": 0.9225,
+      "step": 11442
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3697519334026825e-05,
+      "loss": 1.1449,
+      "step": 11443
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.369647245419927e-05,
+      "loss": 1.0656,
+      "step": 11444
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3695425527446289e-05,
+      "loss": 1.1004,
+      "step": 11445
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3694378553781173e-05,
+      "loss": 0.9857,
+      "step": 11446
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3693331533217213e-05,
+      "loss": 1.0982,
+      "step": 11447
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3692284465767697e-05,
+      "loss": 1.1225,
+      "step": 11448
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3691237351445925e-05,
+      "loss": 0.9248,
+      "step": 11449
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3690190190265185e-05,
+      "loss": 1.0889,
+      "step": 11450
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3689142982238766e-05,
+      "loss": 1.0715,
+      "step": 11451
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3688095727379969e-05,
+      "loss": 1.0965,
+      "step": 11452
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3687048425702085e-05,
+      "loss": 0.9599,
+      "step": 11453
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3686001077218412e-05,
+      "loss": 0.9753,
+      "step": 11454
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368495368194224e-05,
+      "loss": 1.019,
+      "step": 11455
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.368390623988687e-05,
+      "loss": 1.0275,
+      "step": 11456
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3682858751065603e-05,
+      "loss": 0.9774,
+      "step": 11457
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3681811215491724e-05,
+      "loss": 1.224,
+      "step": 11458
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3680763633178544e-05,
+      "loss": 1.1251,
+      "step": 11459
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3679716004139354e-05,
+      "loss": 1.1772,
+      "step": 11460
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3678668328387454e-05,
+      "loss": 1.0743,
+      "step": 11461
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3677620605936148e-05,
+      "loss": 1.1212,
+      "step": 11462
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3676572836798733e-05,
+      "loss": 1.0457,
+      "step": 11463
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3675525020988514e-05,
+      "loss": 0.8589,
+      "step": 11464
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3674477158518787e-05,
+      "loss": 1.1301,
+      "step": 11465
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3673429249402854e-05,
+      "loss": 1.0132,
+      "step": 11466
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3672381293654025e-05,
+      "loss": 0.9866,
+      "step": 11467
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3671333291285595e-05,
+      "loss": 1.0343,
+      "step": 11468
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3670285242310875e-05,
+      "loss": 1.0722,
+      "step": 11469
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3669237146743167e-05,
+      "loss": 1.0232,
+      "step": 11470
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3668189004595776e-05,
+      "loss": 1.1127,
+      "step": 11471
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3667140815882007e-05,
+      "loss": 1.1064,
+      "step": 11472
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3666092580615167e-05,
+      "loss": 1.025,
+      "step": 11473
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3665044298808562e-05,
+      "loss": 0.9299,
+      "step": 11474
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3663995970475502e-05,
+      "loss": 1.0209,
+      "step": 11475
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3662947595629294e-05,
+      "loss": 1.0423,
+      "step": 11476
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3661899174283245e-05,
+      "loss": 0.9826,
+      "step": 11477
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3660850706450669e-05,
+      "loss": 1.0233,
+      "step": 11478
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365980219214487e-05,
+      "loss": 1.1804,
+      "step": 11479
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3658753631379162e-05,
+      "loss": 1.1121,
+      "step": 11480
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3657705024166858e-05,
+      "loss": 1.041,
+      "step": 11481
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3656656370521261e-05,
+      "loss": 0.9856,
+      "step": 11482
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3655607670455698e-05,
+      "loss": 0.9698,
+      "step": 11483
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3654558923983468e-05,
+      "loss": 1.0594,
+      "step": 11484
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3653510131117891e-05,
+      "loss": 1.0613,
+      "step": 11485
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3652461291872282e-05,
+      "loss": 1.0741,
+      "step": 11486
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.365141240625995e-05,
+      "loss": 0.997,
+      "step": 11487
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3650363474294218e-05,
+      "loss": 1.0101,
+      "step": 11488
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.364931449598839e-05,
+      "loss": 1.0629,
+      "step": 11489
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3648265471355797e-05,
+      "loss": 0.9654,
+      "step": 11490
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3647216400409746e-05,
+      "loss": 0.9981,
+      "step": 11491
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3646167283163558e-05,
+      "loss": 0.9917,
+      "step": 11492
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3645118119630554e-05,
+      "loss": 1.0338,
+      "step": 11493
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3644068909824046e-05,
+      "loss": 0.9936,
+      "step": 11494
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3643019653757357e-05,
+      "loss": 1.0894,
+      "step": 11495
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3641970351443808e-05,
+      "loss": 1.0323,
+      "step": 11496
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3640921002896721e-05,
+      "loss": 1.0149,
+      "step": 11497
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363987160812941e-05,
+      "loss": 1.1152,
+      "step": 11498
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3638822167155202e-05,
+      "loss": 1.1426,
+      "step": 11499
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3637772679987422e-05,
+      "loss": 1.1059,
+      "step": 11500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3636723146639386e-05,
+      "loss": 0.8491,
+      "step": 11501
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3635673567124422e-05,
+      "loss": 0.9599,
+      "step": 11502
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3634623941455851e-05,
+      "loss": 1.0215,
+      "step": 11503
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3633574269647003e-05,
+      "loss": 0.9228,
+      "step": 11504
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3632524551711198e-05,
+      "loss": 1.0715,
+      "step": 11505
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.363147478766176e-05,
+      "loss": 0.9685,
+      "step": 11506
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3630424977512025e-05,
+      "loss": 0.9944,
+      "step": 11507
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362937512127531e-05,
+      "loss": 1.0179,
+      "step": 11508
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3628325218964947e-05,
+      "loss": 1.0919,
+      "step": 11509
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3627275270594264e-05,
+      "loss": 0.8907,
+      "step": 11510
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362622527617659e-05,
+      "loss": 1.0733,
+      "step": 11511
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3625175235725252e-05,
+      "loss": 1.1194,
+      "step": 11512
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3624125149253582e-05,
+      "loss": 0.934,
+      "step": 11513
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3623075016774912e-05,
+      "loss": 1.1281,
+      "step": 11514
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.362202483830257e-05,
+      "loss": 1.0894,
+      "step": 11515
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3620974613849887e-05,
+      "loss": 1.0166,
+      "step": 11516
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3619924343430197e-05,
+      "loss": 1.0151,
+      "step": 11517
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3618874027056836e-05,
+      "loss": 0.9885,
+      "step": 11518
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361782366474313e-05,
+      "loss": 0.9135,
+      "step": 11519
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.361677325650242e-05,
+      "loss": 1.0721,
+      "step": 11520
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3615722802348037e-05,
+      "loss": 1.0778,
+      "step": 11521
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3614672302293316e-05,
+      "loss": 0.9568,
+      "step": 11522
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3613621756351596e-05,
+      "loss": 0.9993,
+      "step": 11523
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3612571164536207e-05,
+      "loss": 0.9033,
+      "step": 11524
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3611520526860495e-05,
+      "loss": 1.1433,
+      "step": 11525
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3610469843337788e-05,
+      "loss": 1.0312,
+      "step": 11526
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3609419113981428e-05,
+      "loss": 0.9297,
+      "step": 11527
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3608368338804759e-05,
+      "loss": 1.0724,
+      "step": 11528
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360731751782111e-05,
+      "loss": 0.9522,
+      "step": 11529
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3606266651043827e-05,
+      "loss": 1.0864,
+      "step": 11530
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.360521573848625e-05,
+      "loss": 0.9842,
+      "step": 11531
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3604164780161721e-05,
+      "loss": 0.9835,
+      "step": 11532
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3603113776083578e-05,
+      "loss": 0.9752,
+      "step": 11533
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3602062726265164e-05,
+      "loss": 1.155,
+      "step": 11534
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3601011630719827e-05,
+      "loss": 0.8474,
+      "step": 11535
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3599960489460901e-05,
+      "loss": 0.9977,
+      "step": 11536
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.359890930250174e-05,
+      "loss": 1.1061,
+      "step": 11537
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3597858069855679e-05,
+      "loss": 0.96,
+      "step": 11538
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3596806791536068e-05,
+      "loss": 0.8924,
+      "step": 11539
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3595755467556252e-05,
+      "loss": 0.9972,
+      "step": 11540
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3594704097929578e-05,
+      "loss": 0.9662,
+      "step": 11541
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3593652682669394e-05,
+      "loss": 1.0965,
+      "step": 11542
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3592601221789044e-05,
+      "loss": 1.1459,
+      "step": 11543
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3591549715301874e-05,
+      "loss": 1.0115,
+      "step": 11544
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3590498163221239e-05,
+      "loss": 0.9483,
+      "step": 11545
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3589446565560485e-05,
+      "loss": 0.9526,
+      "step": 11546
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3588394922332962e-05,
+      "loss": 1.0142,
+      "step": 11547
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3587343233552017e-05,
+      "loss": 0.9357,
+      "step": 11548
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3586291499231005e-05,
+      "loss": 1.0754,
+      "step": 11549
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3585239719383276e-05,
+      "loss": 1.0489,
+      "step": 11550
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3584187894022182e-05,
+      "loss": 1.0865,
+      "step": 11551
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3583136023161075e-05,
+      "loss": 0.9976,
+      "step": 11552
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3582084106813309e-05,
+      "loss": 1.124,
+      "step": 11553
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3581032144992237e-05,
+      "loss": 1.1233,
+      "step": 11554
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3579980137711214e-05,
+      "loss": 1.0546,
+      "step": 11555
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3578928084983593e-05,
+      "loss": 0.9511,
+      "step": 11556
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3577875986822732e-05,
+      "loss": 0.9996,
+      "step": 11557
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3576823843241988e-05,
+      "loss": 1.0078,
+      "step": 11558
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3575771654254712e-05,
+      "loss": 1.06,
+      "step": 11559
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3574719419874269e-05,
+      "loss": 0.9945,
+      "step": 11560
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3573667140114008e-05,
+      "loss": 0.9766,
+      "step": 11561
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3572614814987295e-05,
+      "loss": 1.0113,
+      "step": 11562
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3571562444507485e-05,
+      "loss": 1.131,
+      "step": 11563
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3570510028687938e-05,
+      "loss": 1.0183,
+      "step": 11564
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3569457567542012e-05,
+      "loss": 1.0218,
+      "step": 11565
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3568405061083072e-05,
+      "loss": 1.0011,
+      "step": 11566
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356735250932448e-05,
+      "loss": 1.0506,
+      "step": 11567
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3566299912279588e-05,
+      "loss": 1.2289,
+      "step": 11568
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3565247269961769e-05,
+      "loss": 1.0291,
+      "step": 11569
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356419458238438e-05,
+      "loss": 1.1061,
+      "step": 11570
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356314184956079e-05,
+      "loss": 1.0877,
+      "step": 11571
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3562089071504358e-05,
+      "loss": 0.8974,
+      "step": 11572
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.356103624822845e-05,
+      "loss": 1.0122,
+      "step": 11573
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3559983379746433e-05,
+      "loss": 0.9853,
+      "step": 11574
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3558930466071667e-05,
+      "loss": 1.0678,
+      "step": 11575
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3557877507217527e-05,
+      "loss": 1.0331,
+      "step": 11576
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3556824503197374e-05,
+      "loss": 0.8915,
+      "step": 11577
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.355577145402458e-05,
+      "loss": 0.9907,
+      "step": 11578
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3554718359712506e-05,
+      "loss": 1.0434,
+      "step": 11579
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3553665220274528e-05,
+      "loss": 1.137,
+      "step": 11580
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3552612035724013e-05,
+      "loss": 1.0125,
+      "step": 11581
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3551558806074324e-05,
+      "loss": 0.9191,
+      "step": 11582
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3550505531338845e-05,
+      "loss": 1.0101,
+      "step": 11583
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3549452211530934e-05,
+      "loss": 0.8475,
+      "step": 11584
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3548398846663968e-05,
+      "loss": 1.068,
+      "step": 11585
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3547345436751322e-05,
+      "loss": 1.0333,
+      "step": 11586
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3546291981806365e-05,
+      "loss": 1.0702,
+      "step": 11587
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354523848184247e-05,
+      "loss": 1.0391,
+      "step": 11588
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.354418493687301e-05,
+      "loss": 0.9589,
+      "step": 11589
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3543131346911365e-05,
+      "loss": 1.0572,
+      "step": 11590
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3542077711970902e-05,
+      "loss": 1.0471,
+      "step": 11591
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3541024032065002e-05,
+      "loss": 1.014,
+      "step": 11592
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353997030720704e-05,
+      "loss": 1.0638,
+      "step": 11593
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3538916537410391e-05,
+      "loss": 0.9393,
+      "step": 11594
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3537862722688437e-05,
+      "loss": 1.0852,
+      "step": 11595
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3536808863054547e-05,
+      "loss": 1.1052,
+      "step": 11596
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3535754958522111e-05,
+      "loss": 0.9287,
+      "step": 11597
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3534701009104497e-05,
+      "loss": 1.0908,
+      "step": 11598
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353364701481509e-05,
+      "loss": 1.1391,
+      "step": 11599
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3532592975667272e-05,
+      "loss": 1.1037,
+      "step": 11600
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.353153889167442e-05,
+      "loss": 1.0209,
+      "step": 11601
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3530484762849914e-05,
+      "loss": 1.012,
+      "step": 11602
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3529430589207135e-05,
+      "loss": 0.9923,
+      "step": 11603
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3528376370759476e-05,
+      "loss": 0.9973,
+      "step": 11604
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3527322107520306e-05,
+      "loss": 0.967,
+      "step": 11605
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3526267799503017e-05,
+      "loss": 1.0623,
+      "step": 11606
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3525213446720993e-05,
+      "loss": 0.9932,
+      "step": 11607
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3524159049187614e-05,
+      "loss": 1.0096,
+      "step": 11608
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3523104606916268e-05,
+      "loss": 1.0412,
+      "step": 11609
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.352205011992034e-05,
+      "loss": 0.9264,
+      "step": 11610
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3520995588213219e-05,
+      "loss": 0.9035,
+      "step": 11611
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351994101180829e-05,
+      "loss": 1.0594,
+      "step": 11612
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3518886390718936e-05,
+      "loss": 0.9159,
+      "step": 11613
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3517831724958555e-05,
+      "loss": 0.9157,
+      "step": 11614
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3516777014540525e-05,
+      "loss": 1.0745,
+      "step": 11615
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.351572225947824e-05,
+      "loss": 1.1184,
+      "step": 11616
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3514667459785092e-05,
+      "loss": 1.0265,
+      "step": 11617
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3513612615474472e-05,
+      "loss": 1.035,
+      "step": 11618
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3512557726559764e-05,
+      "loss": 0.8727,
+      "step": 11619
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3511502793054363e-05,
+      "loss": 1.042,
+      "step": 11620
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3510447814971663e-05,
+      "loss": 0.9106,
+      "step": 11621
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3509392792325054e-05,
+      "loss": 0.9596,
+      "step": 11622
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3508337725127932e-05,
+      "loss": 1.0385,
+      "step": 11623
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3507282613393687e-05,
+      "loss": 1.112,
+      "step": 11624
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3506227457135715e-05,
+      "loss": 1.0942,
+      "step": 11625
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350517225636741e-05,
+      "loss": 1.0272,
+      "step": 11626
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.350411701110217e-05,
+      "loss": 1.0859,
+      "step": 11627
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3503061721353391e-05,
+      "loss": 1.1463,
+      "step": 11628
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3502006387134467e-05,
+      "loss": 1.013,
+      "step": 11629
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3500951008458794e-05,
+      "loss": 0.9917,
+      "step": 11630
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3499895585339773e-05,
+      "loss": 1.112,
+      "step": 11631
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3498840117790802e-05,
+      "loss": 1.0525,
+      "step": 11632
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3497784605825277e-05,
+      "loss": 1.0233,
+      "step": 11633
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3496729049456599e-05,
+      "loss": 0.9725,
+      "step": 11634
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349567344869817e-05,
+      "loss": 1.0808,
+      "step": 11635
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3494617803563385e-05,
+      "loss": 1.0195,
+      "step": 11636
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.349356211406565e-05,
+      "loss": 0.9652,
+      "step": 11637
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3492506380218367e-05,
+      "loss": 1.1293,
+      "step": 11638
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3491450602034935e-05,
+      "loss": 1.0099,
+      "step": 11639
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3490394779528757e-05,
+      "loss": 1.1598,
+      "step": 11640
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3489338912713238e-05,
+      "loss": 0.9279,
+      "step": 11641
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 1.3488283001601783e-05,
+      "loss": 1.0629,
+      "step": 11642
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3487227046207791e-05,
+      "loss": 1.1303,
+      "step": 11643
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3486171046544671e-05,
+      "loss": 1.044,
+      "step": 11644
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3485115002625831e-05,
+      "loss": 0.9942,
+      "step": 11645
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3484058914464673e-05,
+      "loss": 1.0168,
+      "step": 11646
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3483002782074604e-05,
+      "loss": 1.0222,
+      "step": 11647
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3481946605469032e-05,
+      "loss": 1.0505,
+      "step": 11648
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3480890384661367e-05,
+      "loss": 0.9856,
+      "step": 11649
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3479834119665014e-05,
+      "loss": 0.8912,
+      "step": 11650
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3478777810493382e-05,
+      "loss": 0.9485,
+      "step": 11651
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3477721457159883e-05,
+      "loss": 0.954,
+      "step": 11652
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3476665059677927e-05,
+      "loss": 0.9666,
+      "step": 11653
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3475608618060921e-05,
+      "loss": 1.0662,
+      "step": 11654
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3474552132322278e-05,
+      "loss": 1.0216,
+      "step": 11655
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3473495602475413e-05,
+      "loss": 1.0291,
+      "step": 11656
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3472439028533733e-05,
+      "loss": 1.0592,
+      "step": 11657
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3471382410510654e-05,
+      "loss": 1.1136,
+      "step": 11658
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3470325748419588e-05,
+      "loss": 1.0181,
+      "step": 11659
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3469269042273951e-05,
+      "loss": 1.046,
+      "step": 11660
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3468212292087156e-05,
+      "loss": 0.9722,
+      "step": 11661
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3467155497872618e-05,
+      "loss": 0.9563,
+      "step": 11662
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3466098659643751e-05,
+      "loss": 0.9232,
+      "step": 11663
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3465041777413976e-05,
+      "loss": 0.9696,
+      "step": 11664
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3463984851196703e-05,
+      "loss": 0.9829,
+      "step": 11665
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3462927881005357e-05,
+      "loss": 1.0396,
+      "step": 11666
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.346187086685335e-05,
+      "loss": 0.9934,
+      "step": 11667
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3460813808754102e-05,
+      "loss": 0.9041,
+      "step": 11668
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3459756706721034e-05,
+      "loss": 1.1931,
+      "step": 11669
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3458699560767566e-05,
+      "loss": 0.9992,
+      "step": 11670
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3457642370907111e-05,
+      "loss": 0.9247,
+      "step": 11671
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3456585137153096e-05,
+      "loss": 1.0882,
+      "step": 11672
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345552785951894e-05,
+      "loss": 1.0078,
+      "step": 11673
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345447053801807e-05,
+      "loss": 1.087,
+      "step": 11674
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.34534131726639e-05,
+      "loss": 0.9401,
+      "step": 11675
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3452355763469858e-05,
+      "loss": 0.9613,
+      "step": 11676
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3451298310449368e-05,
+      "loss": 0.9666,
+      "step": 11677
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.345024081361585e-05,
+      "loss": 0.9162,
+      "step": 11678
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344918327298273e-05,
+      "loss": 1.0119,
+      "step": 11679
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3448125688563436e-05,
+      "loss": 1.0367,
+      "step": 11680
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3447068060371394e-05,
+      "loss": 0.9293,
+      "step": 11681
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3446010388420021e-05,
+      "loss": 1.0205,
+      "step": 11682
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3444952672722758e-05,
+      "loss": 0.9178,
+      "step": 11683
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3443894913293023e-05,
+      "loss": 1.0792,
+      "step": 11684
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3442837110144242e-05,
+      "loss": 1.0274,
+      "step": 11685
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.344177926328985e-05,
+      "loss": 1.0092,
+      "step": 11686
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3440721372743273e-05,
+      "loss": 0.9904,
+      "step": 11687
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3439663438517942e-05,
+      "loss": 1.0577,
+      "step": 11688
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3438605460627283e-05,
+      "loss": 1.0456,
+      "step": 11689
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3437547439084732e-05,
+      "loss": 1.0108,
+      "step": 11690
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3436489373903717e-05,
+      "loss": 0.9577,
+      "step": 11691
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3435431265097672e-05,
+      "loss": 1.0677,
+      "step": 11692
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3434373112680028e-05,
+      "loss": 1.0974,
+      "step": 11693
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3433314916664215e-05,
+      "loss": 0.9665,
+      "step": 11694
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.343225667706367e-05,
+      "loss": 0.9366,
+      "step": 11695
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3431198393891827e-05,
+      "loss": 0.9517,
+      "step": 11696
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3430140067162121e-05,
+      "loss": 0.8494,
+      "step": 11697
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3429081696887981e-05,
+      "loss": 1.18,
+      "step": 11698
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3428023283082853e-05,
+      "loss": 1.1666,
+      "step": 11699
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3426964825760167e-05,
+      "loss": 1.0192,
+      "step": 11700
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3425906324933358e-05,
+      "loss": 0.9935,
+      "step": 11701
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3424847780615868e-05,
+      "loss": 1.0642,
+      "step": 11702
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3423789192821128e-05,
+      "loss": 0.988,
+      "step": 11703
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3422730561562585e-05,
+      "loss": 0.9987,
+      "step": 11704
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3421671886853673e-05,
+      "loss": 1.0362,
+      "step": 11705
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3420613168707832e-05,
+      "loss": 0.9485,
+      "step": 11706
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3419554407138504e-05,
+      "loss": 0.9228,
+      "step": 11707
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3418495602159127e-05,
+      "loss": 1.1115,
+      "step": 11708
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3417436753783144e-05,
+      "loss": 1.0316,
+      "step": 11709
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3416377862023993e-05,
+      "loss": 0.9742,
+      "step": 11710
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3415318926895123e-05,
+      "loss": 0.9695,
+      "step": 11711
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341425994840997e-05,
+      "loss": 1.0818,
+      "step": 11712
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3413200926581981e-05,
+      "loss": 1.111,
+      "step": 11713
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3412141861424602e-05,
+      "loss": 1.111,
+      "step": 11714
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.341108275295127e-05,
+      "loss": 1.0699,
+      "step": 11715
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3410023601175439e-05,
+      "loss": 1.124,
+      "step": 11716
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3408964406110548e-05,
+      "loss": 0.968,
+      "step": 11717
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3407905167770045e-05,
+      "loss": 0.9792,
+      "step": 11718
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3406845886167379e-05,
+      "loss": 0.8721,
+      "step": 11719
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3405786561315994e-05,
+      "loss": 1.0097,
+      "step": 11720
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3404727193229341e-05,
+      "loss": 1.0355,
+      "step": 11721
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3403667781920864e-05,
+      "loss": 0.9738,
+      "step": 11722
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3402608327404014e-05,
+      "loss": 0.9919,
+      "step": 11723
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3401548829692241e-05,
+      "loss": 0.9541,
+      "step": 11724
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3400489288798997e-05,
+      "loss": 0.8968,
+      "step": 11725
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3399429704737728e-05,
+      "loss": 1.0451,
+      "step": 11726
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3398370077521887e-05,
+      "loss": 1.0011,
+      "step": 11727
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3397310407164928e-05,
+      "loss": 1.0343,
+      "step": 11728
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33962506936803e-05,
+      "loss": 0.9787,
+      "step": 11729
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3395190937081453e-05,
+      "loss": 0.9755,
+      "step": 11730
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3394131137381847e-05,
+      "loss": 1.0273,
+      "step": 11731
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3393071294594933e-05,
+      "loss": 1.041,
+      "step": 11732
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3392011408734163e-05,
+      "loss": 1.1151,
+      "step": 11733
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3390951479812992e-05,
+      "loss": 1.0409,
+      "step": 11734
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3389891507844882e-05,
+      "loss": 1.1034,
+      "step": 11735
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3388831492843282e-05,
+      "loss": 1.0166,
+      "step": 11736
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3387771434821651e-05,
+      "loss": 0.9861,
+      "step": 11737
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3386711333793445e-05,
+      "loss": 0.9448,
+      "step": 11738
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3385651189772123e-05,
+      "loss": 1.0613,
+      "step": 11739
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3384591002771144e-05,
+      "loss": 1.0359,
+      "step": 11740
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3383530772803964e-05,
+      "loss": 1.0485,
+      "step": 11741
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3382470499884043e-05,
+      "loss": 0.9226,
+      "step": 11742
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3381410184024841e-05,
+      "loss": 0.888,
+      "step": 11743
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3380349825239818e-05,
+      "loss": 1.0096,
+      "step": 11744
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337928942354244e-05,
+      "loss": 1.0373,
+      "step": 11745
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3378228978946159e-05,
+      "loss": 0.9909,
+      "step": 11746
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3377168491464442e-05,
+      "loss": 0.9819,
+      "step": 11747
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3376107961110753e-05,
+      "loss": 1.0155,
+      "step": 11748
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3375047387898557e-05,
+      "loss": 0.9154,
+      "step": 11749
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337398677184131e-05,
+      "loss": 1.039,
+      "step": 11750
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.337292611295248e-05,
+      "loss": 0.9774,
+      "step": 11751
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3371865411245532e-05,
+      "loss": 1.1076,
+      "step": 11752
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3370804666733934e-05,
+      "loss": 1.1465,
+      "step": 11753
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3369743879431147e-05,
+      "loss": 1.0277,
+      "step": 11754
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336868304935064e-05,
+      "loss": 0.9953,
+      "step": 11755
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3367622176505878e-05,
+      "loss": 1.0753,
+      "step": 11756
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3366561260910329e-05,
+      "loss": 1.0953,
+      "step": 11757
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3365500302577463e-05,
+      "loss": 0.9499,
+      "step": 11758
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3364439301520746e-05,
+      "loss": 1.021,
+      "step": 11759
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.336337825775365e-05,
+      "loss": 0.9986,
+      "step": 11760
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3362317171289641e-05,
+      "loss": 0.8986,
+      "step": 11761
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3361256042142193e-05,
+      "loss": 0.9327,
+      "step": 11762
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3360194870324773e-05,
+      "loss": 1.1069,
+      "step": 11763
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3359133655850851e-05,
+      "loss": 1.0743,
+      "step": 11764
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3358072398733904e-05,
+      "loss": 0.9838,
+      "step": 11765
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.33570110989874e-05,
+      "loss": 0.9542,
+      "step": 11766
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3355949756624816e-05,
+      "loss": 1.0338,
+      "step": 11767
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3354888371659623e-05,
+      "loss": 1.0631,
+      "step": 11768
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3353826944105293e-05,
+      "loss": 1.1007,
+      "step": 11769
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3352765473975302e-05,
+      "loss": 1.1346,
+      "step": 11770
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3351703961283127e-05,
+      "loss": 1.058,
+      "step": 11771
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3350642406042239e-05,
+      "loss": 0.8494,
+      "step": 11772
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3349580808266118e-05,
+      "loss": 1.0798,
+      "step": 11773
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334851916796824e-05,
+      "loss": 1.0286,
+      "step": 11774
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3347457485162081e-05,
+      "loss": 0.9818,
+      "step": 11775
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.334639575986112e-05,
+      "loss": 0.8872,
+      "step": 11776
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3345333992078838e-05,
+      "loss": 1.1483,
+      "step": 11777
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3344272181828704e-05,
+      "loss": 0.9699,
+      "step": 11778
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3343210329124205e-05,
+      "loss": 0.9006,
+      "step": 11779
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3342148433978821e-05,
+      "loss": 1.0937,
+      "step": 11780
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3341086496406032e-05,
+      "loss": 1.0362,
+      "step": 11781
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3340024516419315e-05,
+      "loss": 1.0635,
+      "step": 11782
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3338962494032154e-05,
+      "loss": 1.1303,
+      "step": 11783
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3337900429258035e-05,
+      "loss": 0.9988,
+      "step": 11784
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3336838322110434e-05,
+      "loss": 1.0662,
+      "step": 11785
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3335776172602837e-05,
+      "loss": 0.9891,
+      "step": 11786
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3334713980748726e-05,
+      "loss": 1.0242,
+      "step": 11787
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333365174656159e-05,
+      "loss": 1.0444,
+      "step": 11788
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3332589470054907e-05,
+      "loss": 0.9434,
+      "step": 11789
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.333152715124217e-05,
+      "loss": 1.0463,
+      "step": 11790
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3330464790136857e-05,
+      "loss": 0.9987,
+      "step": 11791
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3329402386752457e-05,
+      "loss": 1.173,
+      "step": 11792
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3328339941102459e-05,
+      "loss": 0.9956,
+      "step": 11793
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.332727745320035e-05,
+      "loss": 0.9518,
+      "step": 11794
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3326214923059615e-05,
+      "loss": 1.0726,
+      "step": 11795
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3325152350693742e-05,
+      "loss": 0.9578,
+      "step": 11796
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3324089736116226e-05,
+      "loss": 1.0222,
+      "step": 11797
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3323027079340551e-05,
+      "loss": 0.9349,
+      "step": 11798
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3321964380380208e-05,
+      "loss": 1.0913,
+      "step": 11799
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3320901639248688e-05,
+      "loss": 1.1044,
+      "step": 11800
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3319838855959483e-05,
+      "loss": 1.0263,
+      "step": 11801
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3318776030526088e-05,
+      "loss": 1.0485,
+      "step": 11802
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3317713162961985e-05,
+      "loss": 0.9608,
+      "step": 11803
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3316650253280679e-05,
+      "loss": 1.008,
+      "step": 11804
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3315587301495656e-05,
+      "loss": 1.0406,
+      "step": 11805
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.331452430762041e-05,
+      "loss": 0.9612,
+      "step": 11806
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3313461271668434e-05,
+      "loss": 1.0835,
+      "step": 11807
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3312398193653227e-05,
+      "loss": 0.9534,
+      "step": 11808
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3311335073588286e-05,
+      "loss": 0.9595,
+      "step": 11809
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3310271911487095e-05,
+      "loss": 0.8492,
+      "step": 11810
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3309208707363169e-05,
+      "loss": 0.9743,
+      "step": 11811
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3308145461229992e-05,
+      "loss": 0.9865,
+      "step": 11812
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3307082173101062e-05,
+      "loss": 1.0078,
+      "step": 11813
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3306018842989883e-05,
+      "loss": 1.1033,
+      "step": 11814
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3304955470909948e-05,
+      "loss": 0.9619,
+      "step": 11815
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330389205687476e-05,
+      "loss": 0.9063,
+      "step": 11816
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3302828600897815e-05,
+      "loss": 0.9741,
+      "step": 11817
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330176510299262e-05,
+      "loss": 1.0134,
+      "step": 11818
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.330070156317267e-05,
+      "loss": 1.1299,
+      "step": 11819
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329963798145146e-05,
+      "loss": 1.1037,
+      "step": 11820
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3298574357842506e-05,
+      "loss": 1.1364,
+      "step": 11821
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3297510692359303e-05,
+      "loss": 1.0222,
+      "step": 11822
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3296446985015354e-05,
+      "loss": 1.0095,
+      "step": 11823
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3295383235824159e-05,
+      "loss": 0.95,
+      "step": 11824
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3294319444799233e-05,
+      "loss": 0.9999,
+      "step": 11825
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.329325561195407e-05,
+      "loss": 1.1258,
+      "step": 11826
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3292191737302173e-05,
+      "loss": 1.0372,
+      "step": 11827
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3291127820857059e-05,
+      "loss": 1.0901,
+      "step": 11828
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3290063862632225e-05,
+      "loss": 1.0767,
+      "step": 11829
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3288999862641181e-05,
+      "loss": 0.9814,
+      "step": 11830
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3287935820897433e-05,
+      "loss": 0.9953,
+      "step": 11831
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328687173741449e-05,
+      "loss": 1.0412,
+      "step": 11832
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.328580761220586e-05,
+      "loss": 1.0438,
+      "step": 11833
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3284743445285047e-05,
+      "loss": 0.9835,
+      "step": 11834
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3283679236665568e-05,
+      "loss": 1.0221,
+      "step": 11835
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3282614986360926e-05,
+      "loss": 1.1115,
+      "step": 11836
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3281550694384635e-05,
+      "loss": 1.0613,
+      "step": 11837
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3280486360750205e-05,
+      "loss": 0.9665,
+      "step": 11838
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3279421985471148e-05,
+      "loss": 0.9591,
+      "step": 11839
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3278357568560973e-05,
+      "loss": 0.9877,
+      "step": 11840
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3277293110033198e-05,
+      "loss": 1.1055,
+      "step": 11841
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3276228609901332e-05,
+      "loss": 0.9544,
+      "step": 11842
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3275164068178887e-05,
+      "loss": 1.0101,
+      "step": 11843
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3274099484879379e-05,
+      "loss": 0.9608,
+      "step": 11844
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3273034860016322e-05,
+      "loss": 1.247,
+      "step": 11845
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3271970193603235e-05,
+      "loss": 1.0356,
+      "step": 11846
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3270905485653627e-05,
+      "loss": 0.9089,
+      "step": 11847
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3269840736181016e-05,
+      "loss": 1.0197,
+      "step": 11848
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3268775945198923e-05,
+      "loss": 1.0436,
+      "step": 11849
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3267711112720863e-05,
+      "loss": 1.0745,
+      "step": 11850
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3266646238760348e-05,
+      "loss": 0.9895,
+      "step": 11851
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3265581323330903e-05,
+      "loss": 1.0669,
+      "step": 11852
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3264516366446045e-05,
+      "loss": 0.9692,
+      "step": 11853
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3263451368119293e-05,
+      "loss": 0.9374,
+      "step": 11854
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3262386328364166e-05,
+      "loss": 0.9794,
+      "step": 11855
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3261321247194186e-05,
+      "loss": 1.1399,
+      "step": 11856
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.326025612462287e-05,
+      "loss": 1.0692,
+      "step": 11857
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3259190960663743e-05,
+      "loss": 1.0973,
+      "step": 11858
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3258125755330327e-05,
+      "loss": 0.9876,
+      "step": 11859
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3257060508636144e-05,
+      "loss": 1.0309,
+      "step": 11860
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3255995220594713e-05,
+      "loss": 1.1481,
+      "step": 11861
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3254929891219563e-05,
+      "loss": 1.076,
+      "step": 11862
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3253864520524218e-05,
+      "loss": 0.999,
+      "step": 11863
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3252799108522197e-05,
+      "loss": 0.9932,
+      "step": 11864
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3251733655227028e-05,
+      "loss": 0.9758,
+      "step": 11865
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3250668160652239e-05,
+      "loss": 1.1122,
+      "step": 11866
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3249602624811355e-05,
+      "loss": 1.0532,
+      "step": 11867
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3248537047717897e-05,
+      "loss": 0.8995,
+      "step": 11868
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3247471429385401e-05,
+      "loss": 0.9661,
+      "step": 11869
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3246405769827388e-05,
+      "loss": 0.988,
+      "step": 11870
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3245340069057391e-05,
+      "loss": 0.9756,
+      "step": 11871
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3244274327088934e-05,
+      "loss": 1.0595,
+      "step": 11872
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3243208543935547e-05,
+      "loss": 1.067,
+      "step": 11873
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3242142719610764e-05,
+      "loss": 1.0459,
+      "step": 11874
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.324107685412811e-05,
+      "loss": 1.0759,
+      "step": 11875
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3240010947501119e-05,
+      "loss": 0.9993,
+      "step": 11876
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3238944999743325e-05,
+      "loss": 0.8642,
+      "step": 11877
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3237879010868251e-05,
+      "loss": 1.0226,
+      "step": 11878
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3236812980889438e-05,
+      "loss": 1.095,
+      "step": 11879
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3235746909820413e-05,
+      "loss": 1.07,
+      "step": 11880
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3234680797674715e-05,
+      "loss": 1.0466,
+      "step": 11881
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3233614644465873e-05,
+      "loss": 1.0136,
+      "step": 11882
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3232548450207422e-05,
+      "loss": 1.055,
+      "step": 11883
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3231482214912904e-05,
+      "loss": 1.1141,
+      "step": 11884
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3230415938595844e-05,
+      "loss": 1.0666,
+      "step": 11885
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3229349621269783e-05,
+      "loss": 1.0343,
+      "step": 11886
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3228283262948258e-05,
+      "loss": 0.9874,
+      "step": 11887
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3227216863644808e-05,
+      "loss": 1.0763,
+      "step": 11888
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3226150423372963e-05,
+      "loss": 0.9407,
+      "step": 11889
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3225083942146268e-05,
+      "loss": 0.9137,
+      "step": 11890
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3224017419978262e-05,
+      "loss": 0.9358,
+      "step": 11891
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3222950856882479e-05,
+      "loss": 1.0056,
+      "step": 11892
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3221884252872461e-05,
+      "loss": 0.9354,
+      "step": 11893
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.322081760796175e-05,
+      "loss": 1.0151,
+      "step": 11894
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3219750922163887e-05,
+      "loss": 1.0478,
+      "step": 11895
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3218684195492411e-05,
+      "loss": 1.0636,
+      "step": 11896
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.321761742796086e-05,
+      "loss": 1.0396,
+      "step": 11897
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3216550619582786e-05,
+      "loss": 1.069,
+      "step": 11898
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3215483770371721e-05,
+      "loss": 1.0488,
+      "step": 11899
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3214416880341216e-05,
+      "loss": 0.9701,
+      "step": 11900
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3213349949504813e-05,
+      "loss": 1.0071,
+      "step": 11901
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3212282977876056e-05,
+      "loss": 0.9778,
+      "step": 11902
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3211215965468485e-05,
+      "loss": 0.9074,
+      "step": 11903
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3210148912295656e-05,
+      "loss": 1.0561,
+      "step": 11904
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3209081818371106e-05,
+      "loss": 0.9925,
+      "step": 11905
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3208014683708383e-05,
+      "loss": 1.0743,
+      "step": 11906
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3206947508321035e-05,
+      "loss": 0.8743,
+      "step": 11907
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3205880292222608e-05,
+      "loss": 1.0589,
+      "step": 11908
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3204813035426655e-05,
+      "loss": 1.0091,
+      "step": 11909
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3203745737946717e-05,
+      "loss": 1.0465,
+      "step": 11910
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3202678399796349e-05,
+      "loss": 0.9454,
+      "step": 11911
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.32016110209891e-05,
+      "loss": 1.018,
+      "step": 11912
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3200543601538516e-05,
+      "loss": 0.9349,
+      "step": 11913
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3199476141458147e-05,
+      "loss": 1.0153,
+      "step": 11914
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3198408640761548e-05,
+      "loss": 1.0412,
+      "step": 11915
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3197341099462273e-05,
+      "loss": 1.0475,
+      "step": 11916
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3196273517573865e-05,
+      "loss": 1.0737,
+      "step": 11917
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3195205895109886e-05,
+      "loss": 0.9735,
+      "step": 11918
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3194138232083887e-05,
+      "loss": 0.9441,
+      "step": 11919
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3193070528509413e-05,
+      "loss": 1.0109,
+      "step": 11920
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3192002784400028e-05,
+      "loss": 0.9623,
+      "step": 11921
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3190934999769282e-05,
+      "loss": 1.0401,
+      "step": 11922
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3189867174630735e-05,
+      "loss": 1.0797,
+      "step": 11923
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3188799308997935e-05,
+      "loss": 1.0186,
+      "step": 11924
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3187731402884447e-05,
+      "loss": 1.0751,
+      "step": 11925
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318666345630382e-05,
+      "loss": 0.9368,
+      "step": 11926
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.3185595469269615e-05,
+      "loss": 0.9727,
+      "step": 11927
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.318452744179539e-05,
+      "loss": 1.0377,
+      "step": 11928
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 1.31834593738947e-05,
+      "loss": 0.9658,
+      "step": 11929
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3182391265581107e-05,
+      "loss": 0.8493,
+      "step": 11930
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3181323116868168e-05,
+      "loss": 1.0166,
+      "step": 11931
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3180254927769447e-05,
+      "loss": 1.0791,
+      "step": 11932
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3179186698298501e-05,
+      "loss": 1.0616,
+      "step": 11933
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3178118428468887e-05,
+      "loss": 0.9162,
+      "step": 11934
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3177050118294175e-05,
+      "loss": 1.056,
+      "step": 11935
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317598176778792e-05,
+      "loss": 1.116,
+      "step": 11936
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3174913376963687e-05,
+      "loss": 1.1587,
+      "step": 11937
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3173844945835037e-05,
+      "loss": 1.053,
+      "step": 11938
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3172776474415537e-05,
+      "loss": 1.1321,
+      "step": 11939
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.317170796271875e-05,
+      "loss": 1.0737,
+      "step": 11940
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3170639410758232e-05,
+      "loss": 0.9628,
+      "step": 11941
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3169570818547563e-05,
+      "loss": 1.1028,
+      "step": 11942
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3168502186100296e-05,
+      "loss": 1.157,
+      "step": 11943
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3167433513430004e-05,
+      "loss": 1.1504,
+      "step": 11944
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3166364800550247e-05,
+      "loss": 0.918,
+      "step": 11945
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3165296047474599e-05,
+      "loss": 1.0168,
+      "step": 11946
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3164227254216625e-05,
+      "loss": 1.0031,
+      "step": 11947
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3163158420789888e-05,
+      "loss": 0.93,
+      "step": 11948
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3162089547207964e-05,
+      "loss": 1.0652,
+      "step": 11949
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3161020633484414e-05,
+      "loss": 1.0527,
+      "step": 11950
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3159951679632815e-05,
+      "loss": 1.0455,
+      "step": 11951
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3158882685666731e-05,
+      "loss": 0.9856,
+      "step": 11952
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3157813651599743e-05,
+      "loss": 0.998,
+      "step": 11953
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.315674457744541e-05,
+      "loss": 0.9422,
+      "step": 11954
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3155675463217303e-05,
+      "loss": 0.9938,
+      "step": 11955
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3154606308929005e-05,
+      "loss": 1.0082,
+      "step": 11956
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3153537114594081e-05,
+      "loss": 1.0316,
+      "step": 11957
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3152467880226105e-05,
+      "loss": 1.1076,
+      "step": 11958
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3151398605838652e-05,
+      "loss": 0.963,
+      "step": 11959
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3150329291445297e-05,
+      "loss": 0.9474,
+      "step": 11960
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3149259937059611e-05,
+      "loss": 1.1536,
+      "step": 11961
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314819054269517e-05,
+      "loss": 1.2209,
+      "step": 11962
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3147121108365552e-05,
+      "loss": 0.9555,
+      "step": 11963
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314605163408433e-05,
+      "loss": 0.989,
+      "step": 11964
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3144982119865084e-05,
+      "loss": 0.8769,
+      "step": 11965
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3143912565721387e-05,
+      "loss": 0.9963,
+      "step": 11966
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314284297166682e-05,
+      "loss": 0.9807,
+      "step": 11967
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.314177333771496e-05,
+      "loss": 0.948,
+      "step": 11968
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3140703663879387e-05,
+      "loss": 0.9502,
+      "step": 11969
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3139633950173678e-05,
+      "loss": 0.9969,
+      "step": 11970
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3138564196611412e-05,
+      "loss": 1.0353,
+      "step": 11971
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313749440320617e-05,
+      "loss": 1.0065,
+      "step": 11972
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3136424569971536e-05,
+      "loss": 1.0307,
+      "step": 11973
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3135354696921088e-05,
+      "loss": 0.9261,
+      "step": 11974
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3134284784068408e-05,
+      "loss": 1.1002,
+      "step": 11975
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3133214831427075e-05,
+      "loss": 0.9858,
+      "step": 11976
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.313214483901068e-05,
+      "loss": 0.9775,
+      "step": 11977
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3131074806832798e-05,
+      "loss": 1.0942,
+      "step": 11978
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3130004734907016e-05,
+      "loss": 0.9994,
+      "step": 11979
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3128934623246916e-05,
+      "loss": 1.0824,
+      "step": 11980
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3127864471866088e-05,
+      "loss": 1.0506,
+      "step": 11981
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3126794280778113e-05,
+      "loss": 1.0648,
+      "step": 11982
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3125724049996575e-05,
+      "loss": 1.0129,
+      "step": 11983
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3124653779535068e-05,
+      "loss": 0.9466,
+      "step": 11984
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3123583469407168e-05,
+      "loss": 0.9417,
+      "step": 11985
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312251311962647e-05,
+      "loss": 1.023,
+      "step": 11986
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.312144273020656e-05,
+      "loss": 1.026,
+      "step": 11987
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3120372301161027e-05,
+      "loss": 1.0844,
+      "step": 11988
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3119301832503455e-05,
+      "loss": 1.0224,
+      "step": 11989
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3118231324247437e-05,
+      "loss": 1.1124,
+      "step": 11990
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3117160776406564e-05,
+      "loss": 1.0504,
+      "step": 11991
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3116090188994424e-05,
+      "loss": 1.0089,
+      "step": 11992
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3115019562024607e-05,
+      "loss": 1.0461,
+      "step": 11993
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3113948895510706e-05,
+      "loss": 1.0292,
+      "step": 11994
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3112878189466314e-05,
+      "loss": 1.0372,
+      "step": 11995
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3111807443905019e-05,
+      "loss": 0.9516,
+      "step": 11996
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3110736658840417e-05,
+      "loss": 0.9543,
+      "step": 11997
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3109665834286101e-05,
+      "loss": 0.9115,
+      "step": 11998
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3108594970255664e-05,
+      "loss": 1.0589,
+      "step": 11999
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3107524066762698e-05,
+      "loss": 1.0001,
+      "step": 12000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3106453123820803e-05,
+      "loss": 1.0815,
+      "step": 12001
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3105382141443573e-05,
+      "loss": 1.0296,
+      "step": 12002
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3104311119644597e-05,
+      "loss": 1.037,
+      "step": 12003
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3103240058437479e-05,
+      "loss": 1.0205,
+      "step": 12004
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3102168957835816e-05,
+      "loss": 0.9937,
+      "step": 12005
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3101097817853198e-05,
+      "loss": 0.9134,
+      "step": 12006
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.310002663850323e-05,
+      "loss": 1.0051,
+      "step": 12007
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3098955419799505e-05,
+      "loss": 1.0231,
+      "step": 12008
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3097884161755627e-05,
+      "loss": 1.0881,
+      "step": 12009
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3096812864385189e-05,
+      "loss": 0.9926,
+      "step": 12010
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3095741527701797e-05,
+      "loss": 0.915,
+      "step": 12011
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3094670151719049e-05,
+      "loss": 0.9582,
+      "step": 12012
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3093598736450544e-05,
+      "loss": 0.9948,
+      "step": 12013
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3092527281909881e-05,
+      "loss": 0.9638,
+      "step": 12014
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3091455788110667e-05,
+      "loss": 0.9491,
+      "step": 12015
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3090384255066507e-05,
+      "loss": 1.0502,
+      "step": 12016
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3089312682790993e-05,
+      "loss": 1.0657,
+      "step": 12017
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3088241071297738e-05,
+      "loss": 1.0282,
+      "step": 12018
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308716942060034e-05,
+      "loss": 0.9016,
+      "step": 12019
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3086097730712408e-05,
+      "loss": 1.0311,
+      "step": 12020
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.308502600164754e-05,
+      "loss": 0.975,
+      "step": 12021
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3083954233419347e-05,
+      "loss": 0.967,
+      "step": 12022
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3082882426041437e-05,
+      "loss": 1.0939,
+      "step": 12023
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3081810579527406e-05,
+      "loss": 0.9468,
+      "step": 12024
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3080738693890868e-05,
+      "loss": 0.9505,
+      "step": 12025
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3079666769145433e-05,
+      "loss": 0.9553,
+      "step": 12026
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3078594805304703e-05,
+      "loss": 0.9987,
+      "step": 12027
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3077522802382285e-05,
+      "loss": 0.9031,
+      "step": 12028
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3076450760391794e-05,
+      "loss": 1.1486,
+      "step": 12029
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3075378679346838e-05,
+      "loss": 0.9941,
+      "step": 12030
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3074306559261016e-05,
+      "loss": 1.1065,
+      "step": 12031
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3073234400147956e-05,
+      "loss": 1.115,
+      "step": 12032
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3072162202021257e-05,
+      "loss": 1.1189,
+      "step": 12033
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3071089964894528e-05,
+      "loss": 0.9463,
+      "step": 12034
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.307001768878139e-05,
+      "loss": 0.9084,
+      "step": 12035
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.306894537369545e-05,
+      "loss": 0.9795,
+      "step": 12036
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3067873019650321e-05,
+      "loss": 0.9307,
+      "step": 12037
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3066800626659615e-05,
+      "loss": 1.0694,
+      "step": 12038
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3065728194736948e-05,
+      "loss": 1.017,
+      "step": 12039
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3064655723895936e-05,
+      "loss": 1.0638,
+      "step": 12040
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3063583214150187e-05,
+      "loss": 1.0951,
+      "step": 12041
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3062510665513322e-05,
+      "loss": 0.9383,
+      "step": 12042
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3061438077998952e-05,
+      "loss": 0.9921,
+      "step": 12043
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3060365451620699e-05,
+      "loss": 0.9682,
+      "step": 12044
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305929278639217e-05,
+      "loss": 0.9677,
+      "step": 12045
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3058220082326996e-05,
+      "loss": 0.9635,
+      "step": 12046
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3057147339438785e-05,
+      "loss": 1.0012,
+      "step": 12047
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3056074557741152e-05,
+      "loss": 0.9066,
+      "step": 12048
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3055001737247729e-05,
+      "loss": 1.1032,
+      "step": 12049
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.305392887797212e-05,
+      "loss": 1.0487,
+      "step": 12050
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3052855979927955e-05,
+      "loss": 1.1206,
+      "step": 12051
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3051783043128847e-05,
+      "loss": 1.1132,
+      "step": 12052
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3050710067588424e-05,
+      "loss": 0.9764,
+      "step": 12053
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3049637053320303e-05,
+      "loss": 1.0775,
+      "step": 12054
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3048564000338102e-05,
+      "loss": 0.9561,
+      "step": 12055
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3047490908655449e-05,
+      "loss": 0.9932,
+      "step": 12056
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3046417778285962e-05,
+      "loss": 1.0492,
+      "step": 12057
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3045344609243266e-05,
+      "loss": 1.0174,
+      "step": 12058
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3044271401540987e-05,
+      "loss": 1.0638,
+      "step": 12059
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3043198155192748e-05,
+      "loss": 0.9769,
+      "step": 12060
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3042124870212172e-05,
+      "loss": 0.9117,
+      "step": 12061
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3041051546612879e-05,
+      "loss": 1.0765,
+      "step": 12062
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3039978184408506e-05,
+      "loss": 1.0012,
+      "step": 12063
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3038904783612667e-05,
+      "loss": 1.0735,
+      "step": 12064
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3037831344238996e-05,
+      "loss": 1.0087,
+      "step": 12065
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3036757866301118e-05,
+      "loss": 1.0947,
+      "step": 12066
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.303568434981266e-05,
+      "loss": 0.9633,
+      "step": 12067
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3034610794787253e-05,
+      "loss": 1.1047,
+      "step": 12068
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3033537201238519e-05,
+      "loss": 1.0695,
+      "step": 12069
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3032463569180094e-05,
+      "loss": 1.0138,
+      "step": 12070
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3031389898625602e-05,
+      "loss": 0.8974,
+      "step": 12071
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3030316189588672e-05,
+      "loss": 0.8875,
+      "step": 12072
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3029242442082942e-05,
+      "loss": 1.0369,
+      "step": 12073
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3028168656122035e-05,
+      "loss": 0.9907,
+      "step": 12074
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3027094831719587e-05,
+      "loss": 1.0508,
+      "step": 12075
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3026020968889226e-05,
+      "loss": 1.0613,
+      "step": 12076
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3024947067644591e-05,
+      "loss": 1.1602,
+      "step": 12077
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3023873127999306e-05,
+      "loss": 0.9626,
+      "step": 12078
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.302279914996701e-05,
+      "loss": 1.1141,
+      "step": 12079
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3021725133561334e-05,
+      "loss": 1.033,
+      "step": 12080
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3020651078795919e-05,
+      "loss": 1.0398,
+      "step": 12081
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3019576985684388e-05,
+      "loss": 0.971,
+      "step": 12082
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3018502854240384e-05,
+      "loss": 1.0395,
+      "step": 12083
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3017428684477545e-05,
+      "loss": 0.9591,
+      "step": 12084
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3016354476409499e-05,
+      "loss": 1.023,
+      "step": 12085
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3015280230049889e-05,
+      "loss": 1.0395,
+      "step": 12086
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301420594541235e-05,
+      "loss": 1.1494,
+      "step": 12087
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3013131622510523e-05,
+      "loss": 0.9591,
+      "step": 12088
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.301205726135804e-05,
+      "loss": 1.011,
+      "step": 12089
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3010982861968544e-05,
+      "loss": 1.0675,
+      "step": 12090
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3009908424355672e-05,
+      "loss": 0.9814,
+      "step": 12091
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3008833948533066e-05,
+      "loss": 0.9218,
+      "step": 12092
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3007759434514364e-05,
+      "loss": 0.9885,
+      "step": 12093
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3006684882313208e-05,
+      "loss": 0.9642,
+      "step": 12094
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3005610291943238e-05,
+      "loss": 1.0466,
+      "step": 12095
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3004535663418096e-05,
+      "loss": 0.9691,
+      "step": 12096
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3003460996751422e-05,
+      "loss": 1.0923,
+      "step": 12097
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3002386291956864e-05,
+      "loss": 1.0475,
+      "step": 12098
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.300131154904806e-05,
+      "loss": 0.9027,
+      "step": 12099
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.3000236768038654e-05,
+      "loss": 1.1298,
+      "step": 12100
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2999161948942289e-05,
+      "loss": 0.9785,
+      "step": 12101
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2998087091772616e-05,
+      "loss": 0.8776,
+      "step": 12102
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2997012196543273e-05,
+      "loss": 0.8884,
+      "step": 12103
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2995937263267906e-05,
+      "loss": 0.942,
+      "step": 12104
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2994862291960169e-05,
+      "loss": 0.977,
+      "step": 12105
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2993787282633697e-05,
+      "loss": 0.8747,
+      "step": 12106
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.299271223530214e-05,
+      "loss": 1.0783,
+      "step": 12107
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2991637149979147e-05,
+      "loss": 0.9899,
+      "step": 12108
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2990562026678373e-05,
+      "loss": 1.0648,
+      "step": 12109
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2989486865413454e-05,
+      "loss": 0.9502,
+      "step": 12110
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2988411666198045e-05,
+      "loss": 1.1384,
+      "step": 12111
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2987336429045797e-05,
+      "loss": 1.0053,
+      "step": 12112
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2986261153970354e-05,
+      "loss": 1.0455,
+      "step": 12113
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2985185840985369e-05,
+      "loss": 1.0545,
+      "step": 12114
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2984110490104494e-05,
+      "loss": 1.0235,
+      "step": 12115
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2983035101341382e-05,
+      "loss": 1.1105,
+      "step": 12116
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.298195967470968e-05,
+      "loss": 1.056,
+      "step": 12117
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2980884210223041e-05,
+      "loss": 1.1412,
+      "step": 12118
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297980870789512e-05,
+      "loss": 0.9311,
+      "step": 12119
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.297873316773957e-05,
+      "loss": 1.1061,
+      "step": 12120
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2977657589770042e-05,
+      "loss": 0.8704,
+      "step": 12121
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2976581974000193e-05,
+      "loss": 0.9091,
+      "step": 12122
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2975506320443678e-05,
+      "loss": 1.1062,
+      "step": 12123
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2974430629114143e-05,
+      "loss": 0.9009,
+      "step": 12124
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2973354900025259e-05,
+      "loss": 1.0328,
+      "step": 12125
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2972279133190673e-05,
+      "loss": 0.9288,
+      "step": 12126
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2971203328624039e-05,
+      "loss": 1.1039,
+      "step": 12127
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2970127486339017e-05,
+      "loss": 1.1026,
+      "step": 12128
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2969051606349266e-05,
+      "loss": 1.0523,
+      "step": 12129
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2967975688668444e-05,
+      "loss": 1.0255,
+      "step": 12130
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2966899733310206e-05,
+      "loss": 1.0636,
+      "step": 12131
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2965823740288212e-05,
+      "loss": 0.9727,
+      "step": 12132
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2964747709616127e-05,
+      "loss": 1.0687,
+      "step": 12133
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2963671641307603e-05,
+      "loss": 1.0422,
+      "step": 12134
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2962595535376302e-05,
+      "loss": 1.0355,
+      "step": 12135
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2961519391835888e-05,
+      "loss": 1.0315,
+      "step": 12136
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.296044321070002e-05,
+      "loss": 1.1362,
+      "step": 12137
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2959366991982359e-05,
+      "loss": 0.9769,
+      "step": 12138
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.295829073569657e-05,
+      "loss": 0.8971,
+      "step": 12139
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2957214441856318e-05,
+      "loss": 0.9011,
+      "step": 12140
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2956138110475257e-05,
+      "loss": 1.048,
+      "step": 12141
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2955061741567056e-05,
+      "loss": 1.1219,
+      "step": 12142
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2953985335145381e-05,
+      "loss": 0.9689,
+      "step": 12143
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2952908891223894e-05,
+      "loss": 1.1019,
+      "step": 12144
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2951832409816259e-05,
+      "loss": 1.1159,
+      "step": 12145
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2950755890936147e-05,
+      "loss": 0.9752,
+      "step": 12146
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294967933459722e-05,
+      "loss": 0.9562,
+      "step": 12147
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.294860274081314e-05,
+      "loss": 1.0863,
+      "step": 12148
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2947526109597583e-05,
+      "loss": 1.0552,
+      "step": 12149
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2946449440964213e-05,
+      "loss": 1.053,
+      "step": 12150
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2945372734926698e-05,
+      "loss": 1.0099,
+      "step": 12151
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29442959914987e-05,
+      "loss": 1.0411,
+      "step": 12152
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2943219210693898e-05,
+      "loss": 0.9839,
+      "step": 12153
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2942142392525958e-05,
+      "loss": 0.9769,
+      "step": 12154
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2941065537008544e-05,
+      "loss": 0.9887,
+      "step": 12155
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2939988644155334e-05,
+      "loss": 0.9889,
+      "step": 12156
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2938911713979996e-05,
+      "loss": 1.0199,
+      "step": 12157
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2937834746496201e-05,
+      "loss": 0.9499,
+      "step": 12158
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2936757741717618e-05,
+      "loss": 1.0181,
+      "step": 12159
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2935680699657926e-05,
+      "loss": 1.0634,
+      "step": 12160
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2934603620330792e-05,
+      "loss": 1.0031,
+      "step": 12161
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293352650374989e-05,
+      "loss": 1.04,
+      "step": 12162
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2932449349928895e-05,
+      "loss": 1.0328,
+      "step": 12163
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2931372158881481e-05,
+      "loss": 1.123,
+      "step": 12164
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.293029493062132e-05,
+      "loss": 0.8795,
+      "step": 12165
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292921766516209e-05,
+      "loss": 1.0131,
+      "step": 12166
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2928140362517469e-05,
+      "loss": 1.0421,
+      "step": 12167
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292706302270113e-05,
+      "loss": 1.1114,
+      "step": 12168
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2925985645726741e-05,
+      "loss": 0.8751,
+      "step": 12169
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2924908231607994e-05,
+      "loss": 0.9606,
+      "step": 12170
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2923830780358558e-05,
+      "loss": 0.9683,
+      "step": 12171
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2922753291992109e-05,
+      "loss": 1.0093,
+      "step": 12172
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.292167576652233e-05,
+      "loss": 1.101,
+      "step": 12173
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2920598203962902e-05,
+      "loss": 1.0406,
+      "step": 12174
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.29195206043275e-05,
+      "loss": 1.0219,
+      "step": 12175
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2918442967629798e-05,
+      "loss": 1.0744,
+      "step": 12176
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.291736529388349e-05,
+      "loss": 1.0164,
+      "step": 12177
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2916287583102248e-05,
+      "loss": 1.0416,
+      "step": 12178
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2915209835299751e-05,
+      "loss": 1.1729,
+      "step": 12179
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2914132050489686e-05,
+      "loss": 0.9617,
+      "step": 12180
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2913054228685734e-05,
+      "loss": 0.9053,
+      "step": 12181
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2911976369901575e-05,
+      "loss": 0.9997,
+      "step": 12182
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2910898474150895e-05,
+      "loss": 1.0402,
+      "step": 12183
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2909820541447378e-05,
+      "loss": 1.1335,
+      "step": 12184
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2908742571804704e-05,
+      "loss": 1.0619,
+      "step": 12185
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290766456523656e-05,
+      "loss": 0.9914,
+      "step": 12186
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290658652175663e-05,
+      "loss": 1.0537,
+      "step": 12187
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2905508441378602e-05,
+      "loss": 0.9947,
+      "step": 12188
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290443032411616e-05,
+      "loss": 1.176,
+      "step": 12189
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2903352169982988e-05,
+      "loss": 0.997,
+      "step": 12190
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2902273978992777e-05,
+      "loss": 0.9957,
+      "step": 12191
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.290119575115921e-05,
+      "loss": 1.0928,
+      "step": 12192
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2900117486495977e-05,
+      "loss": 1.0395,
+      "step": 12193
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2899039185016768e-05,
+      "loss": 0.9827,
+      "step": 12194
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.289796084673527e-05,
+      "loss": 1.0603,
+      "step": 12195
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2896882471665173e-05,
+      "loss": 0.8766,
+      "step": 12196
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2895804059820162e-05,
+      "loss": 0.9538,
+      "step": 12197
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2894725611213937e-05,
+      "loss": 0.969,
+      "step": 12198
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2893647125860177e-05,
+      "loss": 1.0141,
+      "step": 12199
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2892568603772578e-05,
+      "loss": 1.0743,
+      "step": 12200
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2891490044964832e-05,
+      "loss": 1.1508,
+      "step": 12201
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2890411449450633e-05,
+      "loss": 0.9641,
+      "step": 12202
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2889332817243669e-05,
+      "loss": 1.0397,
+      "step": 12203
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2888254148357635e-05,
+      "loss": 1.0273,
+      "step": 12204
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2887175442806225e-05,
+      "loss": 0.9066,
+      "step": 12205
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288609670060313e-05,
+      "loss": 1.0726,
+      "step": 12206
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2885017921762049e-05,
+      "loss": 1.0441,
+      "step": 12207
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288393910629667e-05,
+      "loss": 1.1336,
+      "step": 12208
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2882860254220697e-05,
+      "loss": 1.0486,
+      "step": 12209
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2881781365547816e-05,
+      "loss": 0.9207,
+      "step": 12210
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.288070244029173e-05,
+      "loss": 1.0172,
+      "step": 12211
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2879623478466134e-05,
+      "loss": 1.061,
+      "step": 12212
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2878544480084724e-05,
+      "loss": 0.9955,
+      "step": 12213
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2877465445161197e-05,
+      "loss": 1.0051,
+      "step": 12214
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2876386373709253e-05,
+      "loss": 0.9992,
+      "step": 12215
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2875307265742588e-05,
+      "loss": 1.042,
+      "step": 12216
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 1.2874228121274906e-05,
+      "loss": 1.1038,
+      "step": 12217
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.28731489403199e-05,
+      "loss": 0.9825,
+      "step": 12218
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2872069722891274e-05,
+      "loss": 0.9035,
+      "step": 12219
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2870990469002724e-05,
+      "loss": 1.086,
+      "step": 12220
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2869911178667955e-05,
+      "loss": 0.962,
+      "step": 12221
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2868831851900665e-05,
+      "loss": 1.0393,
+      "step": 12222
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286775248871456e-05,
+      "loss": 0.9392,
+      "step": 12223
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2866673089123337e-05,
+      "loss": 1.0168,
+      "step": 12224
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2865593653140704e-05,
+      "loss": 1.0525,
+      "step": 12225
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.286451418078036e-05,
+      "loss": 1.0704,
+      "step": 12226
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2863434672056009e-05,
+      "loss": 1.0451,
+      "step": 12227
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2862355126981356e-05,
+      "loss": 1.0175,
+      "step": 12228
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2861275545570106e-05,
+      "loss": 1.1282,
+      "step": 12229
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2860195927835964e-05,
+      "loss": 1.1529,
+      "step": 12230
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2859116273792633e-05,
+      "loss": 0.8807,
+      "step": 12231
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2858036583453819e-05,
+      "loss": 1.0542,
+      "step": 12232
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.285695685683323e-05,
+      "loss": 1.0112,
+      "step": 12233
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2855877093944575e-05,
+      "loss": 1.0057,
+      "step": 12234
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2854797294801555e-05,
+      "loss": 0.9662,
+      "step": 12235
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2853717459417882e-05,
+      "loss": 0.9893,
+      "step": 12236
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2852637587807268e-05,
+      "loss": 1.0217,
+      "step": 12237
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2851557679983411e-05,
+      "loss": 1.132,
+      "step": 12238
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2850477735960028e-05,
+      "loss": 0.9005,
+      "step": 12239
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2849397755750828e-05,
+      "loss": 0.9466,
+      "step": 12240
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2848317739369518e-05,
+      "loss": 1.1336,
+      "step": 12241
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284723768682981e-05,
+      "loss": 1.109,
+      "step": 12242
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2846157598145412e-05,
+      "loss": 1.0113,
+      "step": 12243
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2845077473330043e-05,
+      "loss": 1.1281,
+      "step": 12244
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2843997312397407e-05,
+      "loss": 1.0626,
+      "step": 12245
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.284291711536122e-05,
+      "loss": 0.9988,
+      "step": 12246
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2841836882235198e-05,
+      "loss": 0.9773,
+      "step": 12247
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2840756613033044e-05,
+      "loss": 0.8919,
+      "step": 12248
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2839676307768479e-05,
+      "loss": 1.0345,
+      "step": 12249
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2838595966455217e-05,
+      "loss": 0.9383,
+      "step": 12250
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2837515589106971e-05,
+      "loss": 1.0728,
+      "step": 12251
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2836435175737456e-05,
+      "loss": 0.9975,
+      "step": 12252
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2835354726360387e-05,
+      "loss": 0.9895,
+      "step": 12253
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2834274240989483e-05,
+      "loss": 0.9792,
+      "step": 12254
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2833193719638455e-05,
+      "loss": 0.9597,
+      "step": 12255
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2832113162321028e-05,
+      "loss": 0.997,
+      "step": 12256
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.283103256905091e-05,
+      "loss": 0.9083,
+      "step": 12257
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2829951939841827e-05,
+      "loss": 1.1116,
+      "step": 12258
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2828871274707488e-05,
+      "loss": 0.8558,
+      "step": 12259
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2827790573661622e-05,
+      "loss": 0.8609,
+      "step": 12260
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2826709836717943e-05,
+      "loss": 1.1152,
+      "step": 12261
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2825629063890166e-05,
+      "loss": 1.0262,
+      "step": 12262
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.282454825519202e-05,
+      "loss": 1.0603,
+      "step": 12263
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2823467410637219e-05,
+      "loss": 1.0793,
+      "step": 12264
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2822386530239489e-05,
+      "loss": 0.9506,
+      "step": 12265
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2821305614012543e-05,
+      "loss": 1.1209,
+      "step": 12266
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2820224661970113e-05,
+      "loss": 0.9844,
+      "step": 12267
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2819143674125916e-05,
+      "loss": 1.0797,
+      "step": 12268
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2818062650493673e-05,
+      "loss": 0.9453,
+      "step": 12269
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2816981591087112e-05,
+      "loss": 1.0072,
+      "step": 12270
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2815900495919953e-05,
+      "loss": 0.9864,
+      "step": 12271
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2814819365005921e-05,
+      "loss": 0.9865,
+      "step": 12272
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2813738198358743e-05,
+      "loss": 0.9254,
+      "step": 12273
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2812656995992143e-05,
+      "loss": 0.9621,
+      "step": 12274
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2811575757919846e-05,
+      "loss": 1.0917,
+      "step": 12275
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2810494484155571e-05,
+      "loss": 1.1441,
+      "step": 12276
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2809413174713056e-05,
+      "loss": 0.9247,
+      "step": 12277
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2808331829606022e-05,
+      "loss": 0.9224,
+      "step": 12278
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2807250448848198e-05,
+      "loss": 1.038,
+      "step": 12279
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2806169032453308e-05,
+      "loss": 1.0269,
+      "step": 12280
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2805087580435087e-05,
+      "loss": 1.0379,
+      "step": 12281
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.280400609280726e-05,
+      "loss": 1.0147,
+      "step": 12282
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2802924569583551e-05,
+      "loss": 0.8743,
+      "step": 12283
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2801843010777697e-05,
+      "loss": 0.966,
+      "step": 12284
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2800761416403424e-05,
+      "loss": 0.9804,
+      "step": 12285
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2799679786474464e-05,
+      "loss": 1.0048,
+      "step": 12286
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2798598121004548e-05,
+      "loss": 0.9202,
+      "step": 12287
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279751642000741e-05,
+      "loss": 1.0428,
+      "step": 12288
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2796434683496777e-05,
+      "loss": 1.051,
+      "step": 12289
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2795352911486381e-05,
+      "loss": 0.9414,
+      "step": 12290
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279427110398996e-05,
+      "loss": 1.0224,
+      "step": 12291
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2793189261021242e-05,
+      "loss": 1.168,
+      "step": 12292
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.279210738259396e-05,
+      "loss": 1.0502,
+      "step": 12293
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2791025468721853e-05,
+      "loss": 0.9234,
+      "step": 12294
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2789943519418657e-05,
+      "loss": 0.9782,
+      "step": 12295
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.27888615346981e-05,
+      "loss": 1.1309,
+      "step": 12296
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2787779514573919e-05,
+      "loss": 0.9702,
+      "step": 12297
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2786697459059856e-05,
+      "loss": 0.9752,
+      "step": 12298
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278561536816964e-05,
+      "loss": 1.0437,
+      "step": 12299
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278453324191701e-05,
+      "loss": 0.8334,
+      "step": 12300
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2783451080315703e-05,
+      "loss": 0.9775,
+      "step": 12301
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2782368883379462e-05,
+      "loss": 1.0718,
+      "step": 12302
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2781286651122015e-05,
+      "loss": 1.0191,
+      "step": 12303
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.278020438355711e-05,
+      "loss": 1.0834,
+      "step": 12304
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277912208069848e-05,
+      "loss": 0.9789,
+      "step": 12305
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2778039742559868e-05,
+      "loss": 0.838,
+      "step": 12306
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2776957369155013e-05,
+      "loss": 1.0121,
+      "step": 12307
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2775874960497652e-05,
+      "loss": 0.9457,
+      "step": 12308
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277479251660153e-05,
+      "loss": 1.0374,
+      "step": 12309
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.277371003748039e-05,
+      "loss": 1.0351,
+      "step": 12310
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2772627523147968e-05,
+      "loss": 1.059,
+      "step": 12311
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2771544973618006e-05,
+      "loss": 1.0316,
+      "step": 12312
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2770462388904253e-05,
+      "loss": 1.0491,
+      "step": 12313
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2769379769020446e-05,
+      "loss": 1.0245,
+      "step": 12314
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2768297113980332e-05,
+      "loss": 0.9163,
+      "step": 12315
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2767214423797657e-05,
+      "loss": 0.9708,
+      "step": 12316
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2766131698486158e-05,
+      "loss": 0.9984,
+      "step": 12317
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2765048938059584e-05,
+      "loss": 0.9911,
+      "step": 12318
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2763966142531681e-05,
+      "loss": 1.0441,
+      "step": 12319
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2762883311916194e-05,
+      "loss": 0.9379,
+      "step": 12320
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2761800446226869e-05,
+      "loss": 1.1811,
+      "step": 12321
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2760717545477452e-05,
+      "loss": 1.056,
+      "step": 12322
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2759634609681693e-05,
+      "loss": 1.0218,
+      "step": 12323
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2758551638853336e-05,
+      "loss": 0.9182,
+      "step": 12324
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.275746863300613e-05,
+      "loss": 0.9558,
+      "step": 12325
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2756385592153824e-05,
+      "loss": 1.1055,
+      "step": 12326
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2755302516310165e-05,
+      "loss": 1.042,
+      "step": 12327
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2754219405488903e-05,
+      "loss": 1.0018,
+      "step": 12328
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2753136259703791e-05,
+      "loss": 1.1802,
+      "step": 12329
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2752053078968576e-05,
+      "loss": 1.1041,
+      "step": 12330
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2750969863297007e-05,
+      "loss": 0.9763,
+      "step": 12331
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2749886612702839e-05,
+      "loss": 1.1715,
+      "step": 12332
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2748803327199824e-05,
+      "loss": 1.0486,
+      "step": 12333
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2747720006801706e-05,
+      "loss": 1.0299,
+      "step": 12334
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2746636651522247e-05,
+      "loss": 1.0628,
+      "step": 12335
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2745553261375193e-05,
+      "loss": 1.0046,
+      "step": 12336
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2744469836374303e-05,
+      "loss": 1.0528,
+      "step": 12337
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2743386376533327e-05,
+      "loss": 0.8836,
+      "step": 12338
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2742302881866017e-05,
+      "loss": 1.0223,
+      "step": 12339
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2741219352386134e-05,
+      "loss": 0.9135,
+      "step": 12340
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2740135788107428e-05,
+      "loss": 0.9946,
+      "step": 12341
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2739052189043654e-05,
+      "loss": 0.8798,
+      "step": 12342
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2737968555208574e-05,
+      "loss": 0.9794,
+      "step": 12343
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2736884886615938e-05,
+      "loss": 1.0993,
+      "step": 12344
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2735801183279505e-05,
+      "loss": 0.8656,
+      "step": 12345
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2734717445213031e-05,
+      "loss": 0.9428,
+      "step": 12346
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.273363367243028e-05,
+      "loss": 1.0277,
+      "step": 12347
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2732549864944998e-05,
+      "loss": 1.052,
+      "step": 12348
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2731466022770955e-05,
+      "loss": 1.0659,
+      "step": 12349
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2730382145921903e-05,
+      "loss": 1.012,
+      "step": 12350
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2729298234411606e-05,
+      "loss": 0.9027,
+      "step": 12351
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2728214288253822e-05,
+      "loss": 1.0881,
+      "step": 12352
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272713030746231e-05,
+      "loss": 0.9801,
+      "step": 12353
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2726046292050833e-05,
+      "loss": 1.0203,
+      "step": 12354
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.272496224203315e-05,
+      "loss": 1.0471,
+      "step": 12355
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2723878157423025e-05,
+      "loss": 0.9588,
+      "step": 12356
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2722794038234217e-05,
+      "loss": 1.0424,
+      "step": 12357
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2721709884480493e-05,
+      "loss": 1.0053,
+      "step": 12358
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2720625696175612e-05,
+      "loss": 1.0866,
+      "step": 12359
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2719541473333339e-05,
+      "loss": 1.0154,
+      "step": 12360
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271845721596744e-05,
+      "loss": 1.0689,
+      "step": 12361
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2717372924091671e-05,
+      "loss": 0.9011,
+      "step": 12362
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2716288597719807e-05,
+      "loss": 0.9696,
+      "step": 12363
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2715204236865609e-05,
+      "loss": 0.9388,
+      "step": 12364
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271411984154284e-05,
+      "loss": 0.9725,
+      "step": 12365
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271303541176527e-05,
+      "loss": 1.1115,
+      "step": 12366
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.271195094754666e-05,
+      "loss": 1.0369,
+      "step": 12367
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2710866448900786e-05,
+      "loss": 1.0254,
+      "step": 12368
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2709781915841406e-05,
+      "loss": 0.9929,
+      "step": 12369
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2708697348382292e-05,
+      "loss": 0.9753,
+      "step": 12370
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2707612746537214e-05,
+      "loss": 0.9733,
+      "step": 12371
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270652811031994e-05,
+      "loss": 1.1023,
+      "step": 12372
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2705443439744234e-05,
+      "loss": 1.0642,
+      "step": 12373
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2704358734823872e-05,
+      "loss": 1.1028,
+      "step": 12374
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270327399557262e-05,
+      "loss": 0.9162,
+      "step": 12375
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2702189222004247e-05,
+      "loss": 1.0538,
+      "step": 12376
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.270110441413253e-05,
+      "loss": 0.9466,
+      "step": 12377
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2700019571971236e-05,
+      "loss": 0.9932,
+      "step": 12378
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2698934695534138e-05,
+      "loss": 1.0113,
+      "step": 12379
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2697849784835004e-05,
+      "loss": 1.0665,
+      "step": 12380
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2696764839887613e-05,
+      "loss": 1.0584,
+      "step": 12381
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2695679860705735e-05,
+      "loss": 0.8912,
+      "step": 12382
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269459484730314e-05,
+      "loss": 1.0054,
+      "step": 12383
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2693509799693608e-05,
+      "loss": 1.0244,
+      "step": 12384
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269242471789091e-05,
+      "loss": 0.9614,
+      "step": 12385
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.269133960190882e-05,
+      "loss": 1.0065,
+      "step": 12386
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2690254451761116e-05,
+      "loss": 0.9776,
+      "step": 12387
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2689169267461574e-05,
+      "loss": 0.9775,
+      "step": 12388
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2688084049023967e-05,
+      "loss": 1.0637,
+      "step": 12389
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2686998796462068e-05,
+      "loss": 1.1025,
+      "step": 12390
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2685913509789665e-05,
+      "loss": 1.0697,
+      "step": 12391
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2684828189020526e-05,
+      "loss": 1.0801,
+      "step": 12392
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2683742834168431e-05,
+      "loss": 0.8964,
+      "step": 12393
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268265744524716e-05,
+      "loss": 1.0153,
+      "step": 12394
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.268157202227049e-05,
+      "loss": 1.0224,
+      "step": 12395
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2680486565252201e-05,
+      "loss": 1.0571,
+      "step": 12396
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.267940107420607e-05,
+      "loss": 1.0603,
+      "step": 12397
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2678315549145881e-05,
+      "loss": 1.0546,
+      "step": 12398
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2677229990085412e-05,
+      "loss": 1.167,
+      "step": 12399
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2676144397038441e-05,
+      "loss": 1.1391,
+      "step": 12400
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2675058770018755e-05,
+      "loss": 1.0793,
+      "step": 12401
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2673973109040134e-05,
+      "loss": 0.9587,
+      "step": 12402
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2672887414116357e-05,
+      "loss": 0.9922,
+      "step": 12403
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2671801685261208e-05,
+      "loss": 0.9983,
+      "step": 12404
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2670715922488475e-05,
+      "loss": 1.0205,
+      "step": 12405
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2669630125811932e-05,
+      "loss": 1.0242,
+      "step": 12406
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2668544295245368e-05,
+      "loss": 1.0945,
+      "step": 12407
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2667458430802568e-05,
+      "loss": 1.0349,
+      "step": 12408
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2666372532497318e-05,
+      "loss": 0.9763,
+      "step": 12409
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2665286600343398e-05,
+      "loss": 1.0106,
+      "step": 12410
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2664200634354597e-05,
+      "loss": 0.967,
+      "step": 12411
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2663114634544703e-05,
+      "loss": 1.137,
+      "step": 12412
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2662028600927494e-05,
+      "loss": 0.9477,
+      "step": 12413
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2660942533516764e-05,
+      "loss": 1.0081,
+      "step": 12414
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2659856432326301e-05,
+      "loss": 1.0479,
+      "step": 12415
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2658770297369889e-05,
+      "loss": 1.0182,
+      "step": 12416
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2657684128661317e-05,
+      "loss": 0.9953,
+      "step": 12417
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2656597926214373e-05,
+      "loss": 0.9744,
+      "step": 12418
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.265551169004285e-05,
+      "loss": 0.9909,
+      "step": 12419
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2654425420160533e-05,
+      "loss": 1.0354,
+      "step": 12420
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2653339116581212e-05,
+      "loss": 1.0116,
+      "step": 12421
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2652252779318678e-05,
+      "loss": 0.8906,
+      "step": 12422
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2651166408386724e-05,
+      "loss": 1.0349,
+      "step": 12423
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2650080003799139e-05,
+      "loss": 0.9808,
+      "step": 12424
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264899356556971e-05,
+      "loss": 0.8927,
+      "step": 12425
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2647907093712241e-05,
+      "loss": 0.9448,
+      "step": 12426
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2646820588240511e-05,
+      "loss": 1.0211,
+      "step": 12427
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264573404916832e-05,
+      "loss": 1.0834,
+      "step": 12428
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2644647476509465e-05,
+      "loss": 1.0594,
+      "step": 12429
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.264356087027773e-05,
+      "loss": 0.9907,
+      "step": 12430
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2642474230486915e-05,
+      "loss": 0.8867,
+      "step": 12431
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2641387557150812e-05,
+      "loss": 1.0184,
+      "step": 12432
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2640300850283218e-05,
+      "loss": 0.9449,
+      "step": 12433
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2639214109897928e-05,
+      "loss": 1.1355,
+      "step": 12434
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2638127336008737e-05,
+      "loss": 0.9356,
+      "step": 12435
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2637040528629441e-05,
+      "loss": 1.0726,
+      "step": 12436
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2635953687773838e-05,
+      "loss": 1.0276,
+      "step": 12437
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2634866813455725e-05,
+      "loss": 0.9662,
+      "step": 12438
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2633779905688896e-05,
+      "loss": 1.1033,
+      "step": 12439
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2632692964487157e-05,
+      "loss": 1.0186,
+      "step": 12440
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2631605989864298e-05,
+      "loss": 0.9895,
+      "step": 12441
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2630518981834119e-05,
+      "loss": 1.0937,
+      "step": 12442
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2629431940410421e-05,
+      "loss": 1.0626,
+      "step": 12443
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2628344865607007e-05,
+      "loss": 1.0183,
+      "step": 12444
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2627257757437672e-05,
+      "loss": 0.9348,
+      "step": 12445
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2626170615916217e-05,
+      "loss": 1.0807,
+      "step": 12446
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2625083441056445e-05,
+      "loss": 0.9616,
+      "step": 12447
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2623996232872155e-05,
+      "loss": 1.0581,
+      "step": 12448
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2622908991377152e-05,
+      "loss": 0.9476,
+      "step": 12449
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2621821716585234e-05,
+      "loss": 1.0762,
+      "step": 12450
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2620734408510208e-05,
+      "loss": 1.0598,
+      "step": 12451
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2619647067165874e-05,
+      "loss": 1.0154,
+      "step": 12452
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2618559692566036e-05,
+      "loss": 0.8748,
+      "step": 12453
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2617472284724498e-05,
+      "loss": 0.9904,
+      "step": 12454
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2616384843655066e-05,
+      "loss": 1.0406,
+      "step": 12455
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.261529736937154e-05,
+      "loss": 0.9504,
+      "step": 12456
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2614209861887728e-05,
+      "loss": 1.1415,
+      "step": 12457
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2613122321217436e-05,
+      "loss": 1.0034,
+      "step": 12458
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2612034747374471e-05,
+      "loss": 1.0224,
+      "step": 12459
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2610947140372636e-05,
+      "loss": 0.9569,
+      "step": 12460
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2609859500225743e-05,
+      "loss": 1.1321,
+      "step": 12461
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2608771826947593e-05,
+      "loss": 0.9761,
+      "step": 12462
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2607684120551995e-05,
+      "loss": 1.0506,
+      "step": 12463
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2606596381052762e-05,
+      "loss": 1.0157,
+      "step": 12464
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2605508608463699e-05,
+      "loss": 1.0261,
+      "step": 12465
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2604420802798614e-05,
+      "loss": 1.0645,
+      "step": 12466
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2603332964071316e-05,
+      "loss": 1.0108,
+      "step": 12467
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2602245092295618e-05,
+      "loss": 0.9842,
+      "step": 12468
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2601157187485327e-05,
+      "loss": 1.1263,
+      "step": 12469
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2600069249654254e-05,
+      "loss": 0.9956,
+      "step": 12470
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.259898127881621e-05,
+      "loss": 1.0223,
+      "step": 12471
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2597893274985009e-05,
+      "loss": 0.9826,
+      "step": 12472
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2596805238174462e-05,
+      "loss": 1.078,
+      "step": 12473
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2595717168398377e-05,
+      "loss": 1.1271,
+      "step": 12474
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2594629065670572e-05,
+      "loss": 1.0904,
+      "step": 12475
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2593540930004854e-05,
+      "loss": 0.9558,
+      "step": 12476
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2592452761415049e-05,
+      "loss": 0.9866,
+      "step": 12477
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2591364559914956e-05,
+      "loss": 1.0144,
+      "step": 12478
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2590276325518395e-05,
+      "loss": 1.1207,
+      "step": 12479
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2589188058239184e-05,
+      "loss": 1.0556,
+      "step": 12480
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2588099758091135e-05,
+      "loss": 0.9646,
+      "step": 12481
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2587011425088065e-05,
+      "loss": 0.952,
+      "step": 12482
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2585923059243785e-05,
+      "loss": 1.0658,
+      "step": 12483
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258483466057212e-05,
+      "loss": 1.0196,
+      "step": 12484
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2583746229086881e-05,
+      "loss": 0.8023,
+      "step": 12485
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.258265776480189e-05,
+      "loss": 1.0146,
+      "step": 12486
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2581569267730957e-05,
+      "loss": 1.1206,
+      "step": 12487
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2580480737887908e-05,
+      "loss": 0.909,
+      "step": 12488
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2579392175286555e-05,
+      "loss": 1.0063,
+      "step": 12489
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257830357994072e-05,
+      "loss": 0.8849,
+      "step": 12490
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2577214951864225e-05,
+      "loss": 1.0331,
+      "step": 12491
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2576126291070884e-05,
+      "loss": 1.0434,
+      "step": 12492
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2575037597574523e-05,
+      "loss": 0.8878,
+      "step": 12493
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257394887138896e-05,
+      "loss": 0.9111,
+      "step": 12494
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2572860112528012e-05,
+      "loss": 1.0444,
+      "step": 12495
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2571771321005508e-05,
+      "loss": 1.0406,
+      "step": 12496
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.257068249683526e-05,
+      "loss": 1.0814,
+      "step": 12497
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2569593640031104e-05,
+      "loss": 0.9002,
+      "step": 12498
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.256850475060685e-05,
+      "loss": 1.1651,
+      "step": 12499
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2567415828576327e-05,
+      "loss": 0.9429,
+      "step": 12500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2566326873953358e-05,
+      "loss": 1.0802,
+      "step": 12501
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2565237886751766e-05,
+      "loss": 0.9458,
+      "step": 12502
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2564148866985378e-05,
+      "loss": 0.9422,
+      "step": 12503
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 1.2563059814668011e-05,
+      "loss": 0.9793,
+      "step": 12504
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2561970729813502e-05,
+      "loss": 1.039,
+      "step": 12505
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2560881612435665e-05,
+      "loss": 0.9115,
+      "step": 12506
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2559792462548332e-05,
+      "loss": 1.0219,
+      "step": 12507
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2558703280165331e-05,
+      "loss": 1.0191,
+      "step": 12508
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2557614065300485e-05,
+      "loss": 0.9158,
+      "step": 12509
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2556524817967623e-05,
+      "loss": 1.1021,
+      "step": 12510
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2555435538180566e-05,
+      "loss": 1.0269,
+      "step": 12511
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2554346225953156e-05,
+      "loss": 0.9538,
+      "step": 12512
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2553256881299211e-05,
+      "loss": 1.0024,
+      "step": 12513
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255216750423256e-05,
+      "loss": 0.9816,
+      "step": 12514
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.255107809476704e-05,
+      "loss": 0.9591,
+      "step": 12515
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254998865291647e-05,
+      "loss": 0.965,
+      "step": 12516
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2548899178694687e-05,
+      "loss": 0.9867,
+      "step": 12517
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254780967211552e-05,
+      "loss": 1.0199,
+      "step": 12518
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25467201331928e-05,
+      "loss": 1.0847,
+      "step": 12519
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2545630561940359e-05,
+      "loss": 1.0698,
+      "step": 12520
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2544540958372024e-05,
+      "loss": 0.9565,
+      "step": 12521
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2543451322501637e-05,
+      "loss": 1.0055,
+      "step": 12522
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254236165434302e-05,
+      "loss": 1.0059,
+      "step": 12523
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.254127195391001e-05,
+      "loss": 1.0269,
+      "step": 12524
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2540182221216442e-05,
+      "loss": 1.0853,
+      "step": 12525
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2539092456276152e-05,
+      "loss": 1.0653,
+      "step": 12526
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2538002659102965e-05,
+      "loss": 0.8555,
+      "step": 12527
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2536912829710725e-05,
+      "loss": 0.9819,
+      "step": 12528
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2535822968113264e-05,
+      "loss": 0.9342,
+      "step": 12529
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2534733074324414e-05,
+      "loss": 1.0328,
+      "step": 12530
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2533643148358015e-05,
+      "loss": 0.9335,
+      "step": 12531
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.25325531902279e-05,
+      "loss": 1.0703,
+      "step": 12532
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2531463199947913e-05,
+      "loss": 0.9653,
+      "step": 12533
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.253037317753188e-05,
+      "loss": 1.0199,
+      "step": 12534
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2529283122993645e-05,
+      "loss": 1.0714,
+      "step": 12535
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2528193036347047e-05,
+      "loss": 1.0917,
+      "step": 12536
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252710291760592e-05,
+      "loss": 1.0998,
+      "step": 12537
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2526012766784103e-05,
+      "loss": 1.0952,
+      "step": 12538
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252492258389544e-05,
+      "loss": 0.9548,
+      "step": 12539
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2523832368953768e-05,
+      "loss": 1.0207,
+      "step": 12540
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2522742121972927e-05,
+      "loss": 1.1309,
+      "step": 12541
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252165184296675e-05,
+      "loss": 1.0575,
+      "step": 12542
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.252056153194909e-05,
+      "loss": 1.0788,
+      "step": 12543
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2519471188933782e-05,
+      "loss": 1.0695,
+      "step": 12544
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2518380813934668e-05,
+      "loss": 1.0156,
+      "step": 12545
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2517290406965588e-05,
+      "loss": 1.0593,
+      "step": 12546
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2516199968040389e-05,
+      "loss": 0.9898,
+      "step": 12547
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2515109497172906e-05,
+      "loss": 1.0389,
+      "step": 12548
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2514018994376992e-05,
+      "loss": 1.1454,
+      "step": 12549
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.251292845966648e-05,
+      "loss": 0.956,
+      "step": 12550
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2511837893055226e-05,
+      "loss": 0.9013,
+      "step": 12551
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2510747294557063e-05,
+      "loss": 1.0035,
+      "step": 12552
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2509656664185843e-05,
+      "loss": 1.0997,
+      "step": 12553
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250856600195541e-05,
+      "loss": 0.9557,
+      "step": 12554
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2507475307879607e-05,
+      "loss": 1.0401,
+      "step": 12555
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.250638458197228e-05,
+      "loss": 0.893,
+      "step": 12556
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2505293824247278e-05,
+      "loss": 0.9741,
+      "step": 12557
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2504203034718447e-05,
+      "loss": 0.9529,
+      "step": 12558
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2503112213399633e-05,
+      "loss": 1.0006,
+      "step": 12559
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2502021360304686e-05,
+      "loss": 0.9824,
+      "step": 12560
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2500930475447453e-05,
+      "loss": 0.998,
+      "step": 12561
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249983955884178e-05,
+      "loss": 1.0444,
+      "step": 12562
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2498748610501517e-05,
+      "loss": 0.9902,
+      "step": 12563
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2497657630440514e-05,
+      "loss": 0.9796,
+      "step": 12564
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2496566618672621e-05,
+      "loss": 0.9922,
+      "step": 12565
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2495475575211686e-05,
+      "loss": 0.9246,
+      "step": 12566
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2494384500071564e-05,
+      "loss": 0.9161,
+      "step": 12567
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2493293393266102e-05,
+      "loss": 1.0233,
+      "step": 12568
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249220225480915e-05,
+      "loss": 1.1442,
+      "step": 12569
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.249111108471456e-05,
+      "loss": 1.078,
+      "step": 12570
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2490019882996187e-05,
+      "loss": 1.0079,
+      "step": 12571
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2488928649667884e-05,
+      "loss": 0.8994,
+      "step": 12572
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2487837384743502e-05,
+      "loss": 1.0504,
+      "step": 12573
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2486746088236892e-05,
+      "loss": 0.8405,
+      "step": 12574
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248565476016191e-05,
+      "loss": 1.0434,
+      "step": 12575
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248456340053241e-05,
+      "loss": 1.0079,
+      "step": 12576
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2483472009362245e-05,
+      "loss": 0.9605,
+      "step": 12577
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248238058666527e-05,
+      "loss": 1.0808,
+      "step": 12578
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2481289132455345e-05,
+      "loss": 1.0237,
+      "step": 12579
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.248019764674632e-05,
+      "loss": 1.1977,
+      "step": 12580
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247910612955205e-05,
+      "loss": 0.9456,
+      "step": 12581
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24780145808864e-05,
+      "loss": 1.0369,
+      "step": 12582
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2476923000763214e-05,
+      "loss": 0.958,
+      "step": 12583
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247583138919636e-05,
+      "loss": 0.9766,
+      "step": 12584
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2474739746199692e-05,
+      "loss": 0.9617,
+      "step": 12585
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.247364807178707e-05,
+      "loss": 0.9297,
+      "step": 12586
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2472556365972346e-05,
+      "loss": 1.0525,
+      "step": 12587
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2471464628769385e-05,
+      "loss": 1.0477,
+      "step": 12588
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2470372860192046e-05,
+      "loss": 1.157,
+      "step": 12589
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2469281060254184e-05,
+      "loss": 1.1705,
+      "step": 12590
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2468189228969665e-05,
+      "loss": 0.9842,
+      "step": 12591
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2467097366352346e-05,
+      "loss": 0.9641,
+      "step": 12592
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2466005472416088e-05,
+      "loss": 0.971,
+      "step": 12593
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2464913547174751e-05,
+      "loss": 1.0524,
+      "step": 12594
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2463821590642198e-05,
+      "loss": 1.1135,
+      "step": 12595
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2462729602832293e-05,
+      "loss": 1.2022,
+      "step": 12596
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2461637583758894e-05,
+      "loss": 0.9612,
+      "step": 12597
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.246054553343587e-05,
+      "loss": 0.9145,
+      "step": 12598
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2459453451877078e-05,
+      "loss": 1.1454,
+      "step": 12599
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2458361339096385e-05,
+      "loss": 1.2274,
+      "step": 12600
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2457269195107653e-05,
+      "loss": 1.0357,
+      "step": 12601
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2456177019924747e-05,
+      "loss": 1.1204,
+      "step": 12602
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2455084813561536e-05,
+      "loss": 0.9752,
+      "step": 12603
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2453992576031877e-05,
+      "loss": 0.9859,
+      "step": 12604
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2452900307349643e-05,
+      "loss": 1.0683,
+      "step": 12605
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2451808007528693e-05,
+      "loss": 0.9508,
+      "step": 12606
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.24507156765829e-05,
+      "loss": 0.9152,
+      "step": 12607
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2449623314526132e-05,
+      "loss": 0.9579,
+      "step": 12608
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2448530921372247e-05,
+      "loss": 0.9977,
+      "step": 12609
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244743849713512e-05,
+      "loss": 0.9536,
+      "step": 12610
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2446346041828614e-05,
+      "loss": 1.0664,
+      "step": 12611
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2445253555466603e-05,
+      "loss": 1.0636,
+      "step": 12612
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2444161038062952e-05,
+      "loss": 0.9601,
+      "step": 12613
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244306848963153e-05,
+      "loss": 1.0634,
+      "step": 12614
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.244197591018621e-05,
+      "loss": 0.9301,
+      "step": 12615
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2440883299740858e-05,
+      "loss": 0.9777,
+      "step": 12616
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2439790658309345e-05,
+      "loss": 0.9892,
+      "step": 12617
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243869798590554e-05,
+      "loss": 0.9688,
+      "step": 12618
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2437605282543321e-05,
+      "loss": 0.9381,
+      "step": 12619
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2436512548236554e-05,
+      "loss": 0.9307,
+      "step": 12620
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243541978299911e-05,
+      "loss": 1.1194,
+      "step": 12621
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2434326986844866e-05,
+      "loss": 1.096,
+      "step": 12622
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2433234159787691e-05,
+      "loss": 1.0926,
+      "step": 12623
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2432141301841456e-05,
+      "loss": 1.1061,
+      "step": 12624
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.243104841302004e-05,
+      "loss": 1.0155,
+      "step": 12625
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2429955493337313e-05,
+      "loss": 0.9665,
+      "step": 12626
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2428862542807149e-05,
+      "loss": 1.1183,
+      "step": 12627
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2427769561443426e-05,
+      "loss": 1.0827,
+      "step": 12628
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2426676549260019e-05,
+      "loss": 1.0226,
+      "step": 12629
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2425583506270797e-05,
+      "loss": 1.0783,
+      "step": 12630
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2424490432489643e-05,
+      "loss": 0.8797,
+      "step": 12631
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2423397327930429e-05,
+      "loss": 1.0317,
+      "step": 12632
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2422304192607036e-05,
+      "loss": 0.9562,
+      "step": 12633
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2421211026533335e-05,
+      "loss": 1.1007,
+      "step": 12634
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2420117829723205e-05,
+      "loss": 1.0189,
+      "step": 12635
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241902460219053e-05,
+      "loss": 0.9303,
+      "step": 12636
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2417931343949178e-05,
+      "loss": 0.9155,
+      "step": 12637
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2416838055013037e-05,
+      "loss": 0.9533,
+      "step": 12638
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2415744735395977e-05,
+      "loss": 0.9124,
+      "step": 12639
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2414651385111887e-05,
+      "loss": 1.047,
+      "step": 12640
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.241355800417464e-05,
+      "loss": 1.0567,
+      "step": 12641
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2412464592598114e-05,
+      "loss": 0.9815,
+      "step": 12642
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2411371150396199e-05,
+      "loss": 0.9619,
+      "step": 12643
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2410277677582767e-05,
+      "loss": 0.9944,
+      "step": 12644
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2409184174171703e-05,
+      "loss": 1.0076,
+      "step": 12645
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2408090640176887e-05,
+      "loss": 0.9532,
+      "step": 12646
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2406997075612204e-05,
+      "loss": 1.0242,
+      "step": 12647
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2405903480491532e-05,
+      "loss": 1.1423,
+      "step": 12648
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2404809854828757e-05,
+      "loss": 1.1367,
+      "step": 12649
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2403716198637766e-05,
+      "loss": 0.962,
+      "step": 12650
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2402622511932431e-05,
+      "loss": 1.1144,
+      "step": 12651
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2401528794726647e-05,
+      "loss": 0.8654,
+      "step": 12652
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2400435047034291e-05,
+      "loss": 0.9291,
+      "step": 12653
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2399341268869256e-05,
+      "loss": 1.0205,
+      "step": 12654
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2398247460245418e-05,
+      "loss": 0.9079,
+      "step": 12655
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2397153621176669e-05,
+      "loss": 0.9225,
+      "step": 12656
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2396059751676895e-05,
+      "loss": 0.9852,
+      "step": 12657
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2394965851759974e-05,
+      "loss": 1.0401,
+      "step": 12658
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2393871921439803e-05,
+      "loss": 1.0848,
+      "step": 12659
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2392777960730263e-05,
+      "loss": 0.9185,
+      "step": 12660
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2391683969645244e-05,
+      "loss": 1.0515,
+      "step": 12661
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2390589948198635e-05,
+      "loss": 1.0545,
+      "step": 12662
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2389495896404318e-05,
+      "loss": 0.9227,
+      "step": 12663
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.238840181427619e-05,
+      "loss": 1.122,
+      "step": 12664
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2387307701828132e-05,
+      "loss": 0.9282,
+      "step": 12665
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2386213559074038e-05,
+      "loss": 1.0549,
+      "step": 12666
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2385119386027798e-05,
+      "loss": 0.8493,
+      "step": 12667
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2384025182703301e-05,
+      "loss": 1.0518,
+      "step": 12668
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2382930949114438e-05,
+      "loss": 0.9362,
+      "step": 12669
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2381836685275097e-05,
+      "loss": 1.1444,
+      "step": 12670
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2380742391199176e-05,
+      "loss": 0.9908,
+      "step": 12671
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237964806690056e-05,
+      "loss": 0.9486,
+      "step": 12672
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2378553712393142e-05,
+      "loss": 1.001,
+      "step": 12673
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2377459327690817e-05,
+      "loss": 1.0342,
+      "step": 12674
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2376364912807479e-05,
+      "loss": 1.1262,
+      "step": 12675
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2375270467757018e-05,
+      "loss": 1.0171,
+      "step": 12676
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2374175992553327e-05,
+      "loss": 1.1131,
+      "step": 12677
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2373081487210305e-05,
+      "loss": 1.0227,
+      "step": 12678
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2371986951741842e-05,
+      "loss": 0.9161,
+      "step": 12679
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.237089238616183e-05,
+      "loss": 1.0563,
+      "step": 12680
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236979779048417e-05,
+      "loss": 0.9697,
+      "step": 12681
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2368703164722757e-05,
+      "loss": 1.1129,
+      "step": 12682
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2367608508891484e-05,
+      "loss": 0.9718,
+      "step": 12683
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2366513823004249e-05,
+      "loss": 1.0711,
+      "step": 12684
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2365419107074948e-05,
+      "loss": 0.9261,
+      "step": 12685
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.236432436111748e-05,
+      "loss": 0.9914,
+      "step": 12686
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2363229585145737e-05,
+      "loss": 1.0139,
+      "step": 12687
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2362134779173622e-05,
+      "loss": 1.0288,
+      "step": 12688
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2361039943215033e-05,
+      "loss": 1.0676,
+      "step": 12689
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2359945077283865e-05,
+      "loss": 0.9655,
+      "step": 12690
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235885018139402e-05,
+      "loss": 0.9678,
+      "step": 12691
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2357755255559395e-05,
+      "loss": 0.9883,
+      "step": 12692
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2356660299793897e-05,
+      "loss": 1.0758,
+      "step": 12693
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2355565314111415e-05,
+      "loss": 1.0541,
+      "step": 12694
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2354470298525856e-05,
+      "loss": 0.9957,
+      "step": 12695
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.235337525305112e-05,
+      "loss": 0.9272,
+      "step": 12696
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2352280177701104e-05,
+      "loss": 0.9699,
+      "step": 12697
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2351185072489717e-05,
+      "loss": 0.9311,
+      "step": 12698
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2350089937430859e-05,
+      "loss": 1.0926,
+      "step": 12699
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2348994772538427e-05,
+      "loss": 1.0864,
+      "step": 12700
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.234789957782633e-05,
+      "loss": 1.0838,
+      "step": 12701
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2346804353308467e-05,
+      "loss": 1.0306,
+      "step": 12702
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2345709098998743e-05,
+      "loss": 0.9878,
+      "step": 12703
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2344613814911062e-05,
+      "loss": 1.0944,
+      "step": 12704
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2343518501059328e-05,
+      "loss": 1.137,
+      "step": 12705
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2342423157457447e-05,
+      "loss": 0.9911,
+      "step": 12706
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2341327784119323e-05,
+      "loss": 1.004,
+      "step": 12707
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2340232381058859e-05,
+      "loss": 0.9358,
+      "step": 12708
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2339136948289965e-05,
+      "loss": 1.0212,
+      "step": 12709
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2338041485826545e-05,
+      "loss": 1.0318,
+      "step": 12710
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2336945993682501e-05,
+      "loss": 1.1438,
+      "step": 12711
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2335850471871751e-05,
+      "loss": 1.104,
+      "step": 12712
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2334754920408191e-05,
+      "loss": 1.0,
+      "step": 12713
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2333659339305735e-05,
+      "loss": 1.0562,
+      "step": 12714
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2332563728578293e-05,
+      "loss": 1.0431,
+      "step": 12715
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2331468088239764e-05,
+      "loss": 1.0252,
+      "step": 12716
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2330372418304068e-05,
+      "loss": 0.9537,
+      "step": 12717
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2329276718785102e-05,
+      "loss": 1.0902,
+      "step": 12718
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2328180989696787e-05,
+      "loss": 1.0076,
+      "step": 12719
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2327085231053027e-05,
+      "loss": 0.9046,
+      "step": 12720
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2325989442867734e-05,
+      "loss": 0.8625,
+      "step": 12721
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2324893625154819e-05,
+      "loss": 1.0443,
+      "step": 12722
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2323797777928188e-05,
+      "loss": 1.0258,
+      "step": 12723
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.232270190120176e-05,
+      "loss": 1.0349,
+      "step": 12724
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2321605994989441e-05,
+      "loss": 1.0209,
+      "step": 12725
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2320510059305148e-05,
+      "loss": 0.8864,
+      "step": 12726
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2319414094162789e-05,
+      "loss": 1.0055,
+      "step": 12727
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2318318099576278e-05,
+      "loss": 1.0797,
+      "step": 12728
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2317222075559533e-05,
+      "loss": 0.8411,
+      "step": 12729
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.231612602212646e-05,
+      "loss": 0.9323,
+      "step": 12730
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2315029939290979e-05,
+      "loss": 1.0238,
+      "step": 12731
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2313933827067e-05,
+      "loss": 0.9895,
+      "step": 12732
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2312837685468441e-05,
+      "loss": 1.0525,
+      "step": 12733
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2311741514509215e-05,
+      "loss": 0.9213,
+      "step": 12734
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2310645314203238e-05,
+      "loss": 1.0385,
+      "step": 12735
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2309549084564429e-05,
+      "loss": 0.8057,
+      "step": 12736
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2308452825606701e-05,
+      "loss": 1.0514,
+      "step": 12737
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2307356537343971e-05,
+      "loss": 0.9979,
+      "step": 12738
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2306260219790154e-05,
+      "loss": 1.0212,
+      "step": 12739
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2305163872959173e-05,
+      "loss": 0.9201,
+      "step": 12740
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.230406749686494e-05,
+      "loss": 1.0234,
+      "step": 12741
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2302971091521376e-05,
+      "loss": 0.9946,
+      "step": 12742
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2301874656942402e-05,
+      "loss": 1.0763,
+      "step": 12743
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2300778193141929e-05,
+      "loss": 1.2558,
+      "step": 12744
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2299681700133883e-05,
+      "loss": 0.9885,
+      "step": 12745
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2298585177932181e-05,
+      "loss": 1.0293,
+      "step": 12746
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2297488626550746e-05,
+      "loss": 1.0418,
+      "step": 12747
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2296392046003494e-05,
+      "loss": 0.876,
+      "step": 12748
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2295295436304349e-05,
+      "loss": 1.1146,
+      "step": 12749
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.229419879746723e-05,
+      "loss": 1.0471,
+      "step": 12750
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2293102129506059e-05,
+      "loss": 1.0442,
+      "step": 12751
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2292005432434757e-05,
+      "loss": 0.953,
+      "step": 12752
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2290908706267247e-05,
+      "loss": 0.9816,
+      "step": 12753
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2289811951017453e-05,
+      "loss": 1.059,
+      "step": 12754
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2288715166699297e-05,
+      "loss": 1.0146,
+      "step": 12755
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.22876183533267e-05,
+      "loss": 0.9792,
+      "step": 12756
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2286521510913591e-05,
+      "loss": 1.0329,
+      "step": 12757
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2285424639473888e-05,
+      "loss": 1.0446,
+      "step": 12758
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2284327739021519e-05,
+      "loss": 1.0756,
+      "step": 12759
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2283230809570408e-05,
+      "loss": 0.9397,
+      "step": 12760
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.228213385113448e-05,
+      "loss": 0.9918,
+      "step": 12761
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2281036863727658e-05,
+      "loss": 1.0145,
+      "step": 12762
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2279939847363874e-05,
+      "loss": 1.0072,
+      "step": 12763
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.227884280205705e-05,
+      "loss": 0.9253,
+      "step": 12764
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2277745727821111e-05,
+      "loss": 1.0434,
+      "step": 12765
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2276648624669987e-05,
+      "loss": 0.877,
+      "step": 12766
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2275551492617604e-05,
+      "loss": 1.1117,
+      "step": 12767
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2274454331677892e-05,
+      "loss": 0.9805,
+      "step": 12768
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2273357141864775e-05,
+      "loss": 0.879,
+      "step": 12769
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2272259923192182e-05,
+      "loss": 0.9762,
+      "step": 12770
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2271162675674048e-05,
+      "loss": 1.0618,
+      "step": 12771
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2270065399324295e-05,
+      "loss": 1.0849,
+      "step": 12772
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2268968094156856e-05,
+      "loss": 1.0037,
+      "step": 12773
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2267870760185661e-05,
+      "loss": 1.0275,
+      "step": 12774
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226677339742464e-05,
+      "loss": 0.9623,
+      "step": 12775
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.226567600588772e-05,
+      "loss": 1.0916,
+      "step": 12776
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2264578585588838e-05,
+      "loss": 0.9583,
+      "step": 12777
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2263481136541921e-05,
+      "loss": 1.0565,
+      "step": 12778
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2262383658760903e-05,
+      "loss": 0.9875,
+      "step": 12779
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2261286152259713e-05,
+      "loss": 0.9381,
+      "step": 12780
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2260188617052286e-05,
+      "loss": 0.8637,
+      "step": 12781
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2259091053152559e-05,
+      "loss": 1.0937,
+      "step": 12782
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2257993460574458e-05,
+      "loss": 1.0062,
+      "step": 12783
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2256895839331918e-05,
+      "loss": 1.0833,
+      "step": 12784
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2255798189438875e-05,
+      "loss": 1.046,
+      "step": 12785
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2254700510909264e-05,
+      "loss": 1.0045,
+      "step": 12786
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2253602803757014e-05,
+      "loss": 0.9786,
+      "step": 12787
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2252505067996065e-05,
+      "loss": 0.956,
+      "step": 12788
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2251407303640356e-05,
+      "loss": 1.0169,
+      "step": 12789
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2250309510703814e-05,
+      "loss": 0.9906,
+      "step": 12790
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.224921168920038e-05,
+      "loss": 1.0348,
+      "step": 12791
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.2248113839143993e-05,
+      "loss": 0.9535,
+      "step": 12792
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2247015960548584e-05,
+      "loss": 1.0881,
+      "step": 12793
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2245918053428092e-05,
+      "loss": 0.851,
+      "step": 12794
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2244820117796455e-05,
+      "loss": 1.0056,
+      "step": 12795
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2243722153667614e-05,
+      "loss": 0.9257,
+      "step": 12796
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2242624161055503e-05,
+      "loss": 1.0354,
+      "step": 12797
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224152613997406e-05,
+      "loss": 1.1307,
+      "step": 12798
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.224042809043723e-05,
+      "loss": 0.8751,
+      "step": 12799
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2239330012458947e-05,
+      "loss": 0.9728,
+      "step": 12800
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223823190605315e-05,
+      "loss": 0.9449,
+      "step": 12801
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2237133771233785e-05,
+      "loss": 1.1212,
+      "step": 12802
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223603560801479e-05,
+      "loss": 0.9877,
+      "step": 12803
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.22349374164101e-05,
+      "loss": 1.0592,
+      "step": 12804
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.223383919643366e-05,
+      "loss": 1.0145,
+      "step": 12805
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2232740948099418e-05,
+      "loss": 0.8978,
+      "step": 12806
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2231642671421304e-05,
+      "loss": 0.9234,
+      "step": 12807
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2230544366413268e-05,
+      "loss": 0.8913,
+      "step": 12808
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2229446033089252e-05,
+      "loss": 0.9129,
+      "step": 12809
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2228347671463198e-05,
+      "loss": 0.9478,
+      "step": 12810
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2227249281549047e-05,
+      "loss": 0.9846,
+      "step": 12811
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2226150863360749e-05,
+      "loss": 0.8911,
+      "step": 12812
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2225052416912244e-05,
+      "loss": 0.9001,
+      "step": 12813
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2223953942217472e-05,
+      "loss": 1.0733,
+      "step": 12814
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2222855439290383e-05,
+      "loss": 0.9586,
+      "step": 12815
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2221756908144924e-05,
+      "loss": 0.9781,
+      "step": 12816
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2220658348795035e-05,
+      "loss": 1.0408,
+      "step": 12817
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2219559761254665e-05,
+      "loss": 1.0034,
+      "step": 12818
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2218461145537762e-05,
+      "loss": 1.0345,
+      "step": 12819
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2217362501658267e-05,
+      "loss": 1.0779,
+      "step": 12820
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2216263829630133e-05,
+      "loss": 0.9805,
+      "step": 12821
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2215165129467301e-05,
+      "loss": 0.9418,
+      "step": 12822
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2214066401183722e-05,
+      "loss": 0.9852,
+      "step": 12823
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2212967644793348e-05,
+      "loss": 0.9249,
+      "step": 12824
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2211868860310119e-05,
+      "loss": 0.9702,
+      "step": 12825
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.221077004774799e-05,
+      "loss": 1.1135,
+      "step": 12826
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2209671207120908e-05,
+      "loss": 0.9303,
+      "step": 12827
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2208572338442821e-05,
+      "loss": 1.025,
+      "step": 12828
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2207473441727683e-05,
+      "loss": 1.1447,
+      "step": 12829
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2206374516989439e-05,
+      "loss": 1.1331,
+      "step": 12830
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2205275564242043e-05,
+      "loss": 1.0497,
+      "step": 12831
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.220417658349944e-05,
+      "loss": 0.9261,
+      "step": 12832
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2203077574775593e-05,
+      "loss": 0.9005,
+      "step": 12833
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2201978538084442e-05,
+      "loss": 1.0167,
+      "step": 12834
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2200879473439941e-05,
+      "loss": 1.1081,
+      "step": 12835
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219978038085605e-05,
+      "loss": 1.0026,
+      "step": 12836
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2198681260346712e-05,
+      "loss": 1.0557,
+      "step": 12837
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2197582111925887e-05,
+      "loss": 0.98,
+      "step": 12838
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2196482935607521e-05,
+      "loss": 1.0035,
+      "step": 12839
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2195383731405576e-05,
+      "loss": 1.0677,
+      "step": 12840
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2194284499333999e-05,
+      "loss": 0.9505,
+      "step": 12841
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2193185239406749e-05,
+      "loss": 1.0532,
+      "step": 12842
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.219208595163778e-05,
+      "loss": 0.9675,
+      "step": 12843
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2190986636041042e-05,
+      "loss": 1.084,
+      "step": 12844
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2189887292630497e-05,
+      "loss": 0.9242,
+      "step": 12845
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2188787921420098e-05,
+      "loss": 0.9882,
+      "step": 12846
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2187688522423802e-05,
+      "loss": 1.0731,
+      "step": 12847
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2186589095655564e-05,
+      "loss": 1.0335,
+      "step": 12848
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2185489641129342e-05,
+      "loss": 0.9938,
+      "step": 12849
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2184390158859094e-05,
+      "loss": 1.0099,
+      "step": 12850
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2183290648858774e-05,
+      "loss": 0.9945,
+      "step": 12851
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2182191111142344e-05,
+      "loss": 0.9208,
+      "step": 12852
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.218109154572376e-05,
+      "loss": 1.0353,
+      "step": 12853
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2179991952616984e-05,
+      "loss": 1.0026,
+      "step": 12854
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.217889233183597e-05,
+      "loss": 0.9641,
+      "step": 12855
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2177792683394678e-05,
+      "loss": 0.9548,
+      "step": 12856
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2176693007307073e-05,
+      "loss": 0.9777,
+      "step": 12857
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2175593303587108e-05,
+      "loss": 1.0117,
+      "step": 12858
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2174493572248745e-05,
+      "loss": 0.9106,
+      "step": 12859
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2173393813305947e-05,
+      "loss": 0.9729,
+      "step": 12860
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2172294026772677e-05,
+      "loss": 0.9509,
+      "step": 12861
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2171194212662892e-05,
+      "loss": 0.9841,
+      "step": 12862
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2170094370990555e-05,
+      "loss": 0.9584,
+      "step": 12863
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.216899450176963e-05,
+      "loss": 1.064,
+      "step": 12864
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2167894605014077e-05,
+      "loss": 0.9604,
+      "step": 12865
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2166794680737857e-05,
+      "loss": 0.9352,
+      "step": 12866
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2165694728954938e-05,
+      "loss": 1.0778,
+      "step": 12867
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2164594749679284e-05,
+      "loss": 0.8881,
+      "step": 12868
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2163494742924853e-05,
+      "loss": 0.918,
+      "step": 12869
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2162394708705614e-05,
+      "loss": 1.081,
+      "step": 12870
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2161294647035531e-05,
+      "loss": 1.0958,
+      "step": 12871
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2160194557928565e-05,
+      "loss": 1.07,
+      "step": 12872
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2159094441398685e-05,
+      "loss": 0.9434,
+      "step": 12873
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2157994297459856e-05,
+      "loss": 1.0363,
+      "step": 12874
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2156894126126046e-05,
+      "loss": 0.9753,
+      "step": 12875
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2155793927411216e-05,
+      "loss": 0.9568,
+      "step": 12876
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2154693701329336e-05,
+      "loss": 1.0644,
+      "step": 12877
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2153593447894374e-05,
+      "loss": 1.074,
+      "step": 12878
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2152493167120297e-05,
+      "loss": 0.9975,
+      "step": 12879
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2151392859021067e-05,
+      "loss": 0.9634,
+      "step": 12880
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2150292523610658e-05,
+      "loss": 1.0664,
+      "step": 12881
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2149192160903041e-05,
+      "loss": 1.0518,
+      "step": 12882
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2148091770912178e-05,
+      "loss": 1.1417,
+      "step": 12883
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214699135365204e-05,
+      "loss": 1.039,
+      "step": 12884
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21458909091366e-05,
+      "loss": 0.9149,
+      "step": 12885
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2144790437379823e-05,
+      "loss": 0.964,
+      "step": 12886
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.214368993839568e-05,
+      "loss": 0.9891,
+      "step": 12887
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2142589412198144e-05,
+      "loss": 1.0543,
+      "step": 12888
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2141488858801185e-05,
+      "loss": 1.0271,
+      "step": 12889
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2140388278218773e-05,
+      "loss": 0.9648,
+      "step": 12890
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2139287670464879e-05,
+      "loss": 0.9629,
+      "step": 12891
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2138187035553477e-05,
+      "loss": 0.9047,
+      "step": 12892
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2137086373498537e-05,
+      "loss": 1.0846,
+      "step": 12893
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2135985684314031e-05,
+      "loss": 0.972,
+      "step": 12894
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2134884968013936e-05,
+      "loss": 1.0123,
+      "step": 12895
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2133784224612221e-05,
+      "loss": 1.0881,
+      "step": 12896
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2132683454122862e-05,
+      "loss": 0.9369,
+      "step": 12897
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2131582656559834e-05,
+      "loss": 1.027,
+      "step": 12898
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2130481831937107e-05,
+      "loss": 1.0547,
+      "step": 12899
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2129380980268656e-05,
+      "loss": 0.9636,
+      "step": 12900
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212828010156846e-05,
+      "loss": 0.911,
+      "step": 12901
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2127179195850491e-05,
+      "loss": 1.0686,
+      "step": 12902
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2126078263128729e-05,
+      "loss": 1.0057,
+      "step": 12903
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2124977303417143e-05,
+      "loss": 1.0187,
+      "step": 12904
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2123876316729712e-05,
+      "loss": 0.9702,
+      "step": 12905
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2122775303080417e-05,
+      "loss": 1.0303,
+      "step": 12906
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.212167426248323e-05,
+      "loss": 1.1804,
+      "step": 12907
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2120573194952127e-05,
+      "loss": 1.1296,
+      "step": 12908
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2119472100501093e-05,
+      "loss": 0.9394,
+      "step": 12909
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.21183709791441e-05,
+      "loss": 1.0802,
+      "step": 12910
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2117269830895127e-05,
+      "loss": 0.8623,
+      "step": 12911
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2116168655768153e-05,
+      "loss": 0.8152,
+      "step": 12912
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2115067453777161e-05,
+      "loss": 0.9017,
+      "step": 12913
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2113966224936125e-05,
+      "loss": 1.0412,
+      "step": 12914
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2112864969259024e-05,
+      "loss": 0.9235,
+      "step": 12915
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2111763686759843e-05,
+      "loss": 1.0073,
+      "step": 12916
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2110662377452562e-05,
+      "loss": 1.0472,
+      "step": 12917
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2109561041351154e-05,
+      "loss": 1.0501,
+      "step": 12918
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2108459678469614e-05,
+      "loss": 0.9364,
+      "step": 12919
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2107358288821911e-05,
+      "loss": 0.9757,
+      "step": 12920
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210625687242203e-05,
+      "loss": 0.9615,
+      "step": 12921
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2105155429283957e-05,
+      "loss": 1.1874,
+      "step": 12922
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.210405395942167e-05,
+      "loss": 1.1275,
+      "step": 12923
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2102952462849153e-05,
+      "loss": 0.9117,
+      "step": 12924
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2101850939580389e-05,
+      "loss": 0.935,
+      "step": 12925
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2100749389629365e-05,
+      "loss": 1.0039,
+      "step": 12926
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209964781301006e-05,
+      "loss": 0.9492,
+      "step": 12927
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2098546209736461e-05,
+      "loss": 1.1362,
+      "step": 12928
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209744457982255e-05,
+      "loss": 1.0278,
+      "step": 12929
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2096342923282314e-05,
+      "loss": 0.953,
+      "step": 12930
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2095241240129739e-05,
+      "loss": 0.9432,
+      "step": 12931
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2094139530378804e-05,
+      "loss": 1.0977,
+      "step": 12932
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2093037794043503e-05,
+      "loss": 0.9487,
+      "step": 12933
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.209193603113782e-05,
+      "loss": 0.9597,
+      "step": 12934
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2090834241675738e-05,
+      "loss": 0.9846,
+      "step": 12935
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2089732425671245e-05,
+      "loss": 1.0284,
+      "step": 12936
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208863058313833e-05,
+      "loss": 0.9467,
+      "step": 12937
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2087528714090983e-05,
+      "loss": 1.0942,
+      "step": 12938
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2086426818543182e-05,
+      "loss": 1.0059,
+      "step": 12939
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.208532489650893e-05,
+      "loss": 1.0084,
+      "step": 12940
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2084222948002204e-05,
+      "loss": 1.0999,
+      "step": 12941
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2083120973036992e-05,
+      "loss": 0.942,
+      "step": 12942
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2082018971627294e-05,
+      "loss": 0.9654,
+      "step": 12943
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2080916943787088e-05,
+      "loss": 1.0044,
+      "step": 12944
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2079814889530373e-05,
+      "loss": 0.96,
+      "step": 12945
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.207871280887113e-05,
+      "loss": 0.9075,
+      "step": 12946
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2077610701823358e-05,
+      "loss": 1.0362,
+      "step": 12947
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2076508568401046e-05,
+      "loss": 0.981,
+      "step": 12948
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2075406408618178e-05,
+      "loss": 1.0368,
+      "step": 12949
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2074304222488755e-05,
+      "loss": 1.0606,
+      "step": 12950
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2073202010026765e-05,
+      "loss": 0.9603,
+      "step": 12951
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.20720997712462e-05,
+      "loss": 0.9409,
+      "step": 12952
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2070997506161048e-05,
+      "loss": 1.0435,
+      "step": 12953
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2069895214785311e-05,
+      "loss": 1.1448,
+      "step": 12954
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2068792897132981e-05,
+      "loss": 1.0747,
+      "step": 12955
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206769055321804e-05,
+      "loss": 1.0301,
+      "step": 12956
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2066588183054497e-05,
+      "loss": 0.8967,
+      "step": 12957
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2065485786656336e-05,
+      "loss": 1.0605,
+      "step": 12958
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2064383364037555e-05,
+      "loss": 0.9409,
+      "step": 12959
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206328091521215e-05,
+      "loss": 1.0497,
+      "step": 12960
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2062178440194116e-05,
+      "loss": 1.1013,
+      "step": 12961
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.206107593899745e-05,
+      "loss": 1.1138,
+      "step": 12962
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2059973411636136e-05,
+      "loss": 0.9418,
+      "step": 12963
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2058870858124188e-05,
+      "loss": 0.9773,
+      "step": 12964
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.205776827847559e-05,
+      "loss": 1.0188,
+      "step": 12965
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2056665672704344e-05,
+      "loss": 1.0591,
+      "step": 12966
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2055563040824444e-05,
+      "loss": 0.9189,
+      "step": 12967
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2054460382849895e-05,
+      "loss": 1.0994,
+      "step": 12968
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2053357698794685e-05,
+      "loss": 1.0175,
+      "step": 12969
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2052254988672818e-05,
+      "loss": 0.9657,
+      "step": 12970
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2051152252498295e-05,
+      "loss": 0.9244,
+      "step": 12971
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2050049490285107e-05,
+      "loss": 1.118,
+      "step": 12972
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2048946702047258e-05,
+      "loss": 0.9884,
+      "step": 12973
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2047843887798745e-05,
+      "loss": 0.9988,
+      "step": 12974
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2046741047553571e-05,
+      "loss": 0.923,
+      "step": 12975
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2045638181325737e-05,
+      "loss": 1.0152,
+      "step": 12976
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2044535289129237e-05,
+      "loss": 0.9912,
+      "step": 12977
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2043432370978081e-05,
+      "loss": 1.0051,
+      "step": 12978
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2042329426886263e-05,
+      "loss": 0.9517,
+      "step": 12979
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2041226456867785e-05,
+      "loss": 0.8966,
+      "step": 12980
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2040123460936653e-05,
+      "loss": 1.0278,
+      "step": 12981
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2039020439106868e-05,
+      "loss": 0.9724,
+      "step": 12982
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203791739139243e-05,
+      "loss": 1.0144,
+      "step": 12983
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2036814317807339e-05,
+      "loss": 0.897,
+      "step": 12984
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.203571121836561e-05,
+      "loss": 0.9299,
+      "step": 12985
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2034608093081233e-05,
+      "loss": 1.0575,
+      "step": 12986
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2033504941968217e-05,
+      "loss": 1.0256,
+      "step": 12987
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2032401765040568e-05,
+      "loss": 1.1201,
+      "step": 12988
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2031298562312291e-05,
+      "loss": 0.9476,
+      "step": 12989
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2030195333797385e-05,
+      "loss": 1.0229,
+      "step": 12990
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202909207950986e-05,
+      "loss": 1.1559,
+      "step": 12991
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2027988799463724e-05,
+      "loss": 0.9329,
+      "step": 12992
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2026885493672976e-05,
+      "loss": 0.9713,
+      "step": 12993
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2025782162151622e-05,
+      "loss": 1.0155,
+      "step": 12994
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2024678804913677e-05,
+      "loss": 1.0984,
+      "step": 12995
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202357542197314e-05,
+      "loss": 1.0135,
+      "step": 12996
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.202247201334402e-05,
+      "loss": 1.0526,
+      "step": 12997
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2021368579040324e-05,
+      "loss": 1.0406,
+      "step": 12998
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2020265119076065e-05,
+      "loss": 0.9925,
+      "step": 12999
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2019161633465241e-05,
+      "loss": 1.0947,
+      "step": 13000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2018058122221868e-05,
+      "loss": 1.04,
+      "step": 13001
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2016954585359953e-05,
+      "loss": 0.9748,
+      "step": 13002
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2015851022893504e-05,
+      "loss": 1.0188,
+      "step": 13003
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.201474743483653e-05,
+      "loss": 1.014,
+      "step": 13004
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2013643821203042e-05,
+      "loss": 0.945,
+      "step": 13005
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2012540182007052e-05,
+      "loss": 1.0063,
+      "step": 13006
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2011436517262565e-05,
+      "loss": 1.0434,
+      "step": 13007
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2010332826983594e-05,
+      "loss": 1.0474,
+      "step": 13008
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2009229111184151e-05,
+      "loss": 0.9464,
+      "step": 13009
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2008125369878249e-05,
+      "loss": 1.0053,
+      "step": 13010
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2007021603079896e-05,
+      "loss": 0.962,
+      "step": 13011
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2005917810803105e-05,
+      "loss": 1.0469,
+      "step": 13012
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200481399306189e-05,
+      "loss": 1.0277,
+      "step": 13013
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.200371014987026e-05,
+      "loss": 1.0651,
+      "step": 13014
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2002606281242229e-05,
+      "loss": 0.9863,
+      "step": 13015
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2001502387191812e-05,
+      "loss": 1.0243,
+      "step": 13016
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.2000398467733026e-05,
+      "loss": 1.0865,
+      "step": 13017
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1999294522879877e-05,
+      "loss": 1.044,
+      "step": 13018
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1998190552646381e-05,
+      "loss": 1.1724,
+      "step": 13019
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1997086557046558e-05,
+      "loss": 0.9446,
+      "step": 13020
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1995982536094417e-05,
+      "loss": 0.9212,
+      "step": 13021
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1994878489803975e-05,
+      "loss": 1.019,
+      "step": 13022
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.199377441818925e-05,
+      "loss": 1.0702,
+      "step": 13023
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1992670321264255e-05,
+      "loss": 1.0628,
+      "step": 13024
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1991566199043008e-05,
+      "loss": 1.1133,
+      "step": 13025
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1990462051539521e-05,
+      "loss": 0.9725,
+      "step": 13026
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1989357878767817e-05,
+      "loss": 0.9789,
+      "step": 13027
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1988253680741907e-05,
+      "loss": 0.8277,
+      "step": 13028
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1987149457475815e-05,
+      "loss": 0.9633,
+      "step": 13029
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1986045208983552e-05,
+      "loss": 0.9181,
+      "step": 13030
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1984940935279144e-05,
+      "loss": 1.0852,
+      "step": 13031
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1983836636376597e-05,
+      "loss": 0.9126,
+      "step": 13032
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1982732312289944e-05,
+      "loss": 1.0475,
+      "step": 13033
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1981627963033196e-05,
+      "loss": 0.9347,
+      "step": 13034
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1980523588620372e-05,
+      "loss": 1.1296,
+      "step": 13035
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1979419189065494e-05,
+      "loss": 0.9497,
+      "step": 13036
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1978314764382581e-05,
+      "loss": 1.0503,
+      "step": 13037
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1977210314585655e-05,
+      "loss": 1.0761,
+      "step": 13038
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1976105839688732e-05,
+      "loss": 0.9229,
+      "step": 13039
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1975001339705839e-05,
+      "loss": 1.0261,
+      "step": 13040
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1973896814650994e-05,
+      "loss": 1.1567,
+      "step": 13041
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1972792264538218e-05,
+      "loss": 1.0486,
+      "step": 13042
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1971687689381536e-05,
+      "loss": 1.0452,
+      "step": 13043
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1970583089194962e-05,
+      "loss": 0.9961,
+      "step": 13044
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1969478463992532e-05,
+      "loss": 1.005,
+      "step": 13045
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1968373813788254e-05,
+      "loss": 1.0194,
+      "step": 13046
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1967269138596164e-05,
+      "loss": 0.9877,
+      "step": 13047
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1966164438430282e-05,
+      "loss": 1.0392,
+      "step": 13048
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1965059713304622e-05,
+      "loss": 0.966,
+      "step": 13049
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196395496323322e-05,
+      "loss": 1.0439,
+      "step": 13050
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1962850188230094e-05,
+      "loss": 1.0389,
+      "step": 13051
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1961745388309276e-05,
+      "loss": 1.0384,
+      "step": 13052
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.196064056348478e-05,
+      "loss": 1.0079,
+      "step": 13053
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1959535713770643e-05,
+      "loss": 0.8506,
+      "step": 13054
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1958430839180883e-05,
+      "loss": 1.0569,
+      "step": 13055
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1957325939729524e-05,
+      "loss": 1.0023,
+      "step": 13056
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1956221015430602e-05,
+      "loss": 1.0831,
+      "step": 13057
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1955116066298135e-05,
+      "loss": 1.008,
+      "step": 13058
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1954011092346155e-05,
+      "loss": 0.9967,
+      "step": 13059
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1952906093588682e-05,
+      "loss": 1.0437,
+      "step": 13060
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1951801070039754e-05,
+      "loss": 0.954,
+      "step": 13061
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1950696021713395e-05,
+      "loss": 0.9292,
+      "step": 13062
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1949590948623627e-05,
+      "loss": 1.0051,
+      "step": 13063
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1948485850784487e-05,
+      "loss": 0.9357,
+      "step": 13064
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1947380728209997e-05,
+      "loss": 1.0604,
+      "step": 13065
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194627558091419e-05,
+      "loss": 0.9713,
+      "step": 13066
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1945170408911095e-05,
+      "loss": 0.9414,
+      "step": 13067
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1944065212214744e-05,
+      "loss": 0.9975,
+      "step": 13068
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1942959990839165e-05,
+      "loss": 1.0057,
+      "step": 13069
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1941854744798382e-05,
+      "loss": 0.9299,
+      "step": 13070
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.194074947410644e-05,
+      "loss": 0.9645,
+      "step": 13071
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1939644178777354e-05,
+      "loss": 0.945,
+      "step": 13072
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1938538858825165e-05,
+      "loss": 0.929,
+      "step": 13073
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1937433514263905e-05,
+      "loss": 1.0088,
+      "step": 13074
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1936328145107606e-05,
+      "loss": 1.0424,
+      "step": 13075
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1935222751370297e-05,
+      "loss": 1.0764,
+      "step": 13076
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1934117333066006e-05,
+      "loss": 1.0425,
+      "step": 13077
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1933011890208777e-05,
+      "loss": 1.0556,
+      "step": 13078
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 1.1931906422812636e-05,
+      "loss": 1.0433,
+      "step": 13079
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1930800930891617e-05,
+      "loss": 0.9425,
+      "step": 13080
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1929695414459757e-05,
+      "loss": 1.0208,
+      "step": 13081
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1928589873531088e-05,
+      "loss": 0.9801,
+      "step": 13082
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1927484308119646e-05,
+      "loss": 1.0516,
+      "step": 13083
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192637871823946e-05,
+      "loss": 1.0515,
+      "step": 13084
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1925273103904573e-05,
+      "loss": 0.9337,
+      "step": 13085
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1924167465129015e-05,
+      "loss": 1.028,
+      "step": 13086
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1923061801926824e-05,
+      "loss": 0.9892,
+      "step": 13087
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1921956114312036e-05,
+      "loss": 1.0226,
+      "step": 13088
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.192085040229869e-05,
+      "loss": 0.9798,
+      "step": 13089
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1919744665900814e-05,
+      "loss": 0.9451,
+      "step": 13090
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1918638905132454e-05,
+      "loss": 1.1219,
+      "step": 13091
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1917533120007642e-05,
+      "loss": 1.0496,
+      "step": 13092
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1916427310540419e-05,
+      "loss": 1.0393,
+      "step": 13093
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.191532147674482e-05,
+      "loss": 0.9609,
+      "step": 13094
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1914215618634883e-05,
+      "loss": 0.9738,
+      "step": 13095
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1913109736224654e-05,
+      "loss": 1.0512,
+      "step": 13096
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1912003829528158e-05,
+      "loss": 0.9868,
+      "step": 13097
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1910897898559445e-05,
+      "loss": 1.0112,
+      "step": 13098
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1909791943332551e-05,
+      "loss": 1.0561,
+      "step": 13099
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1908685963861517e-05,
+      "loss": 1.0116,
+      "step": 13100
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1907579960160381e-05,
+      "loss": 0.9802,
+      "step": 13101
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.190647393224318e-05,
+      "loss": 1.0675,
+      "step": 13102
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1905367880123966e-05,
+      "loss": 1.1713,
+      "step": 13103
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1904261803816768e-05,
+      "loss": 0.9541,
+      "step": 13104
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1903155703335634e-05,
+      "loss": 1.0853,
+      "step": 13105
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1902049578694603e-05,
+      "loss": 0.9764,
+      "step": 13106
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1900943429907716e-05,
+      "loss": 0.9771,
+      "step": 13107
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1899837256989017e-05,
+      "loss": 1.0739,
+      "step": 13108
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1898731059952551e-05,
+      "loss": 1.1135,
+      "step": 13109
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1897624838812355e-05,
+      "loss": 1.1253,
+      "step": 13110
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1896518593582478e-05,
+      "loss": 0.9406,
+      "step": 13111
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1895412324276956e-05,
+      "loss": 1.0921,
+      "step": 13112
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1894306030909842e-05,
+      "loss": 0.9199,
+      "step": 13113
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1893199713495171e-05,
+      "loss": 0.9098,
+      "step": 13114
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1892093372046991e-05,
+      "loss": 1.0181,
+      "step": 13115
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1890987006579348e-05,
+      "loss": 0.9716,
+      "step": 13116
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1889880617106289e-05,
+      "loss": 1.1377,
+      "step": 13117
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1888774203641853e-05,
+      "loss": 1.0618,
+      "step": 13118
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1887667766200088e-05,
+      "loss": 1.0197,
+      "step": 13119
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1886561304795046e-05,
+      "loss": 1.0124,
+      "step": 13120
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1885454819440763e-05,
+      "loss": 1.045,
+      "step": 13121
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1884348310151289e-05,
+      "loss": 0.943,
+      "step": 13122
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1883241776940673e-05,
+      "loss": 1.0126,
+      "step": 13123
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1882135219822963e-05,
+      "loss": 1.0369,
+      "step": 13124
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1881028638812201e-05,
+      "loss": 1.0418,
+      "step": 13125
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187992203392244e-05,
+      "loss": 1.1247,
+      "step": 13126
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1878815405167727e-05,
+      "loss": 1.0191,
+      "step": 13127
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1877708752562107e-05,
+      "loss": 0.9818,
+      "step": 13128
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187660207611963e-05,
+      "loss": 0.9592,
+      "step": 13129
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1875495375854346e-05,
+      "loss": 1.0074,
+      "step": 13130
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187438865178031e-05,
+      "loss": 1.1137,
+      "step": 13131
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1873281903911557e-05,
+      "loss": 0.9659,
+      "step": 13132
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1872175132262148e-05,
+      "loss": 0.9405,
+      "step": 13133
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.187106833684613e-05,
+      "loss": 1.15,
+      "step": 13134
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1869961517677553e-05,
+      "loss": 0.9733,
+      "step": 13135
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186885467477047e-05,
+      "loss": 1.0871,
+      "step": 13136
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1867747808138927e-05,
+      "loss": 0.9355,
+      "step": 13137
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186664091779698e-05,
+      "loss": 0.9542,
+      "step": 13138
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1865534003758678e-05,
+      "loss": 0.9335,
+      "step": 13139
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1864427066038078e-05,
+      "loss": 0.927,
+      "step": 13140
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1863320104649225e-05,
+      "loss": 1.1313,
+      "step": 13141
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1862213119606174e-05,
+      "loss": 1.0185,
+      "step": 13142
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.186110611092298e-05,
+      "loss": 0.9788,
+      "step": 13143
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1859999078613693e-05,
+      "loss": 1.1893,
+      "step": 13144
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1858892022692369e-05,
+      "loss": 0.9894,
+      "step": 13145
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1857784943173058e-05,
+      "loss": 1.1385,
+      "step": 13146
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.185667784006982e-05,
+      "loss": 0.9421,
+      "step": 13147
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1855570713396705e-05,
+      "loss": 1.0799,
+      "step": 13148
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1854463563167768e-05,
+      "loss": 1.1259,
+      "step": 13149
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1853356389397064e-05,
+      "loss": 1.0412,
+      "step": 13150
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1852249192098647e-05,
+      "loss": 0.9935,
+      "step": 13151
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1851141971286576e-05,
+      "loss": 1.0362,
+      "step": 13152
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1850034726974902e-05,
+      "loss": 1.0335,
+      "step": 13153
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1848927459177685e-05,
+      "loss": 1.0037,
+      "step": 13154
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1847820167908986e-05,
+      "loss": 0.984,
+      "step": 13155
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184671285318285e-05,
+      "loss": 1.0932,
+      "step": 13156
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184560551501334e-05,
+      "loss": 1.1187,
+      "step": 13157
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1844498153414512e-05,
+      "loss": 1.0168,
+      "step": 13158
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184339076840043e-05,
+      "loss": 1.0443,
+      "step": 13159
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1842283359985139e-05,
+      "loss": 1.0535,
+      "step": 13160
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.184117592818271e-05,
+      "loss": 0.9253,
+      "step": 13161
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1840068473007195e-05,
+      "loss": 0.9651,
+      "step": 13162
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1838960994472652e-05,
+      "loss": 1.1339,
+      "step": 13163
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1837853492593146e-05,
+      "loss": 0.9918,
+      "step": 13164
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1836745967382729e-05,
+      "loss": 0.8923,
+      "step": 13165
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1835638418855466e-05,
+      "loss": 1.0737,
+      "step": 13166
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1834530847025412e-05,
+      "loss": 1.0082,
+      "step": 13167
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1833423251906633e-05,
+      "loss": 1.0625,
+      "step": 13168
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1832315633513186e-05,
+      "loss": 1.0407,
+      "step": 13169
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.183120799185913e-05,
+      "loss": 0.9254,
+      "step": 13170
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1830100326958531e-05,
+      "loss": 0.9683,
+      "step": 13171
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1828992638825447e-05,
+      "loss": 0.9787,
+      "step": 13172
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1827884927473945e-05,
+      "loss": 1.0383,
+      "step": 13173
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1826777192918075e-05,
+      "loss": 1.0074,
+      "step": 13174
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1825669435171914e-05,
+      "loss": 0.9636,
+      "step": 13175
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1824561654249516e-05,
+      "loss": 0.958,
+      "step": 13176
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.182345385016494e-05,
+      "loss": 0.9719,
+      "step": 13177
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1822346022932263e-05,
+      "loss": 0.9678,
+      "step": 13178
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1821238172565535e-05,
+      "loss": 0.9397,
+      "step": 13179
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1820130299078827e-05,
+      "loss": 0.9514,
+      "step": 13180
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.18190224024862e-05,
+      "loss": 0.8776,
+      "step": 13181
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181791448280172e-05,
+      "loss": 0.8886,
+      "step": 13182
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1816806540039454e-05,
+      "loss": 0.9345,
+      "step": 13183
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1815698574213456e-05,
+      "loss": 0.9896,
+      "step": 13184
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1814590585337808e-05,
+      "loss": 1.0124,
+      "step": 13185
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1813482573426562e-05,
+      "loss": 1.1356,
+      "step": 13186
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.181237453849379e-05,
+      "loss": 1.05,
+      "step": 13187
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1811266480553557e-05,
+      "loss": 0.9962,
+      "step": 13188
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1810158399619929e-05,
+      "loss": 0.9469,
+      "step": 13189
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1809050295706973e-05,
+      "loss": 0.9805,
+      "step": 13190
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1807942168828753e-05,
+      "loss": 0.873,
+      "step": 13191
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1806834018999342e-05,
+      "loss": 1.0672,
+      "step": 13192
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1805725846232802e-05,
+      "loss": 1.1411,
+      "step": 13193
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1804617650543206e-05,
+      "loss": 1.0759,
+      "step": 13194
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180350943194462e-05,
+      "loss": 0.9723,
+      "step": 13195
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1802401190451111e-05,
+      "loss": 0.8655,
+      "step": 13196
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.180129292607675e-05,
+      "loss": 1.0544,
+      "step": 13197
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1800184638835604e-05,
+      "loss": 0.928,
+      "step": 13198
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1799076328741745e-05,
+      "loss": 0.968,
+      "step": 13199
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1797967995809239e-05,
+      "loss": 1.1815,
+      "step": 13200
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1796859640052156e-05,
+      "loss": 0.9686,
+      "step": 13201
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1795751261484569e-05,
+      "loss": 1.0191,
+      "step": 13202
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179464286012055e-05,
+      "loss": 0.9906,
+      "step": 13203
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1793534435974166e-05,
+      "loss": 0.949,
+      "step": 13204
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1792425989059489e-05,
+      "loss": 0.992,
+      "step": 13205
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1791317519390591e-05,
+      "loss": 1.1251,
+      "step": 13206
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.179020902698154e-05,
+      "loss": 1.0105,
+      "step": 13207
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1789100511846415e-05,
+      "loss": 0.9942,
+      "step": 13208
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1787991973999281e-05,
+      "loss": 0.9356,
+      "step": 13209
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.178688341345422e-05,
+      "loss": 0.9938,
+      "step": 13210
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1785774830225293e-05,
+      "loss": 1.031,
+      "step": 13211
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1784666224326579e-05,
+      "loss": 0.9113,
+      "step": 13212
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1783557595772154e-05,
+      "loss": 0.9718,
+      "step": 13213
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1782448944576088e-05,
+      "loss": 1.0148,
+      "step": 13214
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1781340270752453e-05,
+      "loss": 1.0512,
+      "step": 13215
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1780231574315329e-05,
+      "loss": 0.9139,
+      "step": 13216
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1779122855278787e-05,
+      "loss": 1.0956,
+      "step": 13217
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17780141136569e-05,
+      "loss": 1.0305,
+      "step": 13218
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1776905349463745e-05,
+      "loss": 0.9871,
+      "step": 13219
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1775796562713402e-05,
+      "loss": 1.0431,
+      "step": 13220
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.177468775341994e-05,
+      "loss": 0.97,
+      "step": 13221
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1773578921597436e-05,
+      "loss": 1.0527,
+      "step": 13222
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1772470067259967e-05,
+      "loss": 1.0193,
+      "step": 13223
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1771361190421615e-05,
+      "loss": 0.975,
+      "step": 13224
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1770252291096447e-05,
+      "loss": 1.0071,
+      "step": 13225
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1769143369298545e-05,
+      "loss": 0.99,
+      "step": 13226
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176803442504199e-05,
+      "loss": 1.0328,
+      "step": 13227
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1766925458340852e-05,
+      "loss": 1.0078,
+      "step": 13228
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1765816469209214e-05,
+      "loss": 1.0163,
+      "step": 13229
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1764707457661153e-05,
+      "loss": 1.1269,
+      "step": 13230
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176359842371075e-05,
+      "loss": 0.9371,
+      "step": 13231
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176248936737208e-05,
+      "loss": 0.9809,
+      "step": 13232
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1761380288659224e-05,
+      "loss": 1.0019,
+      "step": 13233
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.176027118758626e-05,
+      "loss": 1.1034,
+      "step": 13234
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1759162064167269e-05,
+      "loss": 0.9785,
+      "step": 13235
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.175805291841633e-05,
+      "loss": 0.9037,
+      "step": 13236
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1756943750347523e-05,
+      "loss": 0.9505,
+      "step": 13237
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1755834559974933e-05,
+      "loss": 1.028,
+      "step": 13238
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1754725347312632e-05,
+      "loss": 0.9809,
+      "step": 13239
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1753616112374708e-05,
+      "loss": 1.0075,
+      "step": 13240
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1752506855175242e-05,
+      "loss": 0.9285,
+      "step": 13241
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1751397575728312e-05,
+      "loss": 0.9914,
+      "step": 13242
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1750288274048e-05,
+      "loss": 0.9373,
+      "step": 13243
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1749178950148394e-05,
+      "loss": 1.0729,
+      "step": 13244
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1748069604043573e-05,
+      "loss": 1.0602,
+      "step": 13245
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1746960235747618e-05,
+      "loss": 1.1819,
+      "step": 13246
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.174585084527461e-05,
+      "loss": 0.9738,
+      "step": 13247
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1744741432638642e-05,
+      "loss": 1.0161,
+      "step": 13248
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1743631997853788e-05,
+      "loss": 0.9442,
+      "step": 13249
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1742522540934136e-05,
+      "loss": 1.0537,
+      "step": 13250
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1741413061893768e-05,
+      "loss": 0.961,
+      "step": 13251
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1740303560746772e-05,
+      "loss": 0.9046,
+      "step": 13252
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1739194037507227e-05,
+      "loss": 0.9254,
+      "step": 13253
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1738084492189226e-05,
+      "loss": 1.0027,
+      "step": 13254
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173697492480685e-05,
+      "loss": 1.0859,
+      "step": 13255
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1735865335374181e-05,
+      "loss": 0.9923,
+      "step": 13256
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1734755723905309e-05,
+      "loss": 1.0334,
+      "step": 13257
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.173364609041432e-05,
+      "loss": 0.9606,
+      "step": 13258
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1732536434915301e-05,
+      "loss": 1.0489,
+      "step": 13259
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1731426757422334e-05,
+      "loss": 0.9991,
+      "step": 13260
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1730317057949513e-05,
+      "loss": 1.1119,
+      "step": 13261
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1729207336510923e-05,
+      "loss": 1.0009,
+      "step": 13262
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.172809759312065e-05,
+      "loss": 0.9519,
+      "step": 13263
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1726987827792778e-05,
+      "loss": 0.9826,
+      "step": 13264
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1725878040541401e-05,
+      "loss": 0.945,
+      "step": 13265
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1724768231380606e-05,
+      "loss": 0.9142,
+      "step": 13266
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1723658400324479e-05,
+      "loss": 1.1172,
+      "step": 13267
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1722548547387115e-05,
+      "loss": 1.1206,
+      "step": 13268
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1721438672582598e-05,
+      "loss": 0.8937,
+      "step": 13269
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1720328775925016e-05,
+      "loss": 0.9784,
+      "step": 13270
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1719218857428464e-05,
+      "loss": 0.9754,
+      "step": 13271
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171810891710703e-05,
+      "loss": 0.9097,
+      "step": 13272
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1716998954974805e-05,
+      "loss": 1.0124,
+      "step": 13273
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1715888971045875e-05,
+      "loss": 1.0108,
+      "step": 13274
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1714778965334337e-05,
+      "loss": 1.0097,
+      "step": 13275
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.171366893785428e-05,
+      "loss": 1.0344,
+      "step": 13276
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1712558888619789e-05,
+      "loss": 0.9766,
+      "step": 13277
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1711448817644969e-05,
+      "loss": 0.953,
+      "step": 13278
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.17103387249439e-05,
+      "loss": 0.8996,
+      "step": 13279
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170922861053068e-05,
+      "loss": 1.0262,
+      "step": 13280
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1708118474419396e-05,
+      "loss": 0.9017,
+      "step": 13281
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1707008316624151e-05,
+      "loss": 1.049,
+      "step": 13282
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1705898137159032e-05,
+      "loss": 1.1265,
+      "step": 13283
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1704787936038129e-05,
+      "loss": 1.0553,
+      "step": 13284
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170367771327554e-05,
+      "loss": 1.027,
+      "step": 13285
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.170256746888536e-05,
+      "loss": 0.9708,
+      "step": 13286
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1701457202881681e-05,
+      "loss": 1.0714,
+      "step": 13287
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1700346915278593e-05,
+      "loss": 1.068,
+      "step": 13288
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.16992366060902e-05,
+      "loss": 1.0637,
+      "step": 13289
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1698126275330592e-05,
+      "loss": 0.9308,
+      "step": 13290
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1697015923013862e-05,
+      "loss": 1.0093,
+      "step": 13291
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1695905549154109e-05,
+      "loss": 0.9942,
+      "step": 13292
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1694795153765428e-05,
+      "loss": 1.0573,
+      "step": 13293
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1693684736861914e-05,
+      "loss": 1.0874,
+      "step": 13294
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1692574298457664e-05,
+      "loss": 1.0361,
+      "step": 13295
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1691463838566777e-05,
+      "loss": 0.9574,
+      "step": 13296
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1690353357203347e-05,
+      "loss": 0.9468,
+      "step": 13297
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1689242854381471e-05,
+      "loss": 0.8959,
+      "step": 13298
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168813233011525e-05,
+      "loss": 1.0003,
+      "step": 13299
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1687021784418777e-05,
+      "loss": 0.9881,
+      "step": 13300
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1685911217306151e-05,
+      "loss": 1.0968,
+      "step": 13301
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1684800628791473e-05,
+      "loss": 1.0537,
+      "step": 13302
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.168369001888884e-05,
+      "loss": 0.9754,
+      "step": 13303
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1682579387612354e-05,
+      "loss": 1.1528,
+      "step": 13304
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1681468734976103e-05,
+      "loss": 1.013,
+      "step": 13305
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1680358060994201e-05,
+      "loss": 0.9716,
+      "step": 13306
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1679247365680737e-05,
+      "loss": 0.9605,
+      "step": 13307
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1678136649049817e-05,
+      "loss": 0.8873,
+      "step": 13308
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1677025911115535e-05,
+      "loss": 1.0385,
+      "step": 13309
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1675915151891998e-05,
+      "loss": 0.9326,
+      "step": 13310
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1674804371393305e-05,
+      "loss": 1.0147,
+      "step": 13311
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1673693569633552e-05,
+      "loss": 1.0895,
+      "step": 13312
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167258274662685e-05,
+      "loss": 1.06,
+      "step": 13313
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1671471902387293e-05,
+      "loss": 1.0645,
+      "step": 13314
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.167036103692898e-05,
+      "loss": 0.9631,
+      "step": 13315
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.166925015026602e-05,
+      "loss": 1.0945,
+      "step": 13316
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1668139242412514e-05,
+      "loss": 1.0648,
+      "step": 13317
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1667028313382565e-05,
+      "loss": 0.9406,
+      "step": 13318
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1665917363190268e-05,
+      "loss": 0.9655,
+      "step": 13319
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1664806391849739e-05,
+      "loss": 1.1047,
+      "step": 13320
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1663695399375072e-05,
+      "loss": 1.0166,
+      "step": 13321
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1662584385780374e-05,
+      "loss": 1.0568,
+      "step": 13322
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1661473351079747e-05,
+      "loss": 0.9894,
+      "step": 13323
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1660362295287297e-05,
+      "loss": 0.9413,
+      "step": 13324
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1659251218417128e-05,
+      "loss": 1.0637,
+      "step": 13325
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1658140120483346e-05,
+      "loss": 0.9296,
+      "step": 13326
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1657029001500056e-05,
+      "loss": 1.1235,
+      "step": 13327
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1655917861481359e-05,
+      "loss": 0.8413,
+      "step": 13328
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1654806700441363e-05,
+      "loss": 0.9933,
+      "step": 13329
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1653695518394175e-05,
+      "loss": 0.8966,
+      "step": 13330
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1652584315353904e-05,
+      "loss": 0.8543,
+      "step": 13331
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165147309133465e-05,
+      "loss": 1.0085,
+      "step": 13332
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.165036184635052e-05,
+      "loss": 0.9916,
+      "step": 13333
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1649250580415628e-05,
+      "loss": 0.998,
+      "step": 13334
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1648139293544073e-05,
+      "loss": 0.9556,
+      "step": 13335
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1647027985749965e-05,
+      "loss": 1.0606,
+      "step": 13336
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1645916657047413e-05,
+      "loss": 1.0741,
+      "step": 13337
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1644805307450528e-05,
+      "loss": 0.9976,
+      "step": 13338
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1643693936973412e-05,
+      "loss": 0.9914,
+      "step": 13339
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1642582545630176e-05,
+      "loss": 1.0724,
+      "step": 13340
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1641471133434929e-05,
+      "loss": 1.0955,
+      "step": 13341
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.164035970040178e-05,
+      "loss": 1.0598,
+      "step": 13342
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1639248246544836e-05,
+      "loss": 1.1008,
+      "step": 13343
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163813677187821e-05,
+      "loss": 1.0044,
+      "step": 13344
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1637025276416011e-05,
+      "loss": 0.9716,
+      "step": 13345
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1635913760172348e-05,
+      "loss": 0.9989,
+      "step": 13346
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163480222316133e-05,
+      "loss": 0.9795,
+      "step": 13347
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163369066539707e-05,
+      "loss": 1.0045,
+      "step": 13348
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.163257908689368e-05,
+      "loss": 0.9616,
+      "step": 13349
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1631467487665267e-05,
+      "loss": 0.9968,
+      "step": 13350
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1630355867725945e-05,
+      "loss": 0.9657,
+      "step": 13351
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1629244227089824e-05,
+      "loss": 0.9995,
+      "step": 13352
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.162813256577102e-05,
+      "loss": 1.0894,
+      "step": 13353
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1627020883783639e-05,
+      "loss": 0.9814,
+      "step": 13354
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1625909181141799e-05,
+      "loss": 1.0678,
+      "step": 13355
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1624797457859608e-05,
+      "loss": 1.0362,
+      "step": 13356
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1623685713951182e-05,
+      "loss": 1.1298,
+      "step": 13357
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1622573949430634e-05,
+      "loss": 0.9351,
+      "step": 13358
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1621462164312079e-05,
+      "loss": 0.8972,
+      "step": 13359
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1620350358609624e-05,
+      "loss": 0.9488,
+      "step": 13360
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161923853233739e-05,
+      "loss": 0.985,
+      "step": 13361
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161812668550949e-05,
+      "loss": 1.0123,
+      "step": 13362
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1617014818140036e-05,
+      "loss": 1.0361,
+      "step": 13363
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1615902930243142e-05,
+      "loss": 1.0561,
+      "step": 13364
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1614791021832928e-05,
+      "loss": 0.8727,
+      "step": 13365
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.1613679092923507e-05,
+      "loss": 0.9181,
+      "step": 13366
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.161256714352899e-05,
+      "loss": 0.9653,
+      "step": 13367
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.16114551736635e-05,
+      "loss": 0.9921,
+      "step": 13368
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1610343183341151e-05,
+      "loss": 1.0301,
+      "step": 13369
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1609231172576055e-05,
+      "loss": 1.0018,
+      "step": 13370
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1608119141382332e-05,
+      "loss": 1.079,
+      "step": 13371
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1607007089774098e-05,
+      "loss": 0.9941,
+      "step": 13372
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1605895017765474e-05,
+      "loss": 0.983,
+      "step": 13373
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1604782925370573e-05,
+      "loss": 1.1405,
+      "step": 13374
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1603670812603511e-05,
+      "loss": 0.8675,
+      "step": 13375
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1602558679478412e-05,
+      "loss": 0.9708,
+      "step": 13376
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.160144652600939e-05,
+      "loss": 1.0091,
+      "step": 13377
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1600334352210563e-05,
+      "loss": 0.9974,
+      "step": 13378
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1599222158096051e-05,
+      "loss": 1.1442,
+      "step": 13379
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1598109943679974e-05,
+      "loss": 0.9672,
+      "step": 13380
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1596997708976448e-05,
+      "loss": 0.9875,
+      "step": 13381
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1595885453999593e-05,
+      "loss": 1.0577,
+      "step": 13382
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1594773178763534e-05,
+      "loss": 1.1637,
+      "step": 13383
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1593660883282382e-05,
+      "loss": 0.9188,
+      "step": 13384
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1592548567570268e-05,
+      "loss": 1.1351,
+      "step": 13385
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1591436231641303e-05,
+      "loss": 1.0366,
+      "step": 13386
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1590323875509613e-05,
+      "loss": 1.0448,
+      "step": 13387
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1589211499189317e-05,
+      "loss": 1.001,
+      "step": 13388
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1588099102694534e-05,
+      "loss": 0.8835,
+      "step": 13389
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1586986686039392e-05,
+      "loss": 1.0866,
+      "step": 13390
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1585874249238002e-05,
+      "loss": 1.0156,
+      "step": 13391
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.15847617923045e-05,
+      "loss": 1.1052,
+      "step": 13392
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1583649315252998e-05,
+      "loss": 1.0223,
+      "step": 13393
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1582536818097622e-05,
+      "loss": 0.9966,
+      "step": 13394
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1581424300852492e-05,
+      "loss": 1.0685,
+      "step": 13395
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1580311763531738e-05,
+      "loss": 1.1838,
+      "step": 13396
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1579199206149476e-05,
+      "loss": 0.9915,
+      "step": 13397
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.157808662871983e-05,
+      "loss": 1.0991,
+      "step": 13398
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1576974031256929e-05,
+      "loss": 0.9316,
+      "step": 13399
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1575861413774892e-05,
+      "loss": 0.9848,
+      "step": 13400
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1574748776287845e-05,
+      "loss": 0.9739,
+      "step": 13401
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1573636118809912e-05,
+      "loss": 0.9437,
+      "step": 13402
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.157252344135522e-05,
+      "loss": 0.9915,
+      "step": 13403
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1571410743937897e-05,
+      "loss": 0.9425,
+      "step": 13404
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1570298026572055e-05,
+      "loss": 1.1141,
+      "step": 13405
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1569185289271836e-05,
+      "loss": 1.0558,
+      "step": 13406
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1568072532051352e-05,
+      "loss": 0.9874,
+      "step": 13407
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1566959754924739e-05,
+      "loss": 1.1595,
+      "step": 13408
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1565846957906119e-05,
+      "loss": 1.0414,
+      "step": 13409
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.156473414100962e-05,
+      "loss": 1.0297,
+      "step": 13410
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1563621304249369e-05,
+      "loss": 1.0,
+      "step": 13411
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1562508447639486e-05,
+      "loss": 1.0412,
+      "step": 13412
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1561395571194112e-05,
+      "loss": 0.9313,
+      "step": 13413
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1560282674927363e-05,
+      "loss": 1.0352,
+      "step": 13414
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155916975885337e-05,
+      "loss": 1.1569,
+      "step": 13415
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1558056822986263e-05,
+      "loss": 0.9702,
+      "step": 13416
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1556943867340171e-05,
+      "loss": 0.9262,
+      "step": 13417
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1555830891929221e-05,
+      "loss": 1.028,
+      "step": 13418
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1554717896767538e-05,
+      "loss": 1.1321,
+      "step": 13419
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1553604881869257e-05,
+      "loss": 1.0208,
+      "step": 13420
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1552491847248506e-05,
+      "loss": 1.005,
+      "step": 13421
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.155137879291941e-05,
+      "loss": 1.0852,
+      "step": 13422
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1550265718896105e-05,
+      "loss": 1.073,
+      "step": 13423
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1549152625192721e-05,
+      "loss": 0.9969,
+      "step": 13424
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154803951182338e-05,
+      "loss": 0.9589,
+      "step": 13425
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1546926378802223e-05,
+      "loss": 1.0965,
+      "step": 13426
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1545813226143378e-05,
+      "loss": 1.0409,
+      "step": 13427
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1544700053860972e-05,
+      "loss": 1.0549,
+      "step": 13428
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1543586861969136e-05,
+      "loss": 1.0565,
+      "step": 13429
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1542473650482007e-05,
+      "loss": 1.0846,
+      "step": 13430
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1541360419413717e-05,
+      "loss": 0.9111,
+      "step": 13431
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.154024716877839e-05,
+      "loss": 1.0857,
+      "step": 13432
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1539133898590165e-05,
+      "loss": 0.889,
+      "step": 13433
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1538020608863176e-05,
+      "loss": 0.9297,
+      "step": 13434
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1536907299611551e-05,
+      "loss": 1.0268,
+      "step": 13435
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1535793970849423e-05,
+      "loss": 0.8794,
+      "step": 13436
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1534680622590928e-05,
+      "loss": 1.097,
+      "step": 13437
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1533567254850202e-05,
+      "loss": 0.9997,
+      "step": 13438
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1532453867641374e-05,
+      "loss": 1.1221,
+      "step": 13439
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.153134046097858e-05,
+      "loss": 0.8787,
+      "step": 13440
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1530227034875954e-05,
+      "loss": 1.1361,
+      "step": 13441
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1529113589347629e-05,
+      "loss": 1.0728,
+      "step": 13442
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1528000124407743e-05,
+      "loss": 0.9777,
+      "step": 13443
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152688664007043e-05,
+      "loss": 0.8001,
+      "step": 13444
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1525773136349824e-05,
+      "loss": 0.9153,
+      "step": 13445
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1524659613260063e-05,
+      "loss": 1.0206,
+      "step": 13446
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152354607081528e-05,
+      "loss": 0.9908,
+      "step": 13447
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.152243250902961e-05,
+      "loss": 1.0633,
+      "step": 13448
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1521318927917196e-05,
+      "loss": 0.9531,
+      "step": 13449
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1520205327492164e-05,
+      "loss": 0.9146,
+      "step": 13450
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1519091707768662e-05,
+      "loss": 1.0317,
+      "step": 13451
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.151797806876082e-05,
+      "loss": 0.9008,
+      "step": 13452
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1516864410482776e-05,
+      "loss": 1.0686,
+      "step": 13453
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1515750732948669e-05,
+      "loss": 0.9565,
+      "step": 13454
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1514637036172638e-05,
+      "loss": 0.907,
+      "step": 13455
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1513523320168819e-05,
+      "loss": 1.0562,
+      "step": 13456
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1512409584951348e-05,
+      "loss": 1.0603,
+      "step": 13457
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1511295830534366e-05,
+      "loss": 1.0477,
+      "step": 13458
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1510182056932013e-05,
+      "loss": 1.0662,
+      "step": 13459
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1509068264158428e-05,
+      "loss": 1.0651,
+      "step": 13460
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1507954452227746e-05,
+      "loss": 1.0008,
+      "step": 13461
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1506840621154111e-05,
+      "loss": 0.8397,
+      "step": 13462
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1505726770951662e-05,
+      "loss": 0.8769,
+      "step": 13463
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1504612901634535e-05,
+      "loss": 0.887,
+      "step": 13464
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1503499013216875e-05,
+      "loss": 1.2564,
+      "step": 13465
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1502385105712821e-05,
+      "loss": 1.1111,
+      "step": 13466
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1501271179136513e-05,
+      "loss": 0.9981,
+      "step": 13467
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1500157233502092e-05,
+      "loss": 0.9177,
+      "step": 13468
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1499043268823699e-05,
+      "loss": 0.9663,
+      "step": 13469
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1497929285115474e-05,
+      "loss": 1.0364,
+      "step": 13470
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1496815282391563e-05,
+      "loss": 0.9615,
+      "step": 13471
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1495701260666104e-05,
+      "loss": 1.0602,
+      "step": 13472
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1494587219953241e-05,
+      "loss": 0.84,
+      "step": 13473
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1493473160267115e-05,
+      "loss": 1.0798,
+      "step": 13474
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1492359081621868e-05,
+      "loss": 0.8396,
+      "step": 13475
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.149124498403165e-05,
+      "loss": 1.0276,
+      "step": 13476
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1490130867510591e-05,
+      "loss": 1.0071,
+      "step": 13477
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1489016732072845e-05,
+      "loss": 1.0585,
+      "step": 13478
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1487902577732548e-05,
+      "loss": 0.9675,
+      "step": 13479
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1486788404503854e-05,
+      "loss": 1.0272,
+      "step": 13480
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1485674212400897e-05,
+      "loss": 0.9891,
+      "step": 13481
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1484560001437826e-05,
+      "loss": 1.0505,
+      "step": 13482
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1483445771628787e-05,
+      "loss": 1.1142,
+      "step": 13483
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.148233152298792e-05,
+      "loss": 0.9728,
+      "step": 13484
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1481217255529372e-05,
+      "loss": 0.9335,
+      "step": 13485
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1480102969267289e-05,
+      "loss": 1.0469,
+      "step": 13486
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1478988664215817e-05,
+      "loss": 1.0897,
+      "step": 13487
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1477874340389098e-05,
+      "loss": 0.9821,
+      "step": 13488
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1476759997801283e-05,
+      "loss": 0.8881,
+      "step": 13489
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1475645636466518e-05,
+      "loss": 0.8988,
+      "step": 13490
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1474531256398945e-05,
+      "loss": 0.9975,
+      "step": 13491
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1473416857612714e-05,
+      "loss": 0.9796,
+      "step": 13492
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1472302440121969e-05,
+      "loss": 0.9336,
+      "step": 13493
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1471188003940865e-05,
+      "loss": 0.9428,
+      "step": 13494
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1470073549083538e-05,
+      "loss": 0.8641,
+      "step": 13495
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.146895907556414e-05,
+      "loss": 1.1039,
+      "step": 13496
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1467844583396825e-05,
+      "loss": 1.0582,
+      "step": 13497
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1466730072595733e-05,
+      "loss": 1.0006,
+      "step": 13498
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1465615543175017e-05,
+      "loss": 0.9812,
+      "step": 13499
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1464500995148822e-05,
+      "loss": 1.0436,
+      "step": 13500
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1463386428531303e-05,
+      "loss": 1.0273,
+      "step": 13501
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1462271843336598e-05,
+      "loss": 1.0569,
+      "step": 13502
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1461157239578868e-05,
+      "loss": 1.0387,
+      "step": 13503
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1460042617272259e-05,
+      "loss": 0.9469,
+      "step": 13504
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1458927976430911e-05,
+      "loss": 0.8619,
+      "step": 13505
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.145781331706899e-05,
+      "loss": 1.1689,
+      "step": 13506
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1456698639200636e-05,
+      "loss": 1.08,
+      "step": 13507
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1455583942840005e-05,
+      "loss": 0.9205,
+      "step": 13508
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1454469228001241e-05,
+      "loss": 0.937,
+      "step": 13509
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1453354494698499e-05,
+      "loss": 1.0468,
+      "step": 13510
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1452239742945932e-05,
+      "loss": 1.0753,
+      "step": 13511
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1451124972757684e-05,
+      "loss": 1.0736,
+      "step": 13512
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1450010184147915e-05,
+      "loss": 0.9836,
+      "step": 13513
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1448895377130774e-05,
+      "loss": 1.0365,
+      "step": 13514
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1447780551720412e-05,
+      "loss": 1.0608,
+      "step": 13515
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1446665707930982e-05,
+      "loss": 0.9673,
+      "step": 13516
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1445550845776636e-05,
+      "loss": 0.9719,
+      "step": 13517
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1444435965271529e-05,
+      "loss": 0.8874,
+      "step": 13518
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.144332106642981e-05,
+      "loss": 1.0236,
+      "step": 13519
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1442206149265635e-05,
+      "loss": 0.9624,
+      "step": 13520
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1441091213793159e-05,
+      "loss": 0.9778,
+      "step": 13521
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1439976260026531e-05,
+      "loss": 0.9953,
+      "step": 13522
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1438861287979913e-05,
+      "loss": 0.9416,
+      "step": 13523
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1437746297667447e-05,
+      "loss": 0.9814,
+      "step": 13524
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.14366312891033e-05,
+      "loss": 1.0145,
+      "step": 13525
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1435516262301616e-05,
+      "loss": 1.0194,
+      "step": 13526
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1434401217276558e-05,
+      "loss": 1.1604,
+      "step": 13527
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1433286154042275e-05,
+      "loss": 1.0528,
+      "step": 13528
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1432171072612927e-05,
+      "loss": 1.0182,
+      "step": 13529
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1431055973002669e-05,
+      "loss": 0.9317,
+      "step": 13530
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1429940855225656e-05,
+      "loss": 1.0948,
+      "step": 13531
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1428825719296042e-05,
+      "loss": 1.0984,
+      "step": 13532
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1427710565227982e-05,
+      "loss": 1.0721,
+      "step": 13533
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1426595393035642e-05,
+      "loss": 0.9525,
+      "step": 13534
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1425480202733168e-05,
+      "loss": 1.013,
+      "step": 13535
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1424364994334723e-05,
+      "loss": 0.9561,
+      "step": 13536
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1423249767854459e-05,
+      "loss": 0.9751,
+      "step": 13537
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142213452330654e-05,
+      "loss": 1.0077,
+      "step": 13538
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.142101926070512e-05,
+      "loss": 1.1045,
+      "step": 13539
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1419903980064354e-05,
+      "loss": 0.9463,
+      "step": 13540
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.141878868139841e-05,
+      "loss": 1.0784,
+      "step": 13541
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1417673364721434e-05,
+      "loss": 0.9841,
+      "step": 13542
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1416558030047591e-05,
+      "loss": 1.0038,
+      "step": 13543
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1415442677391038e-05,
+      "loss": 1.0231,
+      "step": 13544
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1414327306765937e-05,
+      "loss": 0.999,
+      "step": 13545
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1413211918186443e-05,
+      "loss": 1.043,
+      "step": 13546
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1412096511666721e-05,
+      "loss": 1.0669,
+      "step": 13547
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1410981087220927e-05,
+      "loss": 1.0583,
+      "step": 13548
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1409865644863218e-05,
+      "loss": 1.0482,
+      "step": 13549
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140875018460776e-05,
+      "loss": 0.9833,
+      "step": 13550
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140763470646871e-05,
+      "loss": 1.0211,
+      "step": 13551
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140651921046023e-05,
+      "loss": 0.9793,
+      "step": 13552
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140540369659648e-05,
+      "loss": 1.0409,
+      "step": 13553
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.140428816489162e-05,
+      "loss": 1.0658,
+      "step": 13554
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1403172615359815e-05,
+      "loss": 1.0633,
+      "step": 13555
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1402057048015223e-05,
+      "loss": 0.8912,
+      "step": 13556
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1400941462872006e-05,
+      "loss": 0.9003,
+      "step": 13557
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1399825859944327e-05,
+      "loss": 0.978,
+      "step": 13558
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1398710239246349e-05,
+      "loss": 0.95,
+      "step": 13559
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1397594600792232e-05,
+      "loss": 1.1507,
+      "step": 13560
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1396478944596142e-05,
+      "loss": 0.9579,
+      "step": 13561
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139536327067224e-05,
+      "loss": 1.0627,
+      "step": 13562
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1394247579034684e-05,
+      "loss": 0.9521,
+      "step": 13563
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1393131869697647e-05,
+      "loss": 1.0234,
+      "step": 13564
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139201614267529e-05,
+      "loss": 1.0419,
+      "step": 13565
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.139090039798177e-05,
+      "loss": 0.9488,
+      "step": 13566
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1389784635631257e-05,
+      "loss": 1.0275,
+      "step": 13567
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1388668855637915e-05,
+      "loss": 0.9498,
+      "step": 13568
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1387553058015906e-05,
+      "loss": 0.9443,
+      "step": 13569
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1386437242779395e-05,
+      "loss": 0.9638,
+      "step": 13570
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1385321409942549e-05,
+      "loss": 0.9477,
+      "step": 13571
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1384205559519532e-05,
+      "loss": 0.9882,
+      "step": 13572
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1383089691524511e-05,
+      "loss": 1.0773,
+      "step": 13573
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138197380597165e-05,
+      "loss": 0.9011,
+      "step": 13574
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.138085790287511e-05,
+      "loss": 1.0225,
+      "step": 13575
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1379741982249066e-05,
+      "loss": 0.9335,
+      "step": 13576
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1378626044107676e-05,
+      "loss": 1.0077,
+      "step": 13577
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1377510088465113e-05,
+      "loss": 1.0779,
+      "step": 13578
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137639411533554e-05,
+      "loss": 1.1167,
+      "step": 13579
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1375278124733127e-05,
+      "loss": 0.9805,
+      "step": 13580
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1374162116672036e-05,
+      "loss": 1.0673,
+      "step": 13581
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1373046091166438e-05,
+      "loss": 1.0554,
+      "step": 13582
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1371930048230499e-05,
+      "loss": 0.871,
+      "step": 13583
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.137081398787839e-05,
+      "loss": 1.0704,
+      "step": 13584
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1369697910124273e-05,
+      "loss": 1.0769,
+      "step": 13585
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1368581814982323e-05,
+      "loss": 0.9499,
+      "step": 13586
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1367465702466705e-05,
+      "loss": 1.0642,
+      "step": 13587
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1366349572591586e-05,
+      "loss": 1.0595,
+      "step": 13588
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1365233425371138e-05,
+      "loss": 0.8978,
+      "step": 13589
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1364117260819532e-05,
+      "loss": 1.1548,
+      "step": 13590
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1363001078950929e-05,
+      "loss": 1.0197,
+      "step": 13591
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1361884879779505e-05,
+      "loss": 1.1155,
+      "step": 13592
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1360768663319428e-05,
+      "loss": 0.8861,
+      "step": 13593
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1359652429584871e-05,
+      "loss": 1.0526,
+      "step": 13594
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135853617859e-05,
+      "loss": 1.0659,
+      "step": 13595
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1357419910348986e-05,
+      "loss": 0.947,
+      "step": 13596
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1356303624876004e-05,
+      "loss": 1.0355,
+      "step": 13597
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.135518732218522e-05,
+      "loss": 0.9335,
+      "step": 13598
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1354071002290805e-05,
+      "loss": 0.9892,
+      "step": 13599
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1352954665206932e-05,
+      "loss": 0.854,
+      "step": 13600
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1351838310947776e-05,
+      "loss": 1.0971,
+      "step": 13601
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13507219395275e-05,
+      "loss": 1.0509,
+      "step": 13602
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1349605550960283e-05,
+      "loss": 0.9701,
+      "step": 13603
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1348489145260299e-05,
+      "loss": 1.0852,
+      "step": 13604
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1347372722441712e-05,
+      "loss": 1.0906,
+      "step": 13605
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1346256282518702e-05,
+      "loss": 1.0103,
+      "step": 13606
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1345139825505438e-05,
+      "loss": 0.9756,
+      "step": 13607
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1344023351416094e-05,
+      "loss": 1.0351,
+      "step": 13608
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1342906860264845e-05,
+      "loss": 0.9372,
+      "step": 13609
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1341790352065858e-05,
+      "loss": 1.1432,
+      "step": 13610
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1340673826833318e-05,
+      "loss": 1.0152,
+      "step": 13611
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1339557284581386e-05,
+      "loss": 1.1061,
+      "step": 13612
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1338440725324248e-05,
+      "loss": 0.9234,
+      "step": 13613
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1337324149076067e-05,
+      "loss": 1.0213,
+      "step": 13614
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.133620755585103e-05,
+      "loss": 1.0738,
+      "step": 13615
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13350909456633e-05,
+      "loss": 0.9413,
+      "step": 13616
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1333974318527059e-05,
+      "loss": 1.0366,
+      "step": 13617
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1332857674456483e-05,
+      "loss": 0.9616,
+      "step": 13618
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1331741013465738e-05,
+      "loss": 0.8711,
+      "step": 13619
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1330624335569014e-05,
+      "loss": 1.0704,
+      "step": 13620
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1329507640780473e-05,
+      "loss": 0.9284,
+      "step": 13621
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.13283909291143e-05,
+      "loss": 0.9499,
+      "step": 13622
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.132727420058467e-05,
+      "loss": 1.047,
+      "step": 13623
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1326157455205756e-05,
+      "loss": 1.0041,
+      "step": 13624
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1325040692991737e-05,
+      "loss": 1.0076,
+      "step": 13625
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1323923913956786e-05,
+      "loss": 1.0859,
+      "step": 13626
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1322807118115089e-05,
+      "loss": 1.0438,
+      "step": 13627
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1321690305480818e-05,
+      "loss": 0.9434,
+      "step": 13628
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1320573476068147e-05,
+      "loss": 0.9172,
+      "step": 13629
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1319456629891259e-05,
+      "loss": 0.9617,
+      "step": 13630
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131833976696433e-05,
+      "loss": 0.9069,
+      "step": 13631
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.131722288730154e-05,
+      "loss": 0.9794,
+      "step": 13632
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1316105990917064e-05,
+      "loss": 1.0937,
+      "step": 13633
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1314989077825083e-05,
+      "loss": 0.9445,
+      "step": 13634
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1313872148039775e-05,
+      "loss": 1.0524,
+      "step": 13635
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1312755201575318e-05,
+      "loss": 1.0509,
+      "step": 13636
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1311638238445896e-05,
+      "loss": 0.9975,
+      "step": 13637
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1310521258665683e-05,
+      "loss": 1.1197,
+      "step": 13638
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1309404262248863e-05,
+      "loss": 0.9757,
+      "step": 13639
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1308287249209608e-05,
+      "loss": 1.0892,
+      "step": 13640
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130717021956211e-05,
+      "loss": 0.9596,
+      "step": 13641
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1306053173320539e-05,
+      "loss": 0.9969,
+      "step": 13642
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.130493611049908e-05,
+      "loss": 1.1826,
+      "step": 13643
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1303819031111915e-05,
+      "loss": 1.0434,
+      "step": 13644
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1302701935173221e-05,
+      "loss": 1.0359,
+      "step": 13645
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1301584822697182e-05,
+      "loss": 0.9755,
+      "step": 13646
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1300467693697979e-05,
+      "loss": 0.8148,
+      "step": 13647
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1299350548189793e-05,
+      "loss": 0.9934,
+      "step": 13648
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1298233386186805e-05,
+      "loss": 0.8586,
+      "step": 13649
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1297116207703198e-05,
+      "loss": 0.9863,
+      "step": 13650
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1295999012753159e-05,
+      "loss": 0.9619,
+      "step": 13651
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1294881801350858e-05,
+      "loss": 1.0367,
+      "step": 13652
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1293764573510491e-05,
+      "loss": 1.0161,
+      "step": 13653
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 1.1292647329246231e-05,
+      "loss": 0.9201,
+      "step": 13654
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1291530068572268e-05,
+      "loss": 1.1357,
+      "step": 13655
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.129041279150278e-05,
+      "loss": 0.9377,
+      "step": 13656
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1289295498051953e-05,
+      "loss": 1.1586,
+      "step": 13657
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1288178188233973e-05,
+      "loss": 1.1188,
+      "step": 13658
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1287060862063018e-05,
+      "loss": 1.0383,
+      "step": 13659
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1285943519553275e-05,
+      "loss": 0.9565,
+      "step": 13660
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1284826160718928e-05,
+      "loss": 0.9797,
+      "step": 13661
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1283708785574168e-05,
+      "loss": 0.9308,
+      "step": 13662
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1282591394133168e-05,
+      "loss": 1.0591,
+      "step": 13663
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1281473986410118e-05,
+      "loss": 1.0205,
+      "step": 13664
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1280356562419209e-05,
+      "loss": 1.1198,
+      "step": 13665
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1279239122174616e-05,
+      "loss": 0.9557,
+      "step": 13666
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.127812166569053e-05,
+      "loss": 0.9445,
+      "step": 13667
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1277004192981137e-05,
+      "loss": 0.9489,
+      "step": 13668
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1275886704060626e-05,
+      "loss": 0.8715,
+      "step": 13669
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1274769198943173e-05,
+      "loss": 1.0359,
+      "step": 13670
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1273651677642974e-05,
+      "loss": 1.0014,
+      "step": 13671
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1272534140174212e-05,
+      "loss": 0.9472,
+      "step": 13672
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1271416586551074e-05,
+      "loss": 0.9416,
+      "step": 13673
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1270299016787744e-05,
+      "loss": 0.9862,
+      "step": 13674
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1269181430898415e-05,
+      "loss": 1.0142,
+      "step": 13675
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.126806382889727e-05,
+      "loss": 1.22,
+      "step": 13676
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1266946210798499e-05,
+      "loss": 0.8888,
+      "step": 13677
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1265828576616286e-05,
+      "loss": 0.8559,
+      "step": 13678
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1264710926364826e-05,
+      "loss": 0.941,
+      "step": 13679
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1263593260058301e-05,
+      "loss": 0.9778,
+      "step": 13680
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1262475577710901e-05,
+      "loss": 0.9594,
+      "step": 13681
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1261357879336815e-05,
+      "loss": 0.9866,
+      "step": 13682
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1260240164950231e-05,
+      "loss": 1.0919,
+      "step": 13683
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125912243456534e-05,
+      "loss": 1.0233,
+      "step": 13684
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1258004688196327e-05,
+      "loss": 1.044,
+      "step": 13685
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1256886925857385e-05,
+      "loss": 0.9582,
+      "step": 13686
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1255769147562706e-05,
+      "loss": 0.9842,
+      "step": 13687
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1254651353326474e-05,
+      "loss": 1.0389,
+      "step": 13688
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125353354316288e-05,
+      "loss": 1.1379,
+      "step": 13689
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.125241571708612e-05,
+      "loss": 0.9658,
+      "step": 13690
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1251297875110377e-05,
+      "loss": 0.9442,
+      "step": 13691
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1250180017249846e-05,
+      "loss": 0.9949,
+      "step": 13692
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1249062143518716e-05,
+      "loss": 1.0887,
+      "step": 13693
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124794425393118e-05,
+      "loss": 1.002,
+      "step": 13694
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1246826348501426e-05,
+      "loss": 0.9767,
+      "step": 13695
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1245708427243649e-05,
+      "loss": 0.9792,
+      "step": 13696
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.124459049017204e-05,
+      "loss": 1.0702,
+      "step": 13697
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1243472537300788e-05,
+      "loss": 0.9647,
+      "step": 13698
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1242354568644085e-05,
+      "loss": 0.9663,
+      "step": 13699
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1241236584216127e-05,
+      "loss": 0.9835,
+      "step": 13700
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1240118584031106e-05,
+      "loss": 1.0047,
+      "step": 13701
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.123900056810321e-05,
+      "loss": 0.9171,
+      "step": 13702
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1237882536446635e-05,
+      "loss": 1.0117,
+      "step": 13703
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1236764489075577e-05,
+      "loss": 1.0176,
+      "step": 13704
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1235646426004223e-05,
+      "loss": 0.9619,
+      "step": 13705
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1234528347246771e-05,
+      "loss": 1.0126,
+      "step": 13706
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1233410252817412e-05,
+      "loss": 0.9875,
+      "step": 13707
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1232292142730343e-05,
+      "loss": 0.9038,
+      "step": 13708
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1231174016999755e-05,
+      "loss": 0.9444,
+      "step": 13709
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1230055875639843e-05,
+      "loss": 0.9222,
+      "step": 13710
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1228937718664802e-05,
+      "loss": 0.8565,
+      "step": 13711
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1227819546088826e-05,
+      "loss": 0.9817,
+      "step": 13712
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1226701357926108e-05,
+      "loss": 0.9726,
+      "step": 13713
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1225583154190849e-05,
+      "loss": 0.8644,
+      "step": 13714
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1224464934897239e-05,
+      "loss": 1.059,
+      "step": 13715
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1223346700059472e-05,
+      "loss": 0.9922,
+      "step": 13716
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1222228449691747e-05,
+      "loss": 1.149,
+      "step": 13717
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1221110183808264e-05,
+      "loss": 1.0655,
+      "step": 13718
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1219991902423206e-05,
+      "loss": 0.9889,
+      "step": 13719
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1218873605550784e-05,
+      "loss": 0.9772,
+      "step": 13720
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1217755293205184e-05,
+      "loss": 0.9671,
+      "step": 13721
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121663696540061e-05,
+      "loss": 1.0476,
+      "step": 13722
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1215518622151248e-05,
+      "loss": 1.0796,
+      "step": 13723
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1214400263471308e-05,
+      "loss": 0.8867,
+      "step": 13724
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121328188937498e-05,
+      "loss": 0.8532,
+      "step": 13725
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1212163499876458e-05,
+      "loss": 0.9108,
+      "step": 13726
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.121104509498995e-05,
+      "loss": 0.92,
+      "step": 13727
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1209926674729646e-05,
+      "loss": 0.9025,
+      "step": 13728
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1208808239109748e-05,
+      "loss": 0.9116,
+      "step": 13729
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1207689788144446e-05,
+      "loss": 1.0697,
+      "step": 13730
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1206571321847947e-05,
+      "loss": 0.9731,
+      "step": 13731
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1205452840234449e-05,
+      "loss": 0.8704,
+      "step": 13732
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1204334343318146e-05,
+      "loss": 0.9459,
+      "step": 13733
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1203215831113242e-05,
+      "loss": 0.9325,
+      "step": 13734
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1202097303633931e-05,
+      "loss": 1.0647,
+      "step": 13735
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1200978760894416e-05,
+      "loss": 1.0595,
+      "step": 13736
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1199860202908897e-05,
+      "loss": 0.9739,
+      "step": 13737
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.119874162969157e-05,
+      "loss": 1.0727,
+      "step": 13738
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1197623041256638e-05,
+      "loss": 1.067,
+      "step": 13739
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1196504437618296e-05,
+      "loss": 0.9574,
+      "step": 13740
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1195385818790757e-05,
+      "loss": 1.064,
+      "step": 13741
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1194267184788205e-05,
+      "loss": 0.9822,
+      "step": 13742
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1193148535624851e-05,
+      "loss": 1.074,
+      "step": 13743
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1192029871314895e-05,
+      "loss": 0.9396,
+      "step": 13744
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1190911191872536e-05,
+      "loss": 1.0183,
+      "step": 13745
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1189792497311975e-05,
+      "loss": 0.912,
+      "step": 13746
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1188673787647413e-05,
+      "loss": 1.1482,
+      "step": 13747
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1187555062893055e-05,
+      "loss": 0.9925,
+      "step": 13748
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1186436323063096e-05,
+      "loss": 1.0083,
+      "step": 13749
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1185317568171747e-05,
+      "loss": 1.0237,
+      "step": 13750
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1184198798233206e-05,
+      "loss": 0.975,
+      "step": 13751
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1183080013261673e-05,
+      "loss": 1.0242,
+      "step": 13752
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1181961213271353e-05,
+      "loss": 1.1307,
+      "step": 13753
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.118084239827645e-05,
+      "loss": 1.1166,
+      "step": 13754
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1179723568291166e-05,
+      "loss": 0.9423,
+      "step": 13755
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11786047233297e-05,
+      "loss": 1.0576,
+      "step": 13756
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1177485863406262e-05,
+      "loss": 1.097,
+      "step": 13757
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1176366988535055e-05,
+      "loss": 1.1113,
+      "step": 13758
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1175248098730276e-05,
+      "loss": 1.1706,
+      "step": 13759
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1174129194006133e-05,
+      "loss": 1.0633,
+      "step": 13760
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.117301027437683e-05,
+      "loss": 0.8364,
+      "step": 13761
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1171891339856575e-05,
+      "loss": 1.1282,
+      "step": 13762
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1170772390459568e-05,
+      "loss": 0.8806,
+      "step": 13763
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1169653426200013e-05,
+      "loss": 0.8723,
+      "step": 13764
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1168534447092119e-05,
+      "loss": 0.945,
+      "step": 13765
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1167415453150086e-05,
+      "loss": 1.0075,
+      "step": 13766
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1166296444388124e-05,
+      "loss": 0.9198,
+      "step": 13767
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1165177420820434e-05,
+      "loss": 1.0524,
+      "step": 13768
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1164058382461228e-05,
+      "loss": 0.9789,
+      "step": 13769
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1162939329324702e-05,
+      "loss": 1.0602,
+      "step": 13770
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1161820261425071e-05,
+      "loss": 1.0218,
+      "step": 13771
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1160701178776538e-05,
+      "loss": 0.949,
+      "step": 13772
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115958208139331e-05,
+      "loss": 1.0761,
+      "step": 13773
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115846296928959e-05,
+      "loss": 1.0155,
+      "step": 13774
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.115734384247959e-05,
+      "loss": 0.9738,
+      "step": 13775
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1156224700977514e-05,
+      "loss": 0.987,
+      "step": 13776
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1155105544797568e-05,
+      "loss": 1.0515,
+      "step": 13777
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1153986373953962e-05,
+      "loss": 0.9325,
+      "step": 13778
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1152867188460905e-05,
+      "loss": 1.0385,
+      "step": 13779
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1151747988332597e-05,
+      "loss": 0.8923,
+      "step": 13780
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1150628773583254e-05,
+      "loss": 0.9715,
+      "step": 13781
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.114950954422708e-05,
+      "loss": 0.9815,
+      "step": 13782
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1148390300278285e-05,
+      "loss": 1.115,
+      "step": 13783
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1147271041751075e-05,
+      "loss": 1.0076,
+      "step": 13784
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1146151768659661e-05,
+      "loss": 0.9324,
+      "step": 13785
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1145032481018252e-05,
+      "loss": 0.9555,
+      "step": 13786
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1143913178841055e-05,
+      "loss": 1.0228,
+      "step": 13787
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1142793862142279e-05,
+      "loss": 1.0255,
+      "step": 13788
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1141674530936136e-05,
+      "loss": 1.0272,
+      "step": 13789
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1140555185236832e-05,
+      "loss": 0.9859,
+      "step": 13790
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.113943582505858e-05,
+      "loss": 0.9755,
+      "step": 13791
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1138316450415587e-05,
+      "loss": 1.1598,
+      "step": 13792
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1137197061322067e-05,
+      "loss": 0.8821,
+      "step": 13793
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1136077657792226e-05,
+      "loss": 0.9433,
+      "step": 13794
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1134958239840277e-05,
+      "loss": 0.9969,
+      "step": 13795
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1133838807480428e-05,
+      "loss": 1.008,
+      "step": 13796
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1132719360726895e-05,
+      "loss": 0.9927,
+      "step": 13797
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1131599899593884e-05,
+      "loss": 0.907,
+      "step": 13798
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1130480424095607e-05,
+      "loss": 0.9655,
+      "step": 13799
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1129360934246278e-05,
+      "loss": 0.9993,
+      "step": 13800
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1128241430060105e-05,
+      "loss": 1.1111,
+      "step": 13801
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1127121911551303e-05,
+      "loss": 0.9451,
+      "step": 13802
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112600237873408e-05,
+      "loss": 1.0511,
+      "step": 13803
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1124882831622651e-05,
+      "loss": 1.0003,
+      "step": 13804
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1123763270231229e-05,
+      "loss": 0.958,
+      "step": 13805
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1122643694574025e-05,
+      "loss": 1.0433,
+      "step": 13806
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1121524104665252e-05,
+      "loss": 0.9057,
+      "step": 13807
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.112040450051912e-05,
+      "loss": 0.8678,
+      "step": 13808
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1119284882149846e-05,
+      "loss": 1.0301,
+      "step": 13809
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111816524957164e-05,
+      "loss": 0.9803,
+      "step": 13810
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1117045602798722e-05,
+      "loss": 0.9445,
+      "step": 13811
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1115925941845295e-05,
+      "loss": 1.0054,
+      "step": 13812
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111480626672558e-05,
+      "loss": 1.0252,
+      "step": 13813
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1113686577453788e-05,
+      "loss": 1.0276,
+      "step": 13814
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.111256687404414e-05,
+      "loss": 0.9195,
+      "step": 13815
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1111447156510837e-05,
+      "loss": 0.8953,
+      "step": 13816
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1110327424868104e-05,
+      "loss": 0.8964,
+      "step": 13817
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1109207679130154e-05,
+      "loss": 1.0779,
+      "step": 13818
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.11080879193112e-05,
+      "loss": 1.0398,
+      "step": 13819
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1106968145425453e-05,
+      "loss": 1.1163,
+      "step": 13820
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1105848357487134e-05,
+      "loss": 0.9849,
+      "step": 13821
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.110472855551046e-05,
+      "loss": 1.0752,
+      "step": 13822
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1103608739509641e-05,
+      "loss": 0.9684,
+      "step": 13823
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1102488909498894e-05,
+      "loss": 1.0237,
+      "step": 13824
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1101369065492437e-05,
+      "loss": 0.9622,
+      "step": 13825
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1100249207504483e-05,
+      "loss": 0.9108,
+      "step": 13826
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1099129335549252e-05,
+      "loss": 0.8744,
+      "step": 13827
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1098009449640956e-05,
+      "loss": 1.0283,
+      "step": 13828
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1096889549793816e-05,
+      "loss": 1.1054,
+      "step": 13829
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1095769636022048e-05,
+      "loss": 0.8803,
+      "step": 13830
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1094649708339862e-05,
+      "loss": 1.073,
+      "step": 13831
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1093529766761486e-05,
+      "loss": 1.0662,
+      "step": 13832
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1092409811301125e-05,
+      "loss": 1.0092,
+      "step": 13833
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.109128984197301e-05,
+      "loss": 0.9042,
+      "step": 13834
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1090169858791348e-05,
+      "loss": 0.9088,
+      "step": 13835
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1089049861770364e-05,
+      "loss": 0.9468,
+      "step": 13836
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1087929850924273e-05,
+      "loss": 0.9977,
+      "step": 13837
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1086809826267289e-05,
+      "loss": 0.976,
+      "step": 13838
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1085689787813638e-05,
+      "loss": 0.977,
+      "step": 13839
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.108456973557753e-05,
+      "loss": 0.87,
+      "step": 13840
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1083449669573192e-05,
+      "loss": 0.8405,
+      "step": 13841
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1082329589814838e-05,
+      "loss": 0.9087,
+      "step": 13842
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1081209496316687e-05,
+      "loss": 1.0248,
+      "step": 13843
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1080089389092962e-05,
+      "loss": 1.0361,
+      "step": 13844
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1078969268157877e-05,
+      "loss": 1.0652,
+      "step": 13845
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1077849133525657e-05,
+      "loss": 1.0043,
+      "step": 13846
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1076728985210516e-05,
+      "loss": 1.0344,
+      "step": 13847
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107560882322668e-05,
+      "loss": 0.8838,
+      "step": 13848
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1074488647588365e-05,
+      "loss": 0.9833,
+      "step": 13849
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.107336845830979e-05,
+      "loss": 0.9957,
+      "step": 13850
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1072248255405182e-05,
+      "loss": 0.9765,
+      "step": 13851
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1071128038888754e-05,
+      "loss": 0.9255,
+      "step": 13852
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1070007808774732e-05,
+      "loss": 0.9529,
+      "step": 13853
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1068887565077329e-05,
+      "loss": 0.993,
+      "step": 13854
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106776730781078e-05,
+      "loss": 0.918,
+      "step": 13855
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1066647036989294e-05,
+      "loss": 1.0894,
+      "step": 13856
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1065526752627096e-05,
+      "loss": 0.9968,
+      "step": 13857
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106440645473841e-05,
+      "loss": 0.8721,
+      "step": 13858
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1063286143337454e-05,
+      "loss": 1.0702,
+      "step": 13859
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1062165818438456e-05,
+      "loss": 0.9292,
+      "step": 13860
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.106104548005563e-05,
+      "loss": 0.8256,
+      "step": 13861
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1059925128203206e-05,
+      "loss": 0.8806,
+      "step": 13862
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.10588047628954e-05,
+      "loss": 0.9231,
+      "step": 13863
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1057684384146439e-05,
+      "loss": 1.0652,
+      "step": 13864
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1056563991970546e-05,
+      "loss": 0.9394,
+      "step": 13865
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.105544358638194e-05,
+      "loss": 1.0612,
+      "step": 13866
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1054323167394847e-05,
+      "loss": 1.0006,
+      "step": 13867
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1053202735023489e-05,
+      "loss": 1.0134,
+      "step": 13868
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1052082289282093e-05,
+      "loss": 0.95,
+      "step": 13869
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1050961830184879e-05,
+      "loss": 1.0438,
+      "step": 13870
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1049841357746072e-05,
+      "loss": 1.0909,
+      "step": 13871
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1048720871979897e-05,
+      "loss": 1.0673,
+      "step": 13872
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1047600372900576e-05,
+      "loss": 1.0212,
+      "step": 13873
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1046479860522332e-05,
+      "loss": 0.9399,
+      "step": 13874
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1045359334859394e-05,
+      "loss": 1.1127,
+      "step": 13875
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1044238795925985e-05,
+      "loss": 0.866,
+      "step": 13876
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1043118243736331e-05,
+      "loss": 0.972,
+      "step": 13877
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1041997678304652e-05,
+      "loss": 0.9851,
+      "step": 13878
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.104087709964518e-05,
+      "loss": 0.9506,
+      "step": 13879
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1039756507772135e-05,
+      "loss": 1.0091,
+      "step": 13880
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1038635902699743e-05,
+      "loss": 1.1204,
+      "step": 13881
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1037515284442236e-05,
+      "loss": 0.9597,
+      "step": 13882
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1036394653013834e-05,
+      "loss": 0.9559,
+      "step": 13883
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.103527400842876e-05,
+      "loss": 1.1209,
+      "step": 13884
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1034153350701247e-05,
+      "loss": 1.0673,
+      "step": 13885
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1033032679845522e-05,
+      "loss": 1.0849,
+      "step": 13886
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1031911995875801e-05,
+      "loss": 0.9849,
+      "step": 13887
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1030791298806322e-05,
+      "loss": 1.0202,
+      "step": 13888
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1029670588651307e-05,
+      "loss": 1.1076,
+      "step": 13889
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1028549865424987e-05,
+      "loss": 0.974,
+      "step": 13890
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1027429129141582e-05,
+      "loss": 0.9796,
+      "step": 13891
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1026308379815325e-05,
+      "loss": 1.0284,
+      "step": 13892
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1025187617460442e-05,
+      "loss": 1.0216,
+      "step": 13893
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102406684209116e-05,
+      "loss": 1.095,
+      "step": 13894
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1022946053721706e-05,
+      "loss": 1.011,
+      "step": 13895
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.102182525236631e-05,
+      "loss": 1.0255,
+      "step": 13896
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1020704438039203e-05,
+      "loss": 1.0539,
+      "step": 13897
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1019583610754605e-05,
+      "loss": 1.0759,
+      "step": 13898
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1018462770526752e-05,
+      "loss": 0.952,
+      "step": 13899
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1017341917369873e-05,
+      "loss": 1.0422,
+      "step": 13900
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1016221051298188e-05,
+      "loss": 0.9549,
+      "step": 13901
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1015100172325935e-05,
+      "loss": 0.9938,
+      "step": 13902
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.101397928046734e-05,
+      "loss": 1.001,
+      "step": 13903
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1012858375736632e-05,
+      "loss": 0.9994,
+      "step": 13904
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1011737458148042e-05,
+      "loss": 1.1095,
+      "step": 13905
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1010616527715797e-05,
+      "loss": 0.9655,
+      "step": 13906
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1009495584454131e-05,
+      "loss": 0.994,
+      "step": 13907
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.100837462837727e-05,
+      "loss": 1.0799,
+      "step": 13908
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1007253659499445e-05,
+      "loss": 1.0102,
+      "step": 13909
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1006132677834886e-05,
+      "loss": 1.071,
+      "step": 13910
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1005011683397826e-05,
+      "loss": 1.0797,
+      "step": 13911
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1003890676202493e-05,
+      "loss": 0.9486,
+      "step": 13912
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1002769656263119e-05,
+      "loss": 1.0218,
+      "step": 13913
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1001648623593938e-05,
+      "loss": 1.1675,
+      "step": 13914
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.1000527578209172e-05,
+      "loss": 0.9032,
+      "step": 13915
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0999406520123061e-05,
+      "loss": 1.0213,
+      "step": 13916
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0998285449349833e-05,
+      "loss": 1.0578,
+      "step": 13917
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0997164365903724e-05,
+      "loss": 0.9581,
+      "step": 13918
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0996043269798957e-05,
+      "loss": 1.0394,
+      "step": 13919
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0994922161049768e-05,
+      "loss": 1.0612,
+      "step": 13920
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0993801039670396e-05,
+      "loss": 0.9678,
+      "step": 13921
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.099267990567506e-05,
+      "loss": 0.947,
+      "step": 13922
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0991558759078005e-05,
+      "loss": 1.0016,
+      "step": 13923
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0990437599893455e-05,
+      "loss": 0.9179,
+      "step": 13924
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0989316428135648e-05,
+      "loss": 1.096,
+      "step": 13925
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0988195243818814e-05,
+      "loss": 1.0303,
+      "step": 13926
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0987074046957187e-05,
+      "loss": 1.0047,
+      "step": 13927
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0985952837565e-05,
+      "loss": 0.9573,
+      "step": 13928
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0984831615656486e-05,
+      "loss": 1.0605,
+      "step": 13929
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0983710381245877e-05,
+      "loss": 1.1359,
+      "step": 13930
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098258913434741e-05,
+      "loss": 1.0138,
+      "step": 13931
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.098146787497532e-05,
+      "loss": 0.9388,
+      "step": 13932
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0980346603143835e-05,
+      "loss": 0.9642,
+      "step": 13933
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0979225318867193e-05,
+      "loss": 1.0454,
+      "step": 13934
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.097810402215963e-05,
+      "loss": 1.0025,
+      "step": 13935
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0976982713035378e-05,
+      "loss": 1.0259,
+      "step": 13936
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0975861391508668e-05,
+      "loss": 1.0532,
+      "step": 13937
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0974740057593742e-05,
+      "loss": 1.0244,
+      "step": 13938
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0973618711304832e-05,
+      "loss": 1.1367,
+      "step": 13939
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0972497352656171e-05,
+      "loss": 1.0701,
+      "step": 13940
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0971375981661998e-05,
+      "loss": 0.9439,
+      "step": 13941
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 1.0970254598336547e-05,
+      "loss": 1.0587,
+      "step": 13942
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.096913320269405e-05,
+      "loss": 1.0193,
+      "step": 13943
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0968011794748747e-05,
+      "loss": 1.0067,
+      "step": 13944
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0966890374514872e-05,
+      "loss": 0.9915,
+      "step": 13945
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0965768942006666e-05,
+      "loss": 1.0074,
+      "step": 13946
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0964647497238352e-05,
+      "loss": 0.9593,
+      "step": 13947
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0963526040224184e-05,
+      "loss": 0.9392,
+      "step": 13948
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0962404570978387e-05,
+      "loss": 1.0195,
+      "step": 13949
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.09612830895152e-05,
+      "loss": 0.9708,
+      "step": 13950
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.096016159584886e-05,
+      "loss": 1.0667,
+      "step": 13951
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0959040089993602e-05,
+      "loss": 0.8875,
+      "step": 13952
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0957918571963668e-05,
+      "loss": 0.9873,
+      "step": 13953
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.095679704177329e-05,
+      "loss": 1.0138,
+      "step": 13954
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0955675499436712e-05,
+      "loss": 1.0061,
+      "step": 13955
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0954553944968162e-05,
+      "loss": 0.9692,
+      "step": 13956
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0953432378381888e-05,
+      "loss": 0.9414,
+      "step": 13957
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0952310799692123e-05,
+      "loss": 0.8895,
+      "step": 13958
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0951189208913105e-05,
+      "loss": 1.2491,
+      "step": 13959
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0950067606059072e-05,
+      "loss": 0.9794,
+      "step": 13960
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0948945991144259e-05,
+      "loss": 0.8487,
+      "step": 13961
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0947824364182913e-05,
+      "loss": 0.9307,
+      "step": 13962
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0946702725189267e-05,
+      "loss": 1.0205,
+      "step": 13963
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.094558107417756e-05,
+      "loss": 0.9947,
+      "step": 13964
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0944459411162033e-05,
+      "loss": 0.8505,
+      "step": 13965
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0943337736156924e-05,
+      "loss": 0.9932,
+      "step": 13966
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0942216049176474e-05,
+      "loss": 0.9459,
+      "step": 13967
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0941094350234915e-05,
+      "loss": 0.971,
+      "step": 13968
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0939972639346496e-05,
+      "loss": 0.9966,
+      "step": 13969
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0938850916525452e-05,
+      "loss": 0.9782,
+      "step": 13970
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0937729181786023e-05,
+      "loss": 0.952,
+      "step": 13971
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0936607435142452e-05,
+      "loss": 1.0563,
+      "step": 13972
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0935485676608974e-05,
+      "loss": 0.9632,
+      "step": 13973
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0934363906199834e-05,
+      "loss": 1.0065,
+      "step": 13974
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.093324212392927e-05,
+      "loss": 1.0234,
+      "step": 13975
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0932120329811526e-05,
+      "loss": 0.9676,
+      "step": 13976
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0930998523860839e-05,
+      "loss": 1.0247,
+      "step": 13977
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0929876706091448e-05,
+      "loss": 0.9101,
+      "step": 13978
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0928754876517602e-05,
+      "loss": 1.0135,
+      "step": 13979
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0927633035153535e-05,
+      "loss": 0.9942,
+      "step": 13980
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092651118201349e-05,
+      "loss": 1.0697,
+      "step": 13981
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0925389317111712e-05,
+      "loss": 0.9443,
+      "step": 13982
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.092426744046244e-05,
+      "loss": 1.0081,
+      "step": 13983
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0923145552079914e-05,
+      "loss": 0.9992,
+      "step": 13984
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0922023651978377e-05,
+      "loss": 1.098,
+      "step": 13985
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0920901740172075e-05,
+      "loss": 0.8949,
+      "step": 13986
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0919779816675247e-05,
+      "loss": 0.9221,
+      "step": 13987
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0918657881502133e-05,
+      "loss": 0.9463,
+      "step": 13988
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0917535934666979e-05,
+      "loss": 1.0073,
+      "step": 13989
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.091641397618403e-05,
+      "loss": 1.1354,
+      "step": 13990
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0915292006067522e-05,
+      "loss": 0.8892,
+      "step": 13991
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0914170024331703e-05,
+      "loss": 0.9931,
+      "step": 13992
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0913048030990818e-05,
+      "loss": 1.0257,
+      "step": 13993
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0911926026059105e-05,
+      "loss": 1.0848,
+      "step": 13994
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0910804009550808e-05,
+      "loss": 0.9908,
+      "step": 13995
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0909681981480174e-05,
+      "loss": 0.9053,
+      "step": 13996
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0908559941861447e-05,
+      "loss": 1.0647,
+      "step": 13997
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0907437890708867e-05,
+      "loss": 1.0281,
+      "step": 13998
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0906315828036682e-05,
+      "loss": 1.0081,
+      "step": 13999
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0905193753859133e-05,
+      "loss": 0.9898,
+      "step": 14000
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0904071668190465e-05,
+      "loss": 1.1094,
+      "step": 14001
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0902949571044925e-05,
+      "loss": 0.9975,
+      "step": 14002
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0901827462436753e-05,
+      "loss": 1.0197,
+      "step": 14003
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0900705342380202e-05,
+      "loss": 1.0647,
+      "step": 14004
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0899583210889506e-05,
+      "loss": 1.1779,
+      "step": 14005
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0898461067978918e-05,
+      "loss": 0.9742,
+      "step": 14006
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0897338913662682e-05,
+      "loss": 1.055,
+      "step": 14007
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0896216747955039e-05,
+      "loss": 0.9852,
+      "step": 14008
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0895094570870239e-05,
+      "loss": 0.9122,
+      "step": 14009
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0893972382422526e-05,
+      "loss": 0.9683,
+      "step": 14010
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0892850182626148e-05,
+      "loss": 1.0339,
+      "step": 14011
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0891727971495347e-05,
+      "loss": 0.9985,
+      "step": 14012
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0890605749044372e-05,
+      "loss": 0.8878,
+      "step": 14013
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0889483515287469e-05,
+      "loss": 1.1172,
+      "step": 14014
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0888361270238883e-05,
+      "loss": 1.238,
+      "step": 14015
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088723901391286e-05,
+      "loss": 0.9516,
+      "step": 14016
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0886116746323646e-05,
+      "loss": 0.9557,
+      "step": 14017
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0884994467485495e-05,
+      "loss": 0.908,
+      "step": 14018
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0883872177412646e-05,
+      "loss": 0.9362,
+      "step": 14019
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0882749876119348e-05,
+      "loss": 0.9251,
+      "step": 14020
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.088162756361985e-05,
+      "loss": 0.9208,
+      "step": 14021
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0880505239928398e-05,
+      "loss": 0.9903,
+      "step": 14022
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.087938290505924e-05,
+      "loss": 0.9032,
+      "step": 14023
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0878260559026622e-05,
+      "loss": 0.9392,
+      "step": 14024
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0877138201844794e-05,
+      "loss": 0.9311,
+      "step": 14025
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0876015833528001e-05,
+      "loss": 0.9625,
+      "step": 14026
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0874893454090496e-05,
+      "loss": 1.1256,
+      "step": 14027
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0873771063546526e-05,
+      "loss": 1.0262,
+      "step": 14028
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0872648661910335e-05,
+      "loss": 0.9023,
+      "step": 14029
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0871526249196174e-05,
+      "loss": 1.0049,
+      "step": 14030
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0870403825418293e-05,
+      "loss": 1.0163,
+      "step": 14031
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086928139059094e-05,
+      "loss": 0.9077,
+      "step": 14032
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0868158944728362e-05,
+      "loss": 0.9527,
+      "step": 14033
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.086703648784481e-05,
+      "loss": 0.9612,
+      "step": 14034
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0865914019954535e-05,
+      "loss": 0.9048,
+      "step": 14035
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0864791541071782e-05,
+      "loss": 1.0595,
+      "step": 14036
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0863669051210803e-05,
+      "loss": 0.9482,
+      "step": 14037
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0862546550385847e-05,
+      "loss": 1.009,
+      "step": 14038
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0861424038611167e-05,
+      "loss": 0.9093,
+      "step": 14039
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0860301515901006e-05,
+      "loss": 1.0178,
+      "step": 14040
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0859178982269619e-05,
+      "loss": 1.0268,
+      "step": 14041
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0858056437731257e-05,
+      "loss": 0.9137,
+      "step": 14042
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0856933882300167e-05,
+      "loss": 0.996,
+      "step": 14043
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0855811315990598e-05,
+      "loss": 0.8413,
+      "step": 14044
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0854688738816807e-05,
+      "loss": 0.9524,
+      "step": 14045
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0853566150793043e-05,
+      "loss": 1.0229,
+      "step": 14046
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0852443551933549e-05,
+      "loss": 0.9848,
+      "step": 14047
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0851320942252588e-05,
+      "loss": 1.0809,
+      "step": 14048
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0850198321764404e-05,
+      "loss": 0.9144,
+      "step": 14049
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0849075690483248e-05,
+      "loss": 1.1152,
+      "step": 14050
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0847953048423373e-05,
+      "loss": 0.9762,
+      "step": 14051
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0846830395599031e-05,
+      "loss": 0.9681,
+      "step": 14052
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0845707732024473e-05,
+      "loss": 0.9306,
+      "step": 14053
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0844585057713947e-05,
+      "loss": 1.0636,
+      "step": 14054
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0843462372681717e-05,
+      "loss": 1.068,
+      "step": 14055
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0842339676942024e-05,
+      "loss": 1.1267,
+      "step": 14056
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.084121697050912e-05,
+      "loss": 0.9403,
+      "step": 14057
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0840094253397262e-05,
+      "loss": 1.0403,
+      "step": 14058
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0838971525620701e-05,
+      "loss": 1.0239,
+      "step": 14059
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0837848787193693e-05,
+      "loss": 0.9602,
+      "step": 14060
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0836726038130483e-05,
+      "loss": 1.0115,
+      "step": 14061
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.083560327844533e-05,
+      "loss": 1.1467,
+      "step": 14062
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0834480508152486e-05,
+      "loss": 0.9697,
+      "step": 14063
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0833357727266205e-05,
+      "loss": 1.1207,
+      "step": 14064
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0832234935800735e-05,
+      "loss": 0.9684,
+      "step": 14065
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0831112133770336e-05,
+      "loss": 1.0371,
+      "step": 14066
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.082998932118926e-05,
+      "loss": 1.0061,
+      "step": 14067
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0828866498071757e-05,
+      "loss": 1.0818,
+      "step": 14068
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0827743664432087e-05,
+      "loss": 0.8624,
+      "step": 14069
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.08266208202845e-05,
+      "loss": 1.0893,
+      "step": 14070
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0825497965643245e-05,
+      "loss": 0.9685,
+      "step": 14071
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0824375100522588e-05,
+      "loss": 1.09,
+      "step": 14072
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0823252224936774e-05,
+      "loss": 1.0329,
+      "step": 14073
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0822129338900064e-05,
+      "loss": 1.0611,
+      "step": 14074
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0821006442426704e-05,
+      "loss": 0.9891,
+      "step": 14075
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.081988353553096e-05,
+      "loss": 0.9932,
+      "step": 14076
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0818760618227079e-05,
+      "loss": 1.15,
+      "step": 14077
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0817637690529317e-05,
+      "loss": 1.0322,
+      "step": 14078
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0816514752451933e-05,
+      "loss": 1.0445,
+      "step": 14079
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0815391804009176e-05,
+      "loss": 0.9946,
+      "step": 14080
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0814268845215308e-05,
+      "loss": 0.9557,
+      "step": 14081
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0813145876084576e-05,
+      "loss": 0.998,
+      "step": 14082
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0812022896631249e-05,
+      "loss": 0.9664,
+      "step": 14083
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0810899906869575e-05,
+      "loss": 0.9503,
+      "step": 14084
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0809776906813802e-05,
+      "loss": 0.8891,
+      "step": 14085
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0808653896478201e-05,
+      "loss": 0.8694,
+      "step": 14086
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080753087587702e-05,
+      "loss": 1.0473,
+      "step": 14087
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0806407845024513e-05,
+      "loss": 1.0858,
+      "step": 14088
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0805284803934946e-05,
+      "loss": 1.1069,
+      "step": 14089
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080416175262257e-05,
+      "loss": 1.0241,
+      "step": 14090
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.080303869110164e-05,
+      "loss": 0.8974,
+      "step": 14091
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0801915619386409e-05,
+      "loss": 0.8661,
+      "step": 14092
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0800792537491146e-05,
+      "loss": 1.0335,
+      "step": 14093
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0799669445430098e-05,
+      "loss": 1.1476,
+      "step": 14094
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.079854634321753e-05,
+      "loss": 1.0326,
+      "step": 14095
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0797423230867692e-05,
+      "loss": 1.0698,
+      "step": 14096
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0796300108394846e-05,
+      "loss": 0.9206,
+      "step": 14097
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0795176975813249e-05,
+      "loss": 1.0793,
+      "step": 14098
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0794053833137157e-05,
+      "loss": 0.9387,
+      "step": 14099
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0792930680380831e-05,
+      "loss": 0.9699,
+      "step": 14100
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0791807517558527e-05,
+      "loss": 0.9373,
+      "step": 14101
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0790684344684499e-05,
+      "loss": 1.002,
+      "step": 14102
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0789561161773012e-05,
+      "loss": 1.0084,
+      "step": 14103
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0788437968838326e-05,
+      "loss": 0.9745,
+      "step": 14104
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0787314765894691e-05,
+      "loss": 1.037,
+      "step": 14105
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0786191552956371e-05,
+      "loss": 1.1493,
+      "step": 14106
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0785068330037626e-05,
+      "loss": 0.9619,
+      "step": 14107
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0783945097152712e-05,
+      "loss": 1.0696,
+      "step": 14108
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0782821854315888e-05,
+      "loss": 0.9143,
+      "step": 14109
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0781698601541415e-05,
+      "loss": 0.9926,
+      "step": 14110
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0780575338843553e-05,
+      "loss": 1.0027,
+      "step": 14111
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0779452066236558e-05,
+      "loss": 1.0073,
+      "step": 14112
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077832878373469e-05,
+      "loss": 0.9549,
+      "step": 14113
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0777205491352215e-05,
+      "loss": 0.9416,
+      "step": 14114
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0776082189103383e-05,
+      "loss": 0.931,
+      "step": 14115
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077495887700246e-05,
+      "loss": 1.0691,
+      "step": 14116
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0773835555063705e-05,
+      "loss": 1.0187,
+      "step": 14117
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.077271222330138e-05,
+      "loss": 1.1258,
+      "step": 14118
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0771588881729742e-05,
+      "loss": 0.9925,
+      "step": 14119
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0770465530363052e-05,
+      "loss": 1.0093,
+      "step": 14120
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0769342169215574e-05,
+      "loss": 0.9555,
+      "step": 14121
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0768218798301564e-05,
+      "loss": 0.8287,
+      "step": 14122
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0767095417635285e-05,
+      "loss": 1.008,
+      "step": 14123
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0765972027230998e-05,
+      "loss": 0.8975,
+      "step": 14124
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0764848627102964e-05,
+      "loss": 1.0188,
+      "step": 14125
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0763725217265443e-05,
+      "loss": 0.9986,
+      "step": 14126
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.07626017977327e-05,
+      "loss": 0.9556,
+      "step": 14127
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0761478368518992e-05,
+      "loss": 0.9425,
+      "step": 14128
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0760354929638584e-05,
+      "loss": 1.0128,
+      "step": 14129
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0759231481105732e-05,
+      "loss": 1.0278,
+      "step": 14130
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0758108022934704e-05,
+      "loss": 0.9848,
+      "step": 14131
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.075698455513976e-05,
+      "loss": 1.066,
+      "step": 14132
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0755861077735161e-05,
+      "loss": 0.9337,
+      "step": 14133
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0754737590735169e-05,
+      "loss": 1.0326,
+      "step": 14134
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.075361409415405e-05,
+      "loss": 0.9844,
+      "step": 14135
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.075249058800606e-05,
+      "loss": 0.9613,
+      "step": 14136
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0751367072305468e-05,
+      "loss": 1.0976,
+      "step": 14137
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0750243547066532e-05,
+      "loss": 0.9913,
+      "step": 14138
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0749120012303518e-05,
+      "loss": 1.0502,
+      "step": 14139
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0747996468030686e-05,
+      "loss": 1.0533,
+      "step": 14140
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0746872914262298e-05,
+      "loss": 0.9595,
+      "step": 14141
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0745749351012624e-05,
+      "loss": 1.223,
+      "step": 14142
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.074462577829592e-05,
+      "loss": 0.8987,
+      "step": 14143
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0743502196126452e-05,
+      "loss": 1.0051,
+      "step": 14144
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0742378604518484e-05,
+      "loss": 0.9448,
+      "step": 14145
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0741255003486281e-05,
+      "loss": 1.0305,
+      "step": 14146
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0740131393044101e-05,
+      "loss": 1.0054,
+      "step": 14147
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0739007773206214e-05,
+      "loss": 1.0972,
+      "step": 14148
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0737884143986882e-05,
+      "loss": 1.152,
+      "step": 14149
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0736760505400368e-05,
+      "loss": 1.0416,
+      "step": 14150
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0735636857460935e-05,
+      "loss": 1.0927,
+      "step": 14151
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0734513200182851e-05,
+      "loss": 1.0654,
+      "step": 14152
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.073338953358038e-05,
+      "loss": 1.1278,
+      "step": 14153
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0732265857667785e-05,
+      "loss": 1.0112,
+      "step": 14154
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0731142172459328e-05,
+      "loss": 0.9267,
+      "step": 14155
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0730018477969279e-05,
+      "loss": 1.1314,
+      "step": 14156
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0728894774211902e-05,
+      "loss": 0.9879,
+      "step": 14157
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0727771061201457e-05,
+      "loss": 0.889,
+      "step": 14158
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0726647338952214e-05,
+      "loss": 0.9817,
+      "step": 14159
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0725523607478439e-05,
+      "loss": 0.9853,
+      "step": 14160
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0724399866794391e-05,
+      "loss": 0.9773,
+      "step": 14161
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0723276116914345e-05,
+      "loss": 0.9625,
+      "step": 14162
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.072215235785256e-05,
+      "loss": 0.9986,
+      "step": 14163
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0721028589623302e-05,
+      "loss": 0.9794,
+      "step": 14164
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0719904812240838e-05,
+      "loss": 1.0096,
+      "step": 14165
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0718781025719433e-05,
+      "loss": 0.9905,
+      "step": 14166
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0717657230073359e-05,
+      "loss": 1.0907,
+      "step": 14167
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0716533425316871e-05,
+      "loss": 0.9697,
+      "step": 14168
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0715409611464247e-05,
+      "loss": 1.1123,
+      "step": 14169
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0714285788529746e-05,
+      "loss": 0.9352,
+      "step": 14170
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0713161956527635e-05,
+      "loss": 0.847,
+      "step": 14171
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0712038115472182e-05,
+      "loss": 0.9847,
+      "step": 14172
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0710914265377656e-05,
+      "loss": 0.9606,
+      "step": 14173
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0709790406258324e-05,
+      "loss": 1.0722,
+      "step": 14174
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0708666538128444e-05,
+      "loss": 1.0801,
+      "step": 14175
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0707542661002296e-05,
+      "loss": 1.0637,
+      "step": 14176
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0706418774894143e-05,
+      "loss": 0.9459,
+      "step": 14177
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0705294879818244e-05,
+      "loss": 1.1433,
+      "step": 14178
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0704170975788877e-05,
+      "loss": 0.9564,
+      "step": 14179
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0703047062820304e-05,
+      "loss": 1.0442,
+      "step": 14180
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0701923140926797e-05,
+      "loss": 0.8001,
+      "step": 14181
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0700799210122616e-05,
+      "loss": 0.9446,
+      "step": 14182
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0699675270422038e-05,
+      "loss": 1.13,
+      "step": 14183
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069855132183933e-05,
+      "loss": 1.0547,
+      "step": 14184
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069742736438875e-05,
+      "loss": 0.9718,
+      "step": 14185
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0696303398084578e-05,
+      "loss": 1.0222,
+      "step": 14186
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0695179422941075e-05,
+      "loss": 0.9192,
+      "step": 14187
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0694055438972514e-05,
+      "loss": 0.902,
+      "step": 14188
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069293144619316e-05,
+      "loss": 1.0359,
+      "step": 14189
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0691807444617286e-05,
+      "loss": 0.9925,
+      "step": 14190
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.069068343425916e-05,
+      "loss": 1.0314,
+      "step": 14191
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0689559415133044e-05,
+      "loss": 1.0037,
+      "step": 14192
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0688435387253216e-05,
+      "loss": 0.9602,
+      "step": 14193
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.068731135063394e-05,
+      "loss": 1.0788,
+      "step": 14194
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0686187305289486e-05,
+      "loss": 0.9852,
+      "step": 14195
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0685063251234122e-05,
+      "loss": 0.9309,
+      "step": 14196
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0683939188482125e-05,
+      "loss": 1.0517,
+      "step": 14197
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0682815117047758e-05,
+      "loss": 0.931,
+      "step": 14198
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0681691036945287e-05,
+      "loss": 0.9712,
+      "step": 14199
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.068056694818899e-05,
+      "loss": 0.9285,
+      "step": 14200
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0679442850793133e-05,
+      "loss": 1.0659,
+      "step": 14201
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0678318744771988e-05,
+      "loss": 1.0311,
+      "step": 14202
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0677194630139819e-05,
+      "loss": 1.0277,
+      "step": 14203
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0676070506910907e-05,
+      "loss": 0.9779,
+      "step": 14204
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0674946375099516e-05,
+      "loss": 0.9967,
+      "step": 14205
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.067382223471991e-05,
+      "loss": 0.8408,
+      "step": 14206
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0672698085786372e-05,
+      "loss": 1.0067,
+      "step": 14207
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0671573928313164e-05,
+      "loss": 1.0665,
+      "step": 14208
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0670449762314561e-05,
+      "loss": 0.9612,
+      "step": 14209
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066932558780483e-05,
+      "loss": 0.922,
+      "step": 14210
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.066820140479825e-05,
+      "loss": 0.9537,
+      "step": 14211
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0667077213309085e-05,
+      "loss": 1.0038,
+      "step": 14212
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0665953013351605e-05,
+      "loss": 1.0349,
+      "step": 14213
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0664828804940085e-05,
+      "loss": 1.0551,
+      "step": 14214
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0663704588088799e-05,
+      "loss": 0.9748,
+      "step": 14215
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0662580362812011e-05,
+      "loss": 1.0035,
+      "step": 14216
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0661456129123999e-05,
+      "loss": 0.9072,
+      "step": 14217
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0660331887039035e-05,
+      "loss": 0.9917,
+      "step": 14218
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0659207636571384e-05,
+      "loss": 0.9174,
+      "step": 14219
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0658083377735324e-05,
+      "loss": 0.9908,
+      "step": 14220
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0656959110545129e-05,
+      "loss": 1.0181,
+      "step": 14221
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0655834835015064e-05,
+      "loss": 0.9837,
+      "step": 14222
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0654710551159404e-05,
+      "loss": 1.0434,
+      "step": 14223
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0653586258992423e-05,
+      "loss": 1.0678,
+      "step": 14224
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0652461958528397e-05,
+      "loss": 0.9389,
+      "step": 14225
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065133764978159e-05,
+      "loss": 0.9066,
+      "step": 14226
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.065021333276628e-05,
+      "loss": 1.0217,
+      "step": 14227
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0649089007496742e-05,
+      "loss": 1.0011,
+      "step": 14228
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.0647964673987244e-05,
+      "loss": 1.0796,
+      "step": 14229
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064684033225206e-05,
+      "loss": 0.9804,
+      "step": 14230
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0645715982305464e-05,
+      "loss": 0.9874,
+      "step": 14231
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0644591624161731e-05,
+      "loss": 1.0685,
+      "step": 14232
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0643467257835132e-05,
+      "loss": 1.0531,
+      "step": 14233
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.064234288333994e-05,
+      "loss": 0.9226,
+      "step": 14234
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0641218500690431e-05,
+      "loss": 1.0028,
+      "step": 14235
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0640094109900875e-05,
+      "loss": 0.9343,
+      "step": 14236
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0638969710985549e-05,
+      "loss": 0.9549,
+      "step": 14237
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0637845303958727e-05,
+      "loss": 1.0913,
+      "step": 14238
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0636720888834685e-05,
+      "loss": 0.8623,
+      "step": 14239
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0635596465627688e-05,
+      "loss": 1.0373,
+      "step": 14240
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0634472034352017e-05,
+      "loss": 0.9632,
+      "step": 14241
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0633347595021949e-05,
+      "loss": 1.0082,
+      "step": 14242
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.063222314765175e-05,
+      "loss": 1.0143,
+      "step": 14243
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0631098692255701e-05,
+      "loss": 1.1102,
+      "step": 14244
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0629974228848074e-05,
+      "loss": 1.0765,
+      "step": 14245
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0628849757443148e-05,
+      "loss": 1.0534,
+      "step": 14246
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0627725278055192e-05,
+      "loss": 1.1231,
+      "step": 14247
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0626600790698481e-05,
+      "loss": 0.9693,
+      "step": 14248
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0625476295387295e-05,
+      "loss": 0.9271,
+      "step": 14249
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0624351792135905e-05,
+      "loss": 1.0239,
+      "step": 14250
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0623227280958586e-05,
+      "loss": 0.8862,
+      "step": 14251
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0622102761869615e-05,
+      "loss": 1.158,
+      "step": 14252
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0620978234883268e-05,
+      "loss": 1.007,
+      "step": 14253
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0619853700013818e-05,
+      "loss": 0.9009,
+      "step": 14254
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0618729157275543e-05,
+      "loss": 0.9629,
+      "step": 14255
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0617604606682716e-05,
+      "loss": 0.9437,
+      "step": 14256
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0616480048249615e-05,
+      "loss": 0.9982,
+      "step": 14257
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0615355481990518e-05,
+      "loss": 1.0623,
+      "step": 14258
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0614230907919694e-05,
+      "loss": 0.9795,
+      "step": 14259
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0613106326051428e-05,
+      "loss": 0.9518,
+      "step": 14260
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0611981736399988e-05,
+      "loss": 1.0379,
+      "step": 14261
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0610857138979653e-05,
+      "loss": 1.0038,
+      "step": 14262
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0609732533804704e-05,
+      "loss": 1.1144,
+      "step": 14263
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060860792088941e-05,
+      "loss": 0.9896,
+      "step": 14264
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0607483300248052e-05,
+      "loss": 1.0148,
+      "step": 14265
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0606358671894905e-05,
+      "loss": 0.9601,
+      "step": 14266
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060523403584425e-05,
+      "loss": 0.9916,
+      "step": 14267
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0604109392110356e-05,
+      "loss": 1.0001,
+      "step": 14268
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0602984740707506e-05,
+      "loss": 1.0503,
+      "step": 14269
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.060186008164998e-05,
+      "loss": 1.0901,
+      "step": 14270
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0600735414952046e-05,
+      "loss": 1.0716,
+      "step": 14271
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0599610740627986e-05,
+      "loss": 0.9947,
+      "step": 14272
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0598486058692079e-05,
+      "loss": 0.9137,
+      "step": 14273
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0597361369158601e-05,
+      "loss": 0.984,
+      "step": 14274
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0596236672041824e-05,
+      "loss": 1.0003,
+      "step": 14275
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0595111967356037e-05,
+      "loss": 0.9058,
+      "step": 14276
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0593987255115512e-05,
+      "loss": 1.0503,
+      "step": 14277
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0592862535334522e-05,
+      "loss": 1.0056,
+      "step": 14278
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0591737808027351e-05,
+      "loss": 1.0553,
+      "step": 14279
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0590613073208275e-05,
+      "loss": 1.0039,
+      "step": 14280
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0589488330891576e-05,
+      "loss": 1.1661,
+      "step": 14281
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0588363581091521e-05,
+      "loss": 0.9628,
+      "step": 14282
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0587238823822403e-05,
+      "loss": 0.8883,
+      "step": 14283
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0586114059098493e-05,
+      "loss": 0.9492,
+      "step": 14284
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0584989286934067e-05,
+      "loss": 0.9049,
+      "step": 14285
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0583864507343406e-05,
+      "loss": 1.0574,
+      "step": 14286
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.058273972034079e-05,
+      "loss": 0.9195,
+      "step": 14287
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0581614925940499e-05,
+      "loss": 1.0339,
+      "step": 14288
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0580490124156806e-05,
+      "loss": 1.0274,
+      "step": 14289
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0579365315003998e-05,
+      "loss": 0.9084,
+      "step": 14290
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0578240498496351e-05,
+      "loss": 0.9131,
+      "step": 14291
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0577115674648136e-05,
+      "loss": 0.9512,
+      "step": 14292
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0575990843473645e-05,
+      "loss": 0.9296,
+      "step": 14293
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0574866004987148e-05,
+      "loss": 0.9435,
+      "step": 14294
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0573741159202933e-05,
+      "loss": 0.9297,
+      "step": 14295
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0572616306135268e-05,
+      "loss": 0.9382,
+      "step": 14296
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0571491445798443e-05,
+      "loss": 0.9683,
+      "step": 14297
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0570366578206736e-05,
+      "loss": 0.9806,
+      "step": 14298
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0569241703374417e-05,
+      "loss": 0.9839,
+      "step": 14299
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0568116821315782e-05,
+      "loss": 1.0557,
+      "step": 14300
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0566991932045098e-05,
+      "loss": 1.0106,
+      "step": 14301
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0565867035576654e-05,
+      "loss": 1.0052,
+      "step": 14302
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.056474213192472e-05,
+      "loss": 0.9571,
+      "step": 14303
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0563617221103583e-05,
+      "loss": 1.0808,
+      "step": 14304
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0562492303127527e-05,
+      "loss": 0.9429,
+      "step": 14305
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0561367378010821e-05,
+      "loss": 1.1497,
+      "step": 14306
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0560242445767759e-05,
+      "loss": 0.9843,
+      "step": 14307
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.055911750641261e-05,
+      "loss": 0.9505,
+      "step": 14308
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0557992559959662e-05,
+      "loss": 1.0896,
+      "step": 14309
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0556867606423191e-05,
+      "loss": 0.8521,
+      "step": 14310
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0555742645817487e-05,
+      "loss": 0.8913,
+      "step": 14311
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.055461767815682e-05,
+      "loss": 0.9247,
+      "step": 14312
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0553492703455472e-05,
+      "loss": 0.9758,
+      "step": 14313
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0552367721727732e-05,
+      "loss": 1.0369,
+      "step": 14314
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0551242732987877e-05,
+      "loss": 1.0192,
+      "step": 14315
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0550117737250185e-05,
+      "loss": 1.0909,
+      "step": 14316
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0548992734528942e-05,
+      "loss": 1.1653,
+      "step": 14317
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054786772483843e-05,
+      "loss": 0.9707,
+      "step": 14318
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054674270819293e-05,
+      "loss": 1.0064,
+      "step": 14319
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0545617684606717e-05,
+      "loss": 1.0738,
+      "step": 14320
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0544492654094084e-05,
+      "loss": 0.993,
+      "step": 14321
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0543367616669303e-05,
+      "loss": 1.0337,
+      "step": 14322
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.054224257234666e-05,
+      "loss": 0.9022,
+      "step": 14323
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0541117521140436e-05,
+      "loss": 1.0637,
+      "step": 14324
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0539992463064918e-05,
+      "loss": 1.0428,
+      "step": 14325
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0538867398134382e-05,
+      "loss": 0.9381,
+      "step": 14326
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0537742326363113e-05,
+      "loss": 0.9706,
+      "step": 14327
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0536617247765397e-05,
+      "loss": 0.9422,
+      "step": 14328
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053549216235551e-05,
+      "loss": 0.9981,
+      "step": 14329
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0534367070147735e-05,
+      "loss": 1.0038,
+      "step": 14330
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0533241971156359e-05,
+      "loss": 0.9332,
+      "step": 14331
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.053211686539566e-05,
+      "loss": 0.9667,
+      "step": 14332
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0530991752879926e-05,
+      "loss": 0.8801,
+      "step": 14333
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0529866633623436e-05,
+      "loss": 0.8992,
+      "step": 14334
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0528741507640476e-05,
+      "loss": 1.1253,
+      "step": 14335
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0527616374945327e-05,
+      "loss": 1.0184,
+      "step": 14336
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0526491235552272e-05,
+      "loss": 1.0585,
+      "step": 14337
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0525366089475593e-05,
+      "loss": 0.998,
+      "step": 14338
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0524240936729577e-05,
+      "loss": 1.0627,
+      "step": 14339
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0523115777328506e-05,
+      "loss": 0.873,
+      "step": 14340
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.052199061128666e-05,
+      "loss": 1.0407,
+      "step": 14341
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0520865438618331e-05,
+      "loss": 1.0668,
+      "step": 14342
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0519740259337793e-05,
+      "loss": 1.1266,
+      "step": 14343
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0518615073459334e-05,
+      "loss": 1.0658,
+      "step": 14344
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0517489880997241e-05,
+      "loss": 0.933,
+      "step": 14345
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0516364681965794e-05,
+      "loss": 1.0293,
+      "step": 14346
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0515239476379277e-05,
+      "loss": 0.8889,
+      "step": 14347
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0514114264251974e-05,
+      "loss": 0.9027,
+      "step": 14348
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0512989045598171e-05,
+      "loss": 0.9201,
+      "step": 14349
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0511863820432153e-05,
+      "loss": 0.9407,
+      "step": 14350
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0510738588768199e-05,
+      "loss": 0.8697,
+      "step": 14351
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.05096133506206e-05,
+      "loss": 0.9983,
+      "step": 14352
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0508488106003637e-05,
+      "loss": 0.9753,
+      "step": 14353
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0507362854931595e-05,
+      "loss": 0.9069,
+      "step": 14354
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0506237597418757e-05,
+      "loss": 0.9778,
+      "step": 14355
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0505112333479414e-05,
+      "loss": 1.0033,
+      "step": 14356
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0503987063127845e-05,
+      "loss": 1.0429,
+      "step": 14357
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0502861786378335e-05,
+      "loss": 0.9526,
+      "step": 14358
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0501736503245171e-05,
+      "loss": 1.0378,
+      "step": 14359
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0500611213742638e-05,
+      "loss": 0.9226,
+      "step": 14360
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049948591788502e-05,
+      "loss": 0.9987,
+      "step": 14361
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0498360615686602e-05,
+      "loss": 0.9124,
+      "step": 14362
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0497235307161673e-05,
+      "loss": 0.97,
+      "step": 14363
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0496109992324513e-05,
+      "loss": 1.0103,
+      "step": 14364
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049498467118941e-05,
+      "loss": 1.0209,
+      "step": 14365
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049385934377065e-05,
+      "loss": 1.0829,
+      "step": 14366
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.049273401008252e-05,
+      "loss": 0.9979,
+      "step": 14367
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0491608670139301e-05,
+      "loss": 1.0698,
+      "step": 14368
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0490483323955283e-05,
+      "loss": 1.135,
+      "step": 14369
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0489357971544753e-05,
+      "loss": 0.9616,
+      "step": 14370
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0488232612921992e-05,
+      "loss": 1.0972,
+      "step": 14371
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0487107248101289e-05,
+      "loss": 0.9447,
+      "step": 14372
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0485981877096929e-05,
+      "loss": 0.9926,
+      "step": 14373
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0484856499923202e-05,
+      "loss": 0.9983,
+      "step": 14374
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048373111659439e-05,
+      "loss": 0.9799,
+      "step": 14375
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0482605727124777e-05,
+      "loss": 1.0377,
+      "step": 14376
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0481480331528658e-05,
+      "loss": 0.9671,
+      "step": 14377
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.048035492982031e-05,
+      "loss": 0.9575,
+      "step": 14378
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0479229522014027e-05,
+      "loss": 0.9698,
+      "step": 14379
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0478104108124093e-05,
+      "loss": 1.1611,
+      "step": 14380
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0476978688164794e-05,
+      "loss": 0.9434,
+      "step": 14381
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0475853262150413e-05,
+      "loss": 1.0419,
+      "step": 14382
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0474727830095247e-05,
+      "loss": 0.9426,
+      "step": 14383
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0473602392013575e-05,
+      "loss": 1.0263,
+      "step": 14384
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0472476947919685e-05,
+      "loss": 1.0395,
+      "step": 14385
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0471351497827867e-05,
+      "loss": 1.0327,
+      "step": 14386
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0470226041752405e-05,
+      "loss": 0.9267,
+      "step": 14387
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0469100579707592e-05,
+      "loss": 0.8556,
+      "step": 14388
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0467975111707705e-05,
+      "loss": 0.99,
+      "step": 14389
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046684963776704e-05,
+      "loss": 1.1487,
+      "step": 14390
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0465724157899883e-05,
+      "loss": 0.9119,
+      "step": 14391
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0464598672120521e-05,
+      "loss": 0.9805,
+      "step": 14392
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.046347318044324e-05,
+      "loss": 0.9526,
+      "step": 14393
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0462347682882328e-05,
+      "loss": 0.9799,
+      "step": 14394
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0461222179452078e-05,
+      "loss": 1.0254,
+      "step": 14395
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0460096670166766e-05,
+      "loss": 0.867,
+      "step": 14396
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0458971155040694e-05,
+      "loss": 1.107,
+      "step": 14397
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0457845634088144e-05,
+      "loss": 0.9674,
+      "step": 14398
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0456720107323398e-05,
+      "loss": 0.9124,
+      "step": 14399
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0455594574760755e-05,
+      "loss": 1.1154,
+      "step": 14400
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0454469036414497e-05,
+      "loss": 1.0203,
+      "step": 14401
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0453343492298913e-05,
+      "loss": 0.8579,
+      "step": 14402
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.045221794242829e-05,
+      "loss": 0.9912,
+      "step": 14403
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0451092386816922e-05,
+      "loss": 0.9798,
+      "step": 14404
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0449966825479091e-05,
+      "loss": 1.0119,
+      "step": 14405
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0448841258429088e-05,
+      "loss": 0.9605,
+      "step": 14406
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0447715685681205e-05,
+      "loss": 0.9068,
+      "step": 14407
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0446590107249726e-05,
+      "loss": 1.0168,
+      "step": 14408
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0445464523148943e-05,
+      "loss": 0.8773,
+      "step": 14409
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0444338933393142e-05,
+      "loss": 0.8924,
+      "step": 14410
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0443213337996615e-05,
+      "loss": 1.1073,
+      "step": 14411
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0442087736973649e-05,
+      "loss": 0.9442,
+      "step": 14412
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.044096213033853e-05,
+      "loss": 0.952,
+      "step": 14413
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0439836518105557e-05,
+      "loss": 1.0032,
+      "step": 14414
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043871090028901e-05,
+      "loss": 1.1161,
+      "step": 14415
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0437585276903184e-05,
+      "loss": 1.0474,
+      "step": 14416
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043645964796236e-05,
+      "loss": 1.1442,
+      "step": 14417
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043533401348084e-05,
+      "loss": 0.9281,
+      "step": 14418
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0434208373472904e-05,
+      "loss": 0.9738,
+      "step": 14419
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0433082727952842e-05,
+      "loss": 0.9024,
+      "step": 14420
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043195707693495e-05,
+      "loss": 1.044,
+      "step": 14421
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.043083142043351e-05,
+      "loss": 0.9102,
+      "step": 14422
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042970575846282e-05,
+      "loss": 1.0905,
+      "step": 14423
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0428580091037164e-05,
+      "loss": 1.0115,
+      "step": 14424
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0427454418170834e-05,
+      "loss": 1.0339,
+      "step": 14425
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042632873987812e-05,
+      "loss": 0.8372,
+      "step": 14426
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0425203056173307e-05,
+      "loss": 0.9797,
+      "step": 14427
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0424077367070696e-05,
+      "loss": 1.0304,
+      "step": 14428
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0422951672584566e-05,
+      "loss": 1.0295,
+      "step": 14429
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0421825972729215e-05,
+      "loss": 1.0079,
+      "step": 14430
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.042070026751893e-05,
+      "loss": 1.0581,
+      "step": 14431
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0419574556968007e-05,
+      "loss": 1.0942,
+      "step": 14432
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0418448841090726e-05,
+      "loss": 1.0172,
+      "step": 14433
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0417323119901384e-05,
+      "loss": 0.9379,
+      "step": 14434
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0416197393414272e-05,
+      "loss": 1.0166,
+      "step": 14435
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0415071661643678e-05,
+      "loss": 0.9741,
+      "step": 14436
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0413945924603894e-05,
+      "loss": 0.8937,
+      "step": 14437
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0412820182309213e-05,
+      "loss": 0.9758,
+      "step": 14438
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0411694434773925e-05,
+      "loss": 0.9861,
+      "step": 14439
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0410568682012319e-05,
+      "loss": 1.0404,
+      "step": 14440
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0409442924038687e-05,
+      "loss": 1.0604,
+      "step": 14441
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040831716086732e-05,
+      "loss": 1.0427,
+      "step": 14442
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0407191392512508e-05,
+      "loss": 1.0507,
+      "step": 14443
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0406065618988545e-05,
+      "loss": 1.0237,
+      "step": 14444
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.040493984030972e-05,
+      "loss": 0.9225,
+      "step": 14445
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0403814056490327e-05,
+      "loss": 1.0199,
+      "step": 14446
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0402688267544653e-05,
+      "loss": 1.1355,
+      "step": 14447
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0401562473486994e-05,
+      "loss": 1.0298,
+      "step": 14448
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0400436674331642e-05,
+      "loss": 0.906,
+      "step": 14449
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0399310870092885e-05,
+      "loss": 0.924,
+      "step": 14450
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0398185060785013e-05,
+      "loss": 0.9928,
+      "step": 14451
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0397059246422324e-05,
+      "loss": 0.8738,
+      "step": 14452
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0395933427019107e-05,
+      "loss": 1.0464,
+      "step": 14453
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0394807602589652e-05,
+      "loss": 1.081,
+      "step": 14454
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.039368177314825e-05,
+      "loss": 1.0041,
+      "step": 14455
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0392555938709201e-05,
+      "loss": 0.9577,
+      "step": 14456
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0391430099286789e-05,
+      "loss": 0.9709,
+      "step": 14457
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0390304254895308e-05,
+      "loss": 0.9622,
+      "step": 14458
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038917840554905e-05,
+      "loss": 1.0409,
+      "step": 14459
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0388052551262312e-05,
+      "loss": 1.075,
+      "step": 14460
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0386926692049377e-05,
+      "loss": 0.997,
+      "step": 14461
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0385800827924546e-05,
+      "loss": 1.0359,
+      "step": 14462
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.038467495890211e-05,
+      "loss": 1.0077,
+      "step": 14463
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0383549084996358e-05,
+      "loss": 1.1857,
+      "step": 14464
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0382423206221584e-05,
+      "loss": 1.0449,
+      "step": 14465
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0381297322592081e-05,
+      "loss": 0.9348,
+      "step": 14466
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0380171434122144e-05,
+      "loss": 0.9746,
+      "step": 14467
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0379045540826061e-05,
+      "loss": 1.1089,
+      "step": 14468
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0377919642718129e-05,
+      "loss": 0.887,
+      "step": 14469
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0376793739812638e-05,
+      "loss": 0.9674,
+      "step": 14470
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0375667832123884e-05,
+      "loss": 0.9608,
+      "step": 14471
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0374541919666154e-05,
+      "loss": 0.8675,
+      "step": 14472
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.037341600245375e-05,
+      "loss": 0.9915,
+      "step": 14473
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0372290080500958e-05,
+      "loss": 1.0336,
+      "step": 14474
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0371164153822072e-05,
+      "loss": 1.0056,
+      "step": 14475
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0370038222431392e-05,
+      "loss": 1.0429,
+      "step": 14476
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0368912286343202e-05,
+      "loss": 0.9412,
+      "step": 14477
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.03677863455718e-05,
+      "loss": 1.0862,
+      "step": 14478
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0366660400131479e-05,
+      "loss": 1.1547,
+      "step": 14479
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0365534450036532e-05,
+      "loss": 1.0363,
+      "step": 14480
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0364408495301254e-05,
+      "loss": 0.8758,
+      "step": 14481
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0363282535939938e-05,
+      "loss": 0.9876,
+      "step": 14482
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0362156571966875e-05,
+      "loss": 1.0598,
+      "step": 14483
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0361030603396365e-05,
+      "loss": 1.0158,
+      "step": 14484
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0359904630242694e-05,
+      "loss": 1.0082,
+      "step": 14485
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035877865252016e-05,
+      "loss": 0.9799,
+      "step": 14486
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0357652670243057e-05,
+      "loss": 0.9465,
+      "step": 14487
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0356526683425678e-05,
+      "loss": 0.9932,
+      "step": 14488
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0355400692082318e-05,
+      "loss": 0.9505,
+      "step": 14489
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0354274696227268e-05,
+      "loss": 1.0879,
+      "step": 14490
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0353148695874831e-05,
+      "loss": 1.0311,
+      "step": 14491
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0352022691039287e-05,
+      "loss": 0.9647,
+      "step": 14492
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.035089668173494e-05,
+      "loss": 0.9455,
+      "step": 14493
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0349770667976084e-05,
+      "loss": 1.026,
+      "step": 14494
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0348644649777014e-05,
+      "loss": 1.0125,
+      "step": 14495
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0347518627152018e-05,
+      "loss": 1.0088,
+      "step": 14496
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0346392600115396e-05,
+      "loss": 0.896,
+      "step": 14497
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0345266568681442e-05,
+      "loss": 1.01,
+      "step": 14498
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0344140532864449e-05,
+      "loss": 1.0669,
+      "step": 14499
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0343014492678708e-05,
+      "loss": 1.0096,
+      "step": 14500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0341888448138523e-05,
+      "loss": 1.0346,
+      "step": 14501
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0340762399258184e-05,
+      "loss": 0.9926,
+      "step": 14502
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0339636346051982e-05,
+      "loss": 1.0749,
+      "step": 14503
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0338510288534216e-05,
+      "loss": 0.954,
+      "step": 14504
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0337384226719183e-05,
+      "loss": 0.9682,
+      "step": 14505
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0336258160621174e-05,
+      "loss": 1.0425,
+      "step": 14506
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0335132090254482e-05,
+      "loss": 0.9694,
+      "step": 14507
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0334006015633409e-05,
+      "loss": 0.9036,
+      "step": 14508
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0332879936772246e-05,
+      "loss": 1.0311,
+      "step": 14509
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0331753853685286e-05,
+      "loss": 0.8946,
+      "step": 14510
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.033062776638683e-05,
+      "loss": 1.0012,
+      "step": 14511
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.032950167489117e-05,
+      "loss": 1.0669,
+      "step": 14512
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0328375579212596e-05,
+      "loss": 1.0122,
+      "step": 14513
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0327249479365414e-05,
+      "loss": 1.1455,
+      "step": 14514
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0326123375363912e-05,
+      "loss": 0.9675,
+      "step": 14515
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 1.0324997267222392e-05,
+      "loss": 1.023,
+      "step": 14516
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.032387115495514e-05,
+      "loss": 1.0305,
+      "step": 14517
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0322745038576459e-05,
+      "loss": 0.9149,
+      "step": 14518
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0321618918100645e-05,
+      "loss": 0.8864,
+      "step": 14519
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0320492793541985e-05,
+      "loss": 0.8831,
+      "step": 14520
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0319366664914785e-05,
+      "loss": 0.942,
+      "step": 14521
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0318240532233337e-05,
+      "loss": 1.0362,
+      "step": 14522
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0317114395511938e-05,
+      "loss": 0.9383,
+      "step": 14523
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0315988254764877e-05,
+      "loss": 1.0955,
+      "step": 14524
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.031486211000646e-05,
+      "loss": 1.0276,
+      "step": 14525
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0313735961250981e-05,
+      "loss": 1.0086,
+      "step": 14526
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0312609808512728e-05,
+      "loss": 1.1188,
+      "step": 14527
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0311483651806006e-05,
+      "loss": 1.0007,
+      "step": 14528
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0310357491145107e-05,
+      "loss": 0.9171,
+      "step": 14529
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030923132654433e-05,
+      "loss": 1.0789,
+      "step": 14530
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0308105158017965e-05,
+      "loss": 1.01,
+      "step": 14531
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0306978985580318e-05,
+      "loss": 1.0722,
+      "step": 14532
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0305852809245677e-05,
+      "loss": 0.9365,
+      "step": 14533
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0304726629028341e-05,
+      "loss": 1.032,
+      "step": 14534
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.030360044494261e-05,
+      "loss": 0.9584,
+      "step": 14535
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0302474257002775e-05,
+      "loss": 0.9903,
+      "step": 14536
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0301348065223133e-05,
+      "loss": 0.9242,
+      "step": 14537
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0300221869617986e-05,
+      "loss": 0.9707,
+      "step": 14538
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0299095670201629e-05,
+      "loss": 0.9987,
+      "step": 14539
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0297969466988356e-05,
+      "loss": 1.021,
+      "step": 14540
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0296843259992459e-05,
+      "loss": 0.9762,
+      "step": 14541
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0295717049228247e-05,
+      "loss": 1.0119,
+      "step": 14542
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0294590834710008e-05,
+      "loss": 1.0052,
+      "step": 14543
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.029346461645204e-05,
+      "loss": 1.0925,
+      "step": 14544
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0292338394468641e-05,
+      "loss": 1.0737,
+      "step": 14545
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0291212168774113e-05,
+      "loss": 0.9585,
+      "step": 14546
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0290085939382746e-05,
+      "loss": 1.0122,
+      "step": 14547
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0288959706308837e-05,
+      "loss": 0.9312,
+      "step": 14548
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0287833469566689e-05,
+      "loss": 1.0466,
+      "step": 14549
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0286707229170593e-05,
+      "loss": 0.9384,
+      "step": 14550
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028558098513485e-05,
+      "loss": 0.9618,
+      "step": 14551
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0284454737473754e-05,
+      "loss": 1.0283,
+      "step": 14552
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.028332848620161e-05,
+      "loss": 1.1111,
+      "step": 14553
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0282202231332706e-05,
+      "loss": 0.9396,
+      "step": 14554
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0281075972881342e-05,
+      "loss": 1.0623,
+      "step": 14555
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027994971086182e-05,
+      "loss": 0.8861,
+      "step": 14556
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0278823445288432e-05,
+      "loss": 0.9871,
+      "step": 14557
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027769717617548e-05,
+      "loss": 0.9974,
+      "step": 14558
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0276570903537257e-05,
+      "loss": 1.0788,
+      "step": 14559
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0275444627388067e-05,
+      "loss": 0.8742,
+      "step": 14560
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02743183477422e-05,
+      "loss": 1.0004,
+      "step": 14561
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0273192064613957e-05,
+      "loss": 1.0359,
+      "step": 14562
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027206577801764e-05,
+      "loss": 1.0072,
+      "step": 14563
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.027093948796754e-05,
+      "loss": 1.0881,
+      "step": 14564
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0269813194477958e-05,
+      "loss": 0.951,
+      "step": 14565
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0268686897563191e-05,
+      "loss": 1.0708,
+      "step": 14566
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0267560597237543e-05,
+      "loss": 0.9721,
+      "step": 14567
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0266434293515302e-05,
+      "loss": 0.9092,
+      "step": 14568
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0265307986410771e-05,
+      "loss": 0.9723,
+      "step": 14569
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026418167593825e-05,
+      "loss": 0.9784,
+      "step": 14570
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0263055362112033e-05,
+      "loss": 0.9512,
+      "step": 14571
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026192904494642e-05,
+      "loss": 1.0641,
+      "step": 14572
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.026080272445571e-05,
+      "loss": 1.0102,
+      "step": 14573
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0259676400654201e-05,
+      "loss": 0.9169,
+      "step": 14574
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025855007355619e-05,
+      "loss": 0.9492,
+      "step": 14575
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0257423743175977e-05,
+      "loss": 0.94,
+      "step": 14576
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.025629740952786e-05,
+      "loss": 1.0544,
+      "step": 14577
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0255171072626135e-05,
+      "loss": 0.931,
+      "step": 14578
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0254044732485103e-05,
+      "loss": 1.0404,
+      "step": 14579
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0252918389119063e-05,
+      "loss": 1.0821,
+      "step": 14580
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0251792042542312e-05,
+      "loss": 1.0034,
+      "step": 14581
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0250665692769148e-05,
+      "loss": 0.915,
+      "step": 14582
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0249539339813873e-05,
+      "loss": 1.0581,
+      "step": 14583
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0248412983690783e-05,
+      "loss": 0.9917,
+      "step": 14584
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0247286624414174e-05,
+      "loss": 0.9613,
+      "step": 14585
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024616026199835e-05,
+      "loss": 1.0566,
+      "step": 14586
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0245033896457606e-05,
+      "loss": 1.0064,
+      "step": 14587
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0243907527806244e-05,
+      "loss": 1.0178,
+      "step": 14588
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.024278115605856e-05,
+      "loss": 0.9582,
+      "step": 14589
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0241654781228856e-05,
+      "loss": 1.0685,
+      "step": 14590
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0240528403331429e-05,
+      "loss": 1.0135,
+      "step": 14591
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0239402022380575e-05,
+      "loss": 1.0667,
+      "step": 14592
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0238275638390597e-05,
+      "loss": 0.905,
+      "step": 14593
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0237149251375795e-05,
+      "loss": 1.1514,
+      "step": 14594
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0236022861350466e-05,
+      "loss": 1.0956,
+      "step": 14595
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0234896468328907e-05,
+      "loss": 0.7372,
+      "step": 14596
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0233770072325419e-05,
+      "loss": 1.0583,
+      "step": 14597
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0232643673354305e-05,
+      "loss": 1.0965,
+      "step": 14598
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.023151727142986e-05,
+      "loss": 1.0115,
+      "step": 14599
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0230390866566381e-05,
+      "loss": 0.9705,
+      "step": 14600
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0229264458778174e-05,
+      "loss": 1.1127,
+      "step": 14601
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0228138048079535e-05,
+      "loss": 0.9861,
+      "step": 14602
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0227011634484763e-05,
+      "loss": 1.0489,
+      "step": 14603
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0225885218008156e-05,
+      "loss": 0.9869,
+      "step": 14604
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0224758798664016e-05,
+      "loss": 1.0029,
+      "step": 14605
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0223632376466642e-05,
+      "loss": 1.0007,
+      "step": 14606
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.022250595143033e-05,
+      "loss": 1.059,
+      "step": 14607
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0221379523569385e-05,
+      "loss": 0.9894,
+      "step": 14608
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0220253092898107e-05,
+      "loss": 1.0774,
+      "step": 14609
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0219126659430789e-05,
+      "loss": 0.9684,
+      "step": 14610
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0218000223181736e-05,
+      "loss": 1.0493,
+      "step": 14611
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0216873784165249e-05,
+      "loss": 0.9534,
+      "step": 14612
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0215747342395621e-05,
+      "loss": 1.1049,
+      "step": 14613
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0214620897887157e-05,
+      "loss": 0.9522,
+      "step": 14614
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0213494450654156e-05,
+      "loss": 1.01,
+      "step": 14615
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.021236800071092e-05,
+      "loss": 1.0594,
+      "step": 14616
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0211241548071742e-05,
+      "loss": 1.0452,
+      "step": 14617
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0210115092750928e-05,
+      "loss": 1.0435,
+      "step": 14618
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020898863476278e-05,
+      "loss": 1.0251,
+      "step": 14619
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0207862174121589e-05,
+      "loss": 0.9589,
+      "step": 14620
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0206735710841661e-05,
+      "loss": 1.0781,
+      "step": 14621
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0205609244937297e-05,
+      "loss": 0.9508,
+      "step": 14622
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.02044827764228e-05,
+      "loss": 1.0193,
+      "step": 14623
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020335630531246e-05,
+      "loss": 0.9939,
+      "step": 14624
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0202229831620583e-05,
+      "loss": 1.0139,
+      "step": 14625
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.020110335536147e-05,
+      "loss": 0.9639,
+      "step": 14626
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0199976876549418e-05,
+      "loss": 0.9823,
+      "step": 14627
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0198850395198734e-05,
+      "loss": 0.9122,
+      "step": 14628
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0197723911323709e-05,
+      "loss": 1.1073,
+      "step": 14629
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0196597424938652e-05,
+      "loss": 1.0621,
+      "step": 14630
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0195470936057859e-05,
+      "loss": 0.9173,
+      "step": 14631
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0194344444695628e-05,
+      "loss": 1.0165,
+      "step": 14632
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0193217950866265e-05,
+      "loss": 1.0677,
+      "step": 14633
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0192091454584066e-05,
+      "loss": 1.0175,
+      "step": 14634
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0190964955863334e-05,
+      "loss": 0.9799,
+      "step": 14635
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0189838454718368e-05,
+      "loss": 0.9338,
+      "step": 14636
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018871195116347e-05,
+      "loss": 0.8804,
+      "step": 14637
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0187585445212939e-05,
+      "loss": 0.9661,
+      "step": 14638
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0186458936881076e-05,
+      "loss": 1.0189,
+      "step": 14639
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0185332426182184e-05,
+      "loss": 0.9042,
+      "step": 14640
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0184205913130555e-05,
+      "loss": 1.0617,
+      "step": 14641
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0183079397740506e-05,
+      "loss": 1.1009,
+      "step": 14642
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0181952880026323e-05,
+      "loss": 1.1206,
+      "step": 14643
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.018082636000231e-05,
+      "loss": 1.0626,
+      "step": 14644
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0179699837682775e-05,
+      "loss": 0.8665,
+      "step": 14645
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0178573313082008e-05,
+      "loss": 1.0699,
+      "step": 14646
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017744678621432e-05,
+      "loss": 0.9417,
+      "step": 14647
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0176320257094e-05,
+      "loss": 1.0653,
+      "step": 14648
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0175193725735364e-05,
+      "loss": 0.9191,
+      "step": 14649
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01740671921527e-05,
+      "loss": 0.9781,
+      "step": 14650
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0172940656360315e-05,
+      "loss": 0.9692,
+      "step": 14651
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.017181411837251e-05,
+      "loss": 0.9342,
+      "step": 14652
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0170687578203582e-05,
+      "loss": 1.219,
+      "step": 14653
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0169561035867839e-05,
+      "loss": 1.0786,
+      "step": 14654
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0168434491379573e-05,
+      "loss": 1.0275,
+      "step": 14655
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0167307944753093e-05,
+      "loss": 0.991,
+      "step": 14656
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0166181396002696e-05,
+      "loss": 0.9165,
+      "step": 14657
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0165054845142686e-05,
+      "loss": 0.9704,
+      "step": 14658
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0163928292187362e-05,
+      "loss": 0.9924,
+      "step": 14659
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0162801737151025e-05,
+      "loss": 1.0673,
+      "step": 14660
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0161675180047976e-05,
+      "loss": 1.0852,
+      "step": 14661
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0160548620892518e-05,
+      "loss": 1.1406,
+      "step": 14662
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0159422059698953e-05,
+      "loss": 0.9806,
+      "step": 14663
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0158295496481578e-05,
+      "loss": 1.0526,
+      "step": 14664
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0157168931254698e-05,
+      "loss": 0.9909,
+      "step": 14665
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0156042364032611e-05,
+      "loss": 1.0471,
+      "step": 14666
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0154915794829625e-05,
+      "loss": 1.0127,
+      "step": 14667
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0153789223660035e-05,
+      "loss": 0.9176,
+      "step": 14668
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0152662650538145e-05,
+      "loss": 0.9925,
+      "step": 14669
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0151536075478257e-05,
+      "loss": 1.0347,
+      "step": 14670
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.015040949849467e-05,
+      "loss": 0.982,
+      "step": 14671
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0149282919601686e-05,
+      "loss": 0.9259,
+      "step": 14672
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0148156338813608e-05,
+      "loss": 1.1433,
+      "step": 14673
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0147029756144739e-05,
+      "loss": 1.0764,
+      "step": 14674
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0145903171609377e-05,
+      "loss": 0.9913,
+      "step": 14675
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0144776585221825e-05,
+      "loss": 1.069,
+      "step": 14676
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0143649996996388e-05,
+      "loss": 0.9521,
+      "step": 14677
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0142523406947362e-05,
+      "loss": 0.9471,
+      "step": 14678
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0141396815089051e-05,
+      "loss": 0.9511,
+      "step": 14679
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0140270221435756e-05,
+      "loss": 1.075,
+      "step": 14680
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0139143626001782e-05,
+      "loss": 1.0138,
+      "step": 14681
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0138017028801426e-05,
+      "loss": 0.9186,
+      "step": 14682
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0136890429848994e-05,
+      "loss": 1.0283,
+      "step": 14683
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0135763829158784e-05,
+      "loss": 1.0374,
+      "step": 14684
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0134637226745101e-05,
+      "loss": 1.0795,
+      "step": 14685
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0133510622622245e-05,
+      "loss": 0.9113,
+      "step": 14686
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0132384016804518e-05,
+      "loss": 1.0145,
+      "step": 14687
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0131257409306223e-05,
+      "loss": 0.9902,
+      "step": 14688
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.013013080014166e-05,
+      "loss": 1.1006,
+      "step": 14689
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0129004189325133e-05,
+      "loss": 0.9668,
+      "step": 14690
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0127877576870944e-05,
+      "loss": 1.0467,
+      "step": 14691
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.012675096279339e-05,
+      "loss": 0.9829,
+      "step": 14692
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0125624347106779e-05,
+      "loss": 0.8726,
+      "step": 14693
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0124497729825412e-05,
+      "loss": 1.0169,
+      "step": 14694
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0123371110963589e-05,
+      "loss": 1.0062,
+      "step": 14695
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0122244490535612e-05,
+      "loss": 0.9621,
+      "step": 14696
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0121117868555785e-05,
+      "loss": 1.021,
+      "step": 14697
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011999124503841e-05,
+      "loss": 0.9576,
+      "step": 14698
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0118864619997786e-05,
+      "loss": 1.0136,
+      "step": 14699
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0117737993448217e-05,
+      "loss": 0.9179,
+      "step": 14700
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0116611365404005e-05,
+      "loss": 1.0366,
+      "step": 14701
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0115484735879456e-05,
+      "loss": 1.0132,
+      "step": 14702
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0114358104888867e-05,
+      "loss": 0.9236,
+      "step": 14703
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0113231472446538e-05,
+      "loss": 1.0979,
+      "step": 14704
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.011210483856678e-05,
+      "loss": 0.9792,
+      "step": 14705
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0110978203263889e-05,
+      "loss": 0.9897,
+      "step": 14706
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0109851566552167e-05,
+      "loss": 0.9358,
+      "step": 14707
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0108724928445918e-05,
+      "loss": 0.8804,
+      "step": 14708
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0107598288959446e-05,
+      "loss": 0.9149,
+      "step": 14709
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010647164810705e-05,
+      "loss": 1.0557,
+      "step": 14710
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0105345005903033e-05,
+      "loss": 0.9586,
+      "step": 14711
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.01042183623617e-05,
+      "loss": 1.0818,
+      "step": 14712
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.010309171749735e-05,
+      "loss": 0.9132,
+      "step": 14713
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0101965071324287e-05,
+      "loss": 0.9835,
+      "step": 14714
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0100838423856812e-05,
+      "loss": 0.9693,
+      "step": 14715
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009971177510923e-05,
+      "loss": 0.938,
+      "step": 14716
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.009858512509584e-05,
+      "loss": 0.9862,
+      "step": 14717
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0097458473830949e-05,
+      "loss": 0.9922,
+      "step": 14718
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0096331821328855e-05,
+      "loss": 1.0579,
+      "step": 14719
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0095205167603861e-05,
+      "loss": 1.0154,
+      "step": 14720
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0094078512670271e-05,
+      "loss": 1.0286,
+      "step": 14721
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0092951856542389e-05,
+      "loss": 1.0859,
+      "step": 14722
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0091825199234515e-05,
+      "loss": 0.9237,
+      "step": 14723
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0090698540760952e-05,
+      "loss": 1.0433,
+      "step": 14724
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0089571881136003e-05,
+      "loss": 0.9534,
+      "step": 14725
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0088445220373971e-05,
+      "loss": 0.9935,
+      "step": 14726
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0087318558489156e-05,
+      "loss": 1.0387,
+      "step": 14727
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0086191895495863e-05,
+      "loss": 1.0206,
+      "step": 14728
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0085065231408394e-05,
+      "loss": 1.0671,
+      "step": 14729
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0083938566241053e-05,
+      "loss": 1.026,
+      "step": 14730
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.008281190000814e-05,
+      "loss": 1.1183,
+      "step": 14731
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0081685232723957e-05,
+      "loss": 1.0638,
+      "step": 14732
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0080558564402813e-05,
+      "loss": 1.0788,
+      "step": 14733
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0079431895058999e-05,
+      "loss": 0.9639,
+      "step": 14734
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0078305224706833e-05,
+      "loss": 1.0764,
+      "step": 14735
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0077178553360604e-05,
+      "loss": 0.9532,
+      "step": 14736
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0076051881034626e-05,
+      "loss": 0.8421,
+      "step": 14737
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0074925207743191e-05,
+      "loss": 1.0336,
+      "step": 14738
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0073798533500608e-05,
+      "loss": 1.0467,
+      "step": 14739
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0072671858321179e-05,
+      "loss": 0.9843,
+      "step": 14740
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0071545182219203e-05,
+      "loss": 0.9181,
+      "step": 14741
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.007041850520899e-05,
+      "loss": 0.8763,
+      "step": 14742
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0069291827304837e-05,
+      "loss": 1.0153,
+      "step": 14743
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.006816514852105e-05,
+      "loss": 0.9805,
+      "step": 14744
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0067038468871928e-05,
+      "loss": 0.9874,
+      "step": 14745
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0065911788371777e-05,
+      "loss": 0.9583,
+      "step": 14746
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00647851070349e-05,
+      "loss": 0.936,
+      "step": 14747
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0063658424875596e-05,
+      "loss": 0.9495,
+      "step": 14748
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0062531741908173e-05,
+      "loss": 1.0344,
+      "step": 14749
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0061405058146928e-05,
+      "loss": 1.191,
+      "step": 14750
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0060278373606171e-05,
+      "loss": 0.8597,
+      "step": 14751
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0059151688300202e-05,
+      "loss": 1.0205,
+      "step": 14752
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0058025002243319e-05,
+      "loss": 1.0991,
+      "step": 14753
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0056898315449833e-05,
+      "loss": 1.1086,
+      "step": 14754
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0055771627934039e-05,
+      "loss": 0.9917,
+      "step": 14755
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0054644939710245e-05,
+      "loss": 1.0708,
+      "step": 14756
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0053518250792753e-05,
+      "loss": 1.0267,
+      "step": 14757
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0052391561195864e-05,
+      "loss": 1.0215,
+      "step": 14758
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0051264870933886e-05,
+      "loss": 1.0238,
+      "step": 14759
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0050138180021116e-05,
+      "loss": 1.0289,
+      "step": 14760
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0049011488471861e-05,
+      "loss": 0.9422,
+      "step": 14761
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0047884796300417e-05,
+      "loss": 1.0556,
+      "step": 14762
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.00467581035211e-05,
+      "loss": 1.0948,
+      "step": 14763
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0045631410148197e-05,
+      "loss": 0.8508,
+      "step": 14764
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0044504716196024e-05,
+      "loss": 0.994,
+      "step": 14765
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0043378021678881e-05,
+      "loss": 1.1034,
+      "step": 14766
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0042251326611066e-05,
+      "loss": 0.9456,
+      "step": 14767
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0041124631006886e-05,
+      "loss": 1.0651,
+      "step": 14768
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.003999793488064e-05,
+      "loss": 1.041,
+      "step": 14769
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0038871238246641e-05,
+      "loss": 1.053,
+      "step": 14770
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0037744541119181e-05,
+      "loss": 0.9838,
+      "step": 14771
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0036617843512566e-05,
+      "loss": 0.952,
+      "step": 14772
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0035491145441105e-05,
+      "loss": 1.0371,
+      "step": 14773
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0034364446919091e-05,
+      "loss": 1.0358,
+      "step": 14774
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0033237747960835e-05,
+      "loss": 1.0006,
+      "step": 14775
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0032111048580638e-05,
+      "loss": 0.9792,
+      "step": 14776
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0030984348792803e-05,
+      "loss": 0.9539,
+      "step": 14777
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002985764861163e-05,
+      "loss": 0.9283,
+      "step": 14778
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0028730948051426e-05,
+      "loss": 1.0685,
+      "step": 14779
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0027604247126492e-05,
+      "loss": 0.8135,
+      "step": 14780
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0026477545851132e-05,
+      "loss": 1.0206,
+      "step": 14781
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.002535084423965e-05,
+      "loss": 0.8987,
+      "step": 14782
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0024224142306345e-05,
+      "loss": 0.9428,
+      "step": 14783
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0023097440065526e-05,
+      "loss": 0.9345,
+      "step": 14784
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0021970737531492e-05,
+      "loss": 0.9896,
+      "step": 14785
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0020844034718547e-05,
+      "loss": 0.9577,
+      "step": 14786
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0019717331640995e-05,
+      "loss": 1.0283,
+      "step": 14787
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0018590628313137e-05,
+      "loss": 1.0342,
+      "step": 14788
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0017463924749278e-05,
+      "loss": 1.0199,
+      "step": 14789
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0016337220963719e-05,
+      "loss": 1.0268,
+      "step": 14790
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0015210516970768e-05,
+      "loss": 1.0923,
+      "step": 14791
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0014083812784723e-05,
+      "loss": 1.0244,
+      "step": 14792
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001295710841989e-05,
+      "loss": 0.9913,
+      "step": 14793
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.001183040389057e-05,
+      "loss": 1.0085,
+      "step": 14794
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0010703699211068e-05,
+      "loss": 1.0361,
+      "step": 14795
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0009576994395685e-05,
+      "loss": 1.0362,
+      "step": 14796
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0008450289458725e-05,
+      "loss": 0.9798,
+      "step": 14797
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0007323584414495e-05,
+      "loss": 1.0037,
+      "step": 14798
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0006196879277291e-05,
+      "loss": 1.0441,
+      "step": 14799
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000507017406142e-05,
+      "loss": 1.0374,
+      "step": 14800
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0003943468781188e-05,
+      "loss": 1.0772,
+      "step": 14801
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.0002816763450891e-05,
+      "loss": 1.0577,
+      "step": 14802
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000169005808484e-05,
+      "loss": 0.9937,
+      "step": 14803
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 1.000056335269733e-05,
+      "loss": 1.1203,
+      "step": 14804
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.999436647302671e-06,
+      "loss": 0.9921,
+      "step": 14805
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.998309941915164e-06,
+      "loss": 0.9874,
+      "step": 14806
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.99718323654911e-06,
+      "loss": 1.0991,
+      "step": 14807
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.996056531218817e-06,
+      "loss": 0.9061,
+      "step": 14808
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.994929825938584e-06,
+      "loss": 1.0252,
+      "step": 14809
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.993803120722712e-06,
+      "loss": 1.0278,
+      "step": 14810
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.992676415585509e-06,
+      "loss": 0.8315,
+      "step": 14811
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.991549710541279e-06,
+      "loss": 0.8618,
+      "step": 14812
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.990423005604319e-06,
+      "loss": 1.1279,
+      "step": 14813
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.989296300788933e-06,
+      "loss": 1.0343,
+      "step": 14814
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.988169596109432e-06,
+      "loss": 1.0791,
+      "step": 14815
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.987042891580113e-06,
+      "loss": 0.9618,
+      "step": 14816
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98591618721528e-06,
+      "loss": 0.9808,
+      "step": 14817
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.984789483029235e-06,
+      "loss": 0.8788,
+      "step": 14818
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.983662779036281e-06,
+      "loss": 1.0824,
+      "step": 14819
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.982536075250725e-06,
+      "loss": 0.9754,
+      "step": 14820
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.981409371686865e-06,
+      "loss": 1.0708,
+      "step": 14821
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.98028266835901e-06,
+      "loss": 1.0412,
+      "step": 14822
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.979155965281457e-06,
+      "loss": 0.8933,
+      "step": 14823
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97802926246851e-06,
+      "loss": 0.9953,
+      "step": 14824
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.976902559934475e-06,
+      "loss": 1.1148,
+      "step": 14825
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.975775857693658e-06,
+      "loss": 0.9778,
+      "step": 14826
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.974649155760353e-06,
+      "loss": 1.0228,
+      "step": 14827
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.973522454148868e-06,
+      "loss": 0.8487,
+      "step": 14828
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.97239575287351e-06,
+      "loss": 1.1555,
+      "step": 14829
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.971269051948577e-06,
+      "loss": 1.0021,
+      "step": 14830
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.970142351388374e-06,
+      "loss": 1.058,
+      "step": 14831
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.969015651207202e-06,
+      "loss": 1.0591,
+      "step": 14832
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.967888951419364e-06,
+      "loss": 1.0937,
+      "step": 14833
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.966762252039167e-06,
+      "loss": 1.1227,
+      "step": 14834
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.96563555308091e-06,
+      "loss": 1.0938,
+      "step": 14835
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.9645088545589e-06,
+      "loss": 0.9346,
+      "step": 14836
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.963382156487436e-06,
+      "loss": 0.9568,
+      "step": 14837
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.962255458880819e-06,
+      "loss": 1.0271,
+      "step": 14838
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.961128761753362e-06,
+      "loss": 1.0754,
+      "step": 14839
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.960002065119361e-06,
+      "loss": 1.0335,
+      "step": 14840
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.958875368993117e-06,
+      "loss": 1.0578,
+      "step": 14841
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.957748673388936e-06,
+      "loss": 1.0391,
+      "step": 14842
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.95662197832112e-06,
+      "loss": 1.0023,
+      "step": 14843
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.955495283803979e-06,
+      "loss": 1.1278,
+      "step": 14844
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.954368589851805e-06,
+      "loss": 0.9888,
+      "step": 14845
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.953241896478908e-06,
+      "loss": 0.9391,
+      "step": 14846
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.952115203699582e-06,
+      "loss": 0.9729,
+      "step": 14847
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.950988511528142e-06,
+      "loss": 1.0328,
+      "step": 14848
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.949861819978887e-06,
+      "loss": 1.0487,
+      "step": 14849
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94873512906612e-06,
+      "loss": 1.0062,
+      "step": 14850
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.947608438804139e-06,
+      "loss": 1.0243,
+      "step": 14851
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.946481749207247e-06,
+      "loss": 1.0019,
+      "step": 14852
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.945355060289757e-06,
+      "loss": 1.0357,
+      "step": 14853
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.944228372065965e-06,
+      "loss": 1.1352,
+      "step": 14854
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.943101684550172e-06,
+      "loss": 0.9106,
+      "step": 14855
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.94197499775668e-06,
+      "loss": 1.0107,
+      "step": 14856
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.940848311699801e-06,
+      "loss": 0.9682,
+      "step": 14857
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.939721626393832e-06,
+      "loss": 1.1397,
+      "step": 14858
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.938594941853073e-06,
+      "loss": 0.9282,
+      "step": 14859
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.937468258091832e-06,
+      "loss": 1.086,
+      "step": 14860
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.936341575124406e-06,
+      "loss": 1.0215,
+      "step": 14861
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.935214892965102e-06,
+      "loss": 1.0293,
+      "step": 14862
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.934088211628225e-06,
+      "loss": 0.9807,
+      "step": 14863
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.932961531128077e-06,
+      "loss": 1.016,
+      "step": 14864
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.931834851478954e-06,
+      "loss": 0.9967,
+      "step": 14865
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.930708172695164e-06,
+      "loss": 0.9482,
+      "step": 14866
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.929581494791012e-06,
+      "loss": 0.8933,
+      "step": 14867
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.928454817780799e-06,
+      "loss": 0.941,
+      "step": 14868
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.927328141678826e-06,
+      "loss": 0.9994,
+      "step": 14869
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.926201466499394e-06,
+      "loss": 1.0045,
+      "step": 14870
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.925074792256812e-06,
+      "loss": 0.9208,
+      "step": 14871
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.923948118965377e-06,
+      "loss": 0.9554,
+      "step": 14872
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.922821446639397e-06,
+      "loss": 0.9707,
+      "step": 14873
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.921694775293172e-06,
+      "loss": 0.8507,
+      "step": 14874
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.920568104941e-06,
+      "loss": 0.9947,
+      "step": 14875
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91944143559719e-06,
+      "loss": 0.8507,
+      "step": 14876
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.918314767276045e-06,
+      "loss": 0.9765,
+      "step": 14877
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.917188099991866e-06,
+      "loss": 1.0595,
+      "step": 14878
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91606143375895e-06,
+      "loss": 0.8951,
+      "step": 14879
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.914934768591608e-06,
+      "loss": 1.0131,
+      "step": 14880
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.91380810450414e-06,
+      "loss": 0.983,
+      "step": 14881
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.912681441510847e-06,
+      "loss": 0.8732,
+      "step": 14882
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.911554779626034e-06,
+      "loss": 0.9704,
+      "step": 14883
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.910428118863999e-06,
+      "loss": 1.0162,
+      "step": 14884
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.909301459239051e-06,
+      "loss": 0.9382,
+      "step": 14885
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.908174800765487e-06,
+      "loss": 1.038,
+      "step": 14886
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.907048143457615e-06,
+      "loss": 1.1123,
+      "step": 14887
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.905921487329734e-06,
+      "loss": 1.0918,
+      "step": 14888
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.904794832396142e-06,
+      "loss": 0.9886,
+      "step": 14889
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.903668178671148e-06,
+      "loss": 0.9879,
+      "step": 14890
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.902541526169056e-06,
+      "loss": 0.9933,
+      "step": 14891
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.901414874904165e-06,
+      "loss": 0.987,
+      "step": 14892
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.900288224890772e-06,
+      "loss": 1.0288,
+      "step": 14893
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.89916157614319e-06,
+      "loss": 1.0342,
+      "step": 14894
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.898034928675718e-06,
+      "loss": 1.0863,
+      "step": 14895
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.896908282502652e-06,
+      "loss": 0.9813,
+      "step": 14896
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.895781637638304e-06,
+      "loss": 0.9311,
+      "step": 14897
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.894654994096967e-06,
+      "loss": 1.0263,
+      "step": 14898
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.893528351892953e-06,
+      "loss": 1.1442,
+      "step": 14899
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.892401711040557e-06,
+      "loss": 0.9649,
+      "step": 14900
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.891275071554085e-06,
+      "loss": 1.0173,
+      "step": 14901
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.890148433447838e-06,
+      "loss": 0.9969,
+      "step": 14902
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.889021796736114e-06,
+      "loss": 0.9636,
+      "step": 14903
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.887895161433221e-06,
+      "loss": 0.9903,
+      "step": 14904
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.886768527553464e-06,
+      "loss": 0.9039,
+      "step": 14905
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.885641895111138e-06,
+      "loss": 1.0682,
+      "step": 14906
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.884515264120546e-06,
+      "loss": 1.0501,
+      "step": 14907
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.883388634595995e-06,
+      "loss": 1.0243,
+      "step": 14908
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.882262006551786e-06,
+      "loss": 0.9935,
+      "step": 14909
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.881135380002218e-06,
+      "loss": 1.0346,
+      "step": 14910
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.880008754961595e-06,
+      "loss": 1.0779,
+      "step": 14911
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.878882131444216e-06,
+      "loss": 0.9358,
+      "step": 14912
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.877755509464391e-06,
+      "loss": 1.1397,
+      "step": 14913
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.876628889036414e-06,
+      "loss": 0.9977,
+      "step": 14914
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.875502270174592e-06,
+      "loss": 1.0868,
+      "step": 14915
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.874375652893226e-06,
+      "loss": 1.0085,
+      "step": 14916
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.873249037206611e-06,
+      "loss": 1.0061,
+      "step": 14917
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87212242312906e-06,
+      "loss": 1.0588,
+      "step": 14918
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.87099581067487e-06,
+      "loss": 0.9094,
+      "step": 14919
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.869869199858342e-06,
+      "loss": 1.0369,
+      "step": 14920
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.868742590693776e-06,
+      "loss": 0.9207,
+      "step": 14921
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.867615983195484e-06,
+      "loss": 0.9902,
+      "step": 14922
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.866489377377758e-06,
+      "loss": 1.0357,
+      "step": 14923
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.865362773254902e-06,
+      "loss": 0.9731,
+      "step": 14924
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.864236170841219e-06,
+      "loss": 1.0293,
+      "step": 14925
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.863109570151008e-06,
+      "loss": 1.1718,
+      "step": 14926
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.861982971198576e-06,
+      "loss": 0.9522,
+      "step": 14927
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.860856373998221e-06,
+      "loss": 0.9334,
+      "step": 14928
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.859729778564246e-06,
+      "loss": 0.9665,
+      "step": 14929
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.858603184910952e-06,
+      "loss": 0.9937,
+      "step": 14930
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.857476593052642e-06,
+      "loss": 0.9646,
+      "step": 14931
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.856350003003614e-06,
+      "loss": 0.8971,
+      "step": 14932
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.855223414778176e-06,
+      "loss": 1.0045,
+      "step": 14933
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.854096828390625e-06,
+      "loss": 0.9922,
+      "step": 14934
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.852970243855261e-06,
+      "loss": 0.9101,
+      "step": 14935
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.851843661186392e-06,
+      "loss": 0.9983,
+      "step": 14936
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.850717080398316e-06,
+      "loss": 1.0943,
+      "step": 14937
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.849590501505332e-06,
+      "loss": 0.9341,
+      "step": 14938
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.848463924521748e-06,
+      "loss": 1.0429,
+      "step": 14939
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.847337349461856e-06,
+      "loss": 0.9178,
+      "step": 14940
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.846210776339967e-06,
+      "loss": 1.0499,
+      "step": 14941
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.845084205170378e-06,
+      "loss": 1.1377,
+      "step": 14942
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84395763596739e-06,
+      "loss": 1.132,
+      "step": 14943
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.842831068745306e-06,
+      "loss": 0.8464,
+      "step": 14944
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.841704503518424e-06,
+      "loss": 1.0204,
+      "step": 14945
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.84057794030105e-06,
+      "loss": 0.9091,
+      "step": 14946
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.839451379107485e-06,
+      "loss": 1.017,
+      "step": 14947
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.838324819952026e-06,
+      "loss": 0.8346,
+      "step": 14948
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.837198262848976e-06,
+      "loss": 1.0122,
+      "step": 14949
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83607170781264e-06,
+      "loss": 0.9783,
+      "step": 14950
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.834945154857317e-06,
+      "loss": 1.1047,
+      "step": 14951
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.833818603997306e-06,
+      "loss": 1.0473,
+      "step": 14952
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.83269205524691e-06,
+      "loss": 0.9649,
+      "step": 14953
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.831565508620427e-06,
+      "loss": 0.8928,
+      "step": 14954
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.830438964132164e-06,
+      "loss": 0.9746,
+      "step": 14955
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82931242179642e-06,
+      "loss": 1.0552,
+      "step": 14956
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.828185881627495e-06,
+      "loss": 0.9961,
+      "step": 14957
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.827059343639689e-06,
+      "loss": 0.993,
+      "step": 14958
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8259328078473e-06,
+      "loss": 0.9221,
+      "step": 14959
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.82480627426464e-06,
+      "loss": 0.971,
+      "step": 14960
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.823679742906001e-06,
+      "loss": 1.0689,
+      "step": 14961
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.822553213785686e-06,
+      "loss": 0.8945,
+      "step": 14962
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.821426686917992e-06,
+      "loss": 0.9746,
+      "step": 14963
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.820300162317228e-06,
+      "loss": 0.9759,
+      "step": 14964
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.819173639997692e-06,
+      "loss": 0.9415,
+      "step": 14965
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.818047119973682e-06,
+      "loss": 1.0305,
+      "step": 14966
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.8169206022595e-06,
+      "loss": 0.9389,
+      "step": 14967
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.815794086869443e-06,
+      "loss": 0.9668,
+      "step": 14968
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.814667573817819e-06,
+      "loss": 1.0836,
+      "step": 14969
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.813541063118927e-06,
+      "loss": 0.9127,
+      "step": 14970
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.812414554787066e-06,
+      "loss": 0.9674,
+      "step": 14971
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.811288048836534e-06,
+      "loss": 1.0292,
+      "step": 14972
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.810161545281634e-06,
+      "loss": 0.8357,
+      "step": 14973
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.809035044136669e-06,
+      "loss": 1.0631,
+      "step": 14974
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.807908545415938e-06,
+      "loss": 1.0289,
+      "step": 14975
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80678204913374e-06,
+      "loss": 0.9825,
+      "step": 14976
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.805655555304372e-06,
+      "loss": 1.0192,
+      "step": 14977
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.804529063942145e-06,
+      "loss": 1.018,
+      "step": 14978
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.80340257506135e-06,
+      "loss": 0.9572,
+      "step": 14979
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.802276088676293e-06,
+      "loss": 0.9986,
+      "step": 14980
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.801149604801273e-06,
+      "loss": 1.0644,
+      "step": 14981
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.800023123450582e-06,
+      "loss": 1.0355,
+      "step": 14982
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.798896644638532e-06,
+      "loss": 0.9988,
+      "step": 14983
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79777016837942e-06,
+      "loss": 0.9493,
+      "step": 14984
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.796643694687546e-06,
+      "loss": 0.925,
+      "step": 14985
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.795517223577207e-06,
+      "loss": 1.0228,
+      "step": 14986
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.794390755062703e-06,
+      "loss": 1.08,
+      "step": 14987
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.79326428915834e-06,
+      "loss": 0.908,
+      "step": 14988
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.792137825878415e-06,
+      "loss": 0.9073,
+      "step": 14989
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.791011365237226e-06,
+      "loss": 0.9845,
+      "step": 14990
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.789884907249072e-06,
+      "loss": 1.0324,
+      "step": 14991
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.788758451928261e-06,
+      "loss": 0.9068,
+      "step": 14992
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.787631999289084e-06,
+      "loss": 0.9453,
+      "step": 14993
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.786505549345848e-06,
+      "loss": 0.991,
+      "step": 14994
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.785379102112848e-06,
+      "loss": 0.9705,
+      "step": 14995
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.784252657604382e-06,
+      "loss": 1.0845,
+      "step": 14996
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.783126215834755e-06,
+      "loss": 0.9449,
+      "step": 14997
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.781999776818267e-06,
+      "loss": 0.9506,
+      "step": 14998
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.780873340569216e-06,
+      "loss": 1.0803,
+      "step": 14999
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7797469071019e-06,
+      "loss": 1.0143,
+      "step": 15000
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.778620476430616e-06,
+      "loss": 1.0589,
+      "step": 15001
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.777494048569672e-06,
+      "loss": 1.115,
+      "step": 15002
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.776367623533363e-06,
+      "loss": 1.0328,
+      "step": 15003
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.775241201335989e-06,
+      "loss": 0.9988,
+      "step": 15004
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.774114781991845e-06,
+      "loss": 0.9958,
+      "step": 15005
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.772988365515242e-06,
+      "loss": 1.0328,
+      "step": 15006
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.771861951920467e-06,
+      "loss": 1.122,
+      "step": 15007
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.770735541221829e-06,
+      "loss": 1.0681,
+      "step": 15008
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.769609133433622e-06,
+      "loss": 1.0194,
+      "step": 15009
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.768482728570142e-06,
+      "loss": 1.0798,
+      "step": 15010
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.767356326645697e-06,
+      "loss": 1.0448,
+      "step": 15011
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.766229927674583e-06,
+      "loss": 0.9397,
+      "step": 15012
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.765103531671096e-06,
+      "loss": 1.0868,
+      "step": 15013
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.76397713864954e-06,
+      "loss": 1.0579,
+      "step": 15014
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.762850748624206e-06,
+      "loss": 0.9579,
+      "step": 15015
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.761724361609404e-06,
+      "loss": 0.9265,
+      "step": 15016
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.760597977619427e-06,
+      "loss": 1.1208,
+      "step": 15017
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.759471596668576e-06,
+      "loss": 0.9683,
+      "step": 15018
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.758345218771145e-06,
+      "loss": 1.0974,
+      "step": 15019
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.757218843941441e-06,
+      "loss": 1.0268,
+      "step": 15020
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.756092472193757e-06,
+      "loss": 1.0583,
+      "step": 15021
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.754966103542398e-06,
+      "loss": 1.0509,
+      "step": 15022
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.753839738001656e-06,
+      "loss": 0.9702,
+      "step": 15023
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.752713375585828e-06,
+      "loss": 1.1871,
+      "step": 15024
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.75158701630922e-06,
+      "loss": 1.0189,
+      "step": 15025
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.750460660186132e-06,
+      "loss": 1.0023,
+      "step": 15026
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.749334307230854e-06,
+      "loss": 1.0447,
+      "step": 15027
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.748207957457688e-06,
+      "loss": 0.8813,
+      "step": 15028
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.747081610880939e-06,
+      "loss": 0.9557,
+      "step": 15029
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.7459552675149e-06,
+      "loss": 0.9491,
+      "step": 15030
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.744828927373868e-06,
+      "loss": 1.0665,
+      "step": 15031
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.743702590472146e-06,
+      "loss": 1.1964,
+      "step": 15032
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.742576256824025e-06,
+      "loss": 1.0745,
+      "step": 15033
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.741449926443813e-06,
+      "loss": 0.9639,
+      "step": 15034
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.740323599345802e-06,
+      "loss": 0.9766,
+      "step": 15035
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.739197275544293e-06,
+      "loss": 1.0995,
+      "step": 15036
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.738070955053584e-06,
+      "loss": 0.8786,
+      "step": 15037
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.73694463788797e-06,
+      "loss": 0.9499,
+      "step": 15038
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.735818324061753e-06,
+      "loss": 0.9625,
+      "step": 15039
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.734692013589232e-06,
+      "loss": 0.9398,
+      "step": 15040
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.733565706484703e-06,
+      "loss": 0.9739,
+      "step": 15041
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.732439402762459e-06,
+      "loss": 1.0775,
+      "step": 15042
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.731313102436809e-06,
+      "loss": 1.0123,
+      "step": 15043
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.730186805522045e-06,
+      "loss": 0.8992,
+      "step": 15044
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.729060512032464e-06,
+      "loss": 1.0022,
+      "step": 15045
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.727934221982365e-06,
+      "loss": 1.0191,
+      "step": 15046
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.726807935386043e-06,
+      "loss": 0.9211,
+      "step": 15047
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.725681652257803e-06,
+      "loss": 0.8895,
+      "step": 15048
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.724555372611937e-06,
+      "loss": 1.0741,
+      "step": 15049
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.723429096462745e-06,
+      "loss": 0.935,
+      "step": 15050
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.722302823824523e-06,
+      "loss": 0.9181,
+      "step": 15051
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.721176554711567e-06,
+      "loss": 0.8222,
+      "step": 15052
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.720050289138181e-06,
+      "loss": 0.9728,
+      "step": 15053
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.71892402711866e-06,
+      "loss": 0.9439,
+      "step": 15054
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.717797768667298e-06,
+      "loss": 1.0077,
+      "step": 15055
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.716671513798392e-06,
+      "loss": 1.0841,
+      "step": 15056
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.715545262526244e-06,
+      "loss": 0.98,
+      "step": 15057
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.714419014865152e-06,
+      "loss": 1.0201,
+      "step": 15058
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.713292770829408e-06,
+      "loss": 0.9127,
+      "step": 15059
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.712166530433316e-06,
+      "loss": 0.9052,
+      "step": 15060
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.711040293691165e-06,
+      "loss": 1.0476,
+      "step": 15061
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.709914060617259e-06,
+      "loss": 0.9782,
+      "step": 15062
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70878783122589e-06,
+      "loss": 0.9398,
+      "step": 15063
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70766160553136e-06,
+      "loss": 1.0248,
+      "step": 15064
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.706535383547962e-06,
+      "loss": 1.0614,
+      "step": 15065
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.705409165289992e-06,
+      "loss": 1.0528,
+      "step": 15066
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.704282950771756e-06,
+      "loss": 0.9664,
+      "step": 15067
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.703156740007543e-06,
+      "loss": 1.063,
+      "step": 15068
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.70203053301165e-06,
+      "loss": 0.9278,
+      "step": 15069
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.700904329798373e-06,
+      "loss": 1.0747,
+      "step": 15070
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.699778130382014e-06,
+      "loss": 1.0539,
+      "step": 15071
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.698651934776868e-06,
+      "loss": 0.9832,
+      "step": 15072
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69752574299723e-06,
+      "loss": 1.0212,
+      "step": 15073
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.696399555057395e-06,
+      "loss": 1.0792,
+      "step": 15074
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.69527337097166e-06,
+      "loss": 1.0313,
+      "step": 15075
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.694147190754325e-06,
+      "loss": 0.8235,
+      "step": 15076
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.693021014419685e-06,
+      "loss": 1.0598,
+      "step": 15077
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.691894841982039e-06,
+      "loss": 0.936,
+      "step": 15078
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.690768673455676e-06,
+      "loss": 1.1104,
+      "step": 15079
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.689642508854893e-06,
+      "loss": 0.9748,
+      "step": 15080
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.688516348193995e-06,
+      "loss": 1.0042,
+      "step": 15081
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.687390191487275e-06,
+      "loss": 0.8323,
+      "step": 15082
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.686264038749024e-06,
+      "loss": 1.0406,
+      "step": 15083
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68513788999354e-06,
+      "loss": 0.9897,
+      "step": 15084
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.684011745235124e-06,
+      "loss": 1.0684,
+      "step": 15085
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.682885604488065e-06,
+      "loss": 0.9938,
+      "step": 15086
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.681759467766666e-06,
+      "loss": 0.9833,
+      "step": 15087
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.68063333508522e-06,
+      "loss": 0.9928,
+      "step": 15088
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.679507206458016e-06,
+      "loss": 1.0034,
+      "step": 15089
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.67838108189936e-06,
+      "loss": 0.8424,
+      "step": 15090
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.677254961423544e-06,
+      "loss": 0.9876,
+      "step": 15091
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.676128845044865e-06,
+      "loss": 0.9002,
+      "step": 15092
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.675002732777613e-06,
+      "loss": 1.11,
+      "step": 15093
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.673876624636088e-06,
+      "loss": 0.9181,
+      "step": 15094
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.672750520634588e-06,
+      "loss": 0.9861,
+      "step": 15095
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.671624420787406e-06,
+      "loss": 1.0549,
+      "step": 15096
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.670498325108836e-06,
+      "loss": 0.9323,
+      "step": 15097
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.669372233613172e-06,
+      "loss": 0.9263,
+      "step": 15098
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.668246146314715e-06,
+      "loss": 0.8838,
+      "step": 15099
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.667120063227758e-06,
+      "loss": 0.94,
+      "step": 15100
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.665993984366595e-06,
+      "loss": 1.0388,
+      "step": 15101
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.664867909745523e-06,
+      "loss": 0.9546,
+      "step": 15102
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.66374183937883e-06,
+      "loss": 0.9705,
+      "step": 15103
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.662615773280818e-06,
+      "loss": 1.152,
+      "step": 15104
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.661489711465785e-06,
+      "loss": 1.0443,
+      "step": 15105
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.660363653948023e-06,
+      "loss": 0.9638,
+      "step": 15106
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.659237600741823e-06,
+      "loss": 1.0932,
+      "step": 15107
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.658111551861477e-06,
+      "loss": 0.9471,
+      "step": 15108
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.656985507321293e-06,
+      "loss": 0.9056,
+      "step": 15109
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.655859467135555e-06,
+      "loss": 1.0185,
+      "step": 15110
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.654733431318563e-06,
+      "loss": 0.9218,
+      "step": 15111
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.653607399884604e-06,
+      "loss": 1.0184,
+      "step": 15112
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.652481372847985e-06,
+      "loss": 0.9538,
+      "step": 15113
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.65135535022299e-06,
+      "loss": 1.1034,
+      "step": 15114
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.650229332023919e-06,
+      "loss": 0.9753,
+      "step": 15115
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.649103318265063e-06,
+      "loss": 1.1091,
+      "step": 15116
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.647977308960714e-06,
+      "loss": 0.9093,
+      "step": 15117
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.646851304125174e-06,
+      "loss": 0.9491,
+      "step": 15118
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.645725303772734e-06,
+      "loss": 0.96,
+      "step": 15119
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.644599307917687e-06,
+      "loss": 0.9949,
+      "step": 15120
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.643473316574327e-06,
+      "loss": 0.9598,
+      "step": 15121
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.642347329756944e-06,
+      "loss": 1.0084,
+      "step": 15122
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.641221347479843e-06,
+      "loss": 0.973,
+      "step": 15123
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.640095369757309e-06,
+      "loss": 0.8939,
+      "step": 15124
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.63896939660364e-06,
+      "loss": 1.0907,
+      "step": 15125
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.637843428033125e-06,
+      "loss": 1.0431,
+      "step": 15126
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.636717464060065e-06,
+      "loss": 0.9755,
+      "step": 15127
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.635591504698747e-06,
+      "loss": 0.9595,
+      "step": 15128
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.634465549963471e-06,
+      "loss": 0.9447,
+      "step": 15129
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.633339599868525e-06,
+      "loss": 0.918,
+      "step": 15130
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.632213654428202e-06,
+      "loss": 0.9532,
+      "step": 15131
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.631087713656801e-06,
+      "loss": 1.0194,
+      "step": 15132
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.629961777568613e-06,
+      "loss": 0.8793,
+      "step": 15133
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62883584617793e-06,
+      "loss": 1.0347,
+      "step": 15134
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.627709919499046e-06,
+      "loss": 1.0986,
+      "step": 15135
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.626583997546252e-06,
+      "loss": 0.9399,
+      "step": 15136
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.625458080333848e-06,
+      "loss": 1.0718,
+      "step": 15137
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.62433216787612e-06,
+      "loss": 0.9963,
+      "step": 15138
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.623206260187367e-06,
+      "loss": 0.8652,
+      "step": 15139
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.622080357281873e-06,
+      "loss": 1.0384,
+      "step": 15140
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.620954459173942e-06,
+      "loss": 0.9761,
+      "step": 15141
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.619828565877859e-06,
+      "loss": 0.9489,
+      "step": 15142
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.618702677407924e-06,
+      "loss": 1.129,
+      "step": 15143
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.617576793778419e-06,
+      "loss": 1.0382,
+      "step": 15144
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.616450915003645e-06,
+      "loss": 0.9327,
+      "step": 15145
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.615325041097892e-06,
+      "loss": 1.004,
+      "step": 15146
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.614199172075457e-06,
+      "loss": 0.9625,
+      "step": 15147
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.613073307950624e-06,
+      "loss": 0.9692,
+      "step": 15148
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.611947448737695e-06,
+      "loss": 0.9242,
+      "step": 15149
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.610821594450952e-06,
+      "loss": 0.9665,
+      "step": 15150
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.609695745104697e-06,
+      "loss": 0.8669,
+      "step": 15151
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.608569900713215e-06,
+      "loss": 0.9692,
+      "step": 15152
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.607444061290805e-06,
+      "loss": 0.9657,
+      "step": 15153
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60631822685175e-06,
+      "loss": 1.0082,
+      "step": 15154
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.605192397410353e-06,
+      "loss": 0.8193,
+      "step": 15155
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.604066572980898e-06,
+      "loss": 0.9305,
+      "step": 15156
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.60294075357768e-06,
+      "loss": 0.924,
+      "step": 15157
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.601814939214989e-06,
+      "loss": 0.9249,
+      "step": 15158
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.600689129907116e-06,
+      "loss": 0.9764,
+      "step": 15159
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.59956332566836e-06,
+      "loss": 0.998,
+      "step": 15160
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.598437526513007e-06,
+      "loss": 0.9465,
+      "step": 15161
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.597311732455348e-06,
+      "loss": 0.9688,
+      "step": 15162
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.596185943509673e-06,
+      "loss": 0.9605,
+      "step": 15163
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.595060159690281e-06,
+      "loss": 0.9546,
+      "step": 15164
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.593934381011459e-06,
+      "loss": 1.0021,
+      "step": 15165
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.592808607487494e-06,
+      "loss": 1.0108,
+      "step": 15166
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.591682839132685e-06,
+      "loss": 0.9924,
+      "step": 15167
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.590557075961315e-06,
+      "loss": 0.9234,
+      "step": 15168
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.589431317987684e-06,
+      "loss": 0.9827,
+      "step": 15169
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.588305565226078e-06,
+      "loss": 1.0525,
+      "step": 15170
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58717981769079e-06,
+      "loss": 1.0168,
+      "step": 15171
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.586054075396107e-06,
+      "loss": 1.0012,
+      "step": 15172
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.584928338356322e-06,
+      "loss": 1.0058,
+      "step": 15173
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58380260658573e-06,
+      "loss": 1.0574,
+      "step": 15174
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.58267688009862e-06,
+      "loss": 1.1161,
+      "step": 15175
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.581551158909279e-06,
+      "loss": 1.0464,
+      "step": 15176
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.580425443031995e-06,
+      "loss": 1.0602,
+      "step": 15177
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57929973248107e-06,
+      "loss": 1.0041,
+      "step": 15178
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.578174027270786e-06,
+      "loss": 0.8488,
+      "step": 15179
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.577048327415435e-06,
+      "loss": 0.9555,
+      "step": 15180
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57592263292931e-06,
+      "loss": 1.0586,
+      "step": 15181
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.574796943826693e-06,
+      "loss": 1.0027,
+      "step": 15182
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.573671260121884e-06,
+      "loss": 0.9658,
+      "step": 15183
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57254558182917e-06,
+      "loss": 0.9971,
+      "step": 15184
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.57141990896284e-06,
+      "loss": 1.065,
+      "step": 15185
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.570294241537183e-06,
+      "loss": 0.9153,
+      "step": 15186
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.569168579566487e-06,
+      "loss": 1.0749,
+      "step": 15187
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.568042923065053e-06,
+      "loss": 0.9973,
+      "step": 15188
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56691727204716e-06,
+      "loss": 1.0223,
+      "step": 15189
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5657916265271e-06,
+      "loss": 1.024,
+      "step": 15190
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.564665986519161e-06,
+      "loss": 0.9781,
+      "step": 15191
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56354035203764e-06,
+      "loss": 0.9954,
+      "step": 15192
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.56241472309682e-06,
+      "loss": 1.0558,
+      "step": 15193
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.561289099710993e-06,
+      "loss": 0.9413,
+      "step": 15194
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.560163481894448e-06,
+      "loss": 0.9037,
+      "step": 15195
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.559037869661469e-06,
+      "loss": 0.9417,
+      "step": 15196
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.557912263026353e-06,
+      "loss": 0.9844,
+      "step": 15197
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.556786662003389e-06,
+      "loss": 0.8664,
+      "step": 15198
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.555661066606863e-06,
+      "loss": 0.9528,
+      "step": 15199
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.55453547685106e-06,
+      "loss": 0.9876,
+      "step": 15200
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.553409892750275e-06,
+      "loss": 0.9292,
+      "step": 15201
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.552284314318797e-06,
+      "loss": 0.9363,
+      "step": 15202
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.551158741570914e-06,
+      "loss": 1.0805,
+      "step": 15203
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.550033174520912e-06,
+      "loss": 0.9119,
+      "step": 15204
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54890761318308e-06,
+      "loss": 0.9728,
+      "step": 15205
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54778205757171e-06,
+      "loss": 0.9794,
+      "step": 15206
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.546656507701089e-06,
+      "loss": 0.9276,
+      "step": 15207
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.545530963585507e-06,
+      "loss": 1.0122,
+      "step": 15208
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.544405425239249e-06,
+      "loss": 0.8845,
+      "step": 15209
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.543279892676603e-06,
+      "loss": 1.033,
+      "step": 15210
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.54215436591186e-06,
+      "loss": 0.9753,
+      "step": 15211
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.541028844959308e-06,
+      "loss": 1.0776,
+      "step": 15212
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.539903329833237e-06,
+      "loss": 0.8498,
+      "step": 15213
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.53877782054793e-06,
+      "loss": 1.0116,
+      "step": 15214
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.537652317117672e-06,
+      "loss": 0.9839,
+      "step": 15215
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.536526819556762e-06,
+      "loss": 0.9009,
+      "step": 15216
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.535401327879482e-06,
+      "loss": 0.9538,
+      "step": 15217
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.53427584210012e-06,
+      "loss": 0.9712,
+      "step": 15218
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.533150362232961e-06,
+      "loss": 0.9939,
+      "step": 15219
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.532024888292297e-06,
+      "loss": 0.928,
+      "step": 15220
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.530899420292412e-06,
+      "loss": 0.9982,
+      "step": 15221
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.529773958247598e-06,
+      "loss": 1.0798,
+      "step": 15222
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.528648502172138e-06,
+      "loss": 1.0538,
+      "step": 15223
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.527523052080317e-06,
+      "loss": 0.9304,
+      "step": 15224
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.526397607986427e-06,
+      "loss": 1.0788,
+      "step": 15225
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.525272169904757e-06,
+      "loss": 0.9601,
+      "step": 15226
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.52414673784959e-06,
+      "loss": 0.9292,
+      "step": 15227
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.523021311835213e-06,
+      "loss": 1.1871,
+      "step": 15228
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.521895891875909e-06,
+      "loss": 0.9972,
+      "step": 15229
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.520770477985976e-06,
+      "loss": 1.0922,
+      "step": 15230
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.519645070179691e-06,
+      "loss": 1.0356,
+      "step": 15231
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.518519668471347e-06,
+      "loss": 0.978,
+      "step": 15232
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.517394272875223e-06,
+      "loss": 1.0653,
+      "step": 15233
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.516268883405614e-06,
+      "loss": 0.9666,
+      "step": 15234
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.515143500076801e-06,
+      "loss": 0.9874,
+      "step": 15235
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.514018122903075e-06,
+      "loss": 0.86,
+      "step": 15236
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.512892751898718e-06,
+      "loss": 0.8983,
+      "step": 15237
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.51176738707801e-06,
+      "loss": 0.8571,
+      "step": 15238
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.510642028455248e-06,
+      "loss": 1.0138,
+      "step": 15239
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.509516676044718e-06,
+      "loss": 0.9619,
+      "step": 15240
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.5083913298607e-06,
+      "loss": 1.0936,
+      "step": 15241
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.507265989917484e-06,
+      "loss": 1.0964,
+      "step": 15242
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.506140656229352e-06,
+      "loss": 1.0869,
+      "step": 15243
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.505015328810593e-06,
+      "loss": 0.894,
+      "step": 15244
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.50389000767549e-06,
+      "loss": 0.9652,
+      "step": 15245
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.502764692838332e-06,
+      "loss": 1.0726,
+      "step": 15246
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.501639384313397e-06,
+      "loss": 1.0631,
+      "step": 15247
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.500514082114984e-06,
+      "loss": 0.9233,
+      "step": 15248
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.499388786257364e-06,
+      "loss": 1.0578,
+      "step": 15249
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.498263496754832e-06,
+      "loss": 0.9963,
+      "step": 15250
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.49713821362167e-06,
+      "loss": 0.9324,
+      "step": 15251
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.496012936872158e-06,
+      "loss": 1.0125,
+      "step": 15252
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.494887666520588e-06,
+      "loss": 0.9308,
+      "step": 15253
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.493762402581244e-06,
+      "loss": 1.0396,
+      "step": 15254
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.492637145068408e-06,
+      "loss": 0.904,
+      "step": 15255
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.491511893996366e-06,
+      "loss": 0.9656,
+      "step": 15256
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.490386649379401e-06,
+      "loss": 1.1034,
+      "step": 15257
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.489261411231803e-06,
+      "loss": 1.0623,
+      "step": 15258
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48813617956785e-06,
+      "loss": 1.0257,
+      "step": 15259
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.487010954401832e-06,
+      "loss": 1.0981,
+      "step": 15260
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.485885735748026e-06,
+      "loss": 1.0063,
+      "step": 15261
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.484760523620726e-06,
+      "loss": 1.0007,
+      "step": 15262
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.48363531803421e-06,
+      "loss": 1.045,
+      "step": 15263
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.482510119002762e-06,
+      "loss": 0.924,
+      "step": 15264
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.481384926540667e-06,
+      "loss": 0.8667,
+      "step": 15265
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.480259740662207e-06,
+      "loss": 0.927,
+      "step": 15266
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.479134561381672e-06,
+      "loss": 0.9765,
+      "step": 15267
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.478009388713341e-06,
+      "loss": 0.9378,
+      "step": 15268
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.476884222671498e-06,
+      "loss": 1.0588,
+      "step": 15269
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.475759063270427e-06,
+      "loss": 0.9529,
+      "step": 15270
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.474633910524409e-06,
+      "loss": 1.0605,
+      "step": 15271
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.473508764447731e-06,
+      "loss": 0.9766,
+      "step": 15272
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.472383625054676e-06,
+      "loss": 1.0094,
+      "step": 15273
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.471258492359529e-06,
+      "loss": 1.0767,
+      "step": 15274
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.470133366376564e-06,
+      "loss": 0.9484,
+      "step": 15275
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.469008247120076e-06,
+      "loss": 0.9673,
+      "step": 15276
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.467883134604342e-06,
+      "loss": 0.9357,
+      "step": 15277
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.466758028843646e-06,
+      "loss": 0.922,
+      "step": 15278
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.465632929852268e-06,
+      "loss": 1.015,
+      "step": 15279
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.464507837644491e-06,
+      "loss": 0.957,
+      "step": 15280
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.463382752234606e-06,
+      "loss": 1.0204,
+      "step": 15281
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.462257673636888e-06,
+      "loss": 1.1035,
+      "step": 15282
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.46113260186562e-06,
+      "loss": 1.0337,
+      "step": 15283
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.460007536935087e-06,
+      "loss": 0.8626,
+      "step": 15284
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.458882478859563e-06,
+      "loss": 1.1214,
+      "step": 15285
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.457757427653344e-06,
+      "loss": 0.9818,
+      "step": 15286
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.4566323833307e-06,
+      "loss": 0.9703,
+      "step": 15287
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.455507345905923e-06,
+      "loss": 0.9612,
+      "step": 15288
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.454382315393285e-06,
+      "loss": 0.9529,
+      "step": 15289
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.453257291807072e-06,
+      "loss": 0.9738,
+      "step": 15290
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.452132275161572e-06,
+      "loss": 0.9763,
+      "step": 15291
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.45100726547106e-06,
+      "loss": 0.9507,
+      "step": 15292
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.449882262749818e-06,
+      "loss": 0.969,
+      "step": 15293
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.448757267012125e-06,
+      "loss": 0.9744,
+      "step": 15294
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44763227827227e-06,
+      "loss": 0.9778,
+      "step": 15295
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44650729654453e-06,
+      "loss": 1.0612,
+      "step": 15296
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.445382321843186e-06,
+      "loss": 1.007,
+      "step": 15297
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.44425735418252e-06,
+      "loss": 0.9411,
+      "step": 15298
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.443132393576809e-06,
+      "loss": 1.0515,
+      "step": 15299
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.442007440040341e-06,
+      "loss": 1.0442,
+      "step": 15300
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.440882493587392e-06,
+      "loss": 0.9002,
+      "step": 15301
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.439757554232248e-06,
+      "loss": 0.9355,
+      "step": 15302
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.438632621989179e-06,
+      "loss": 1.1269,
+      "step": 15303
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.437507696872476e-06,
+      "loss": 0.9048,
+      "step": 15304
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.436382778896418e-06,
+      "loss": 1.0788,
+      "step": 15305
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.435257868075285e-06,
+      "loss": 1.1006,
+      "step": 15306
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.434132964423353e-06,
+      "loss": 1.0826,
+      "step": 15307
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.433008067954903e-06,
+      "loss": 0.9868,
+      "step": 15308
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.431883178684221e-06,
+      "loss": 1.0178,
+      "step": 15309
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.430758296625585e-06,
+      "loss": 0.8577,
+      "step": 15310
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.429633421793269e-06,
+      "loss": 0.8999,
+      "step": 15311
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.428508554201556e-06,
+      "loss": 1.1561,
+      "step": 15312
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.427383693864734e-06,
+      "loss": 0.9472,
+      "step": 15313
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.426258840797072e-06,
+      "loss": 1.0433,
+      "step": 15314
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.425133995012853e-06,
+      "loss": 1.0145,
+      "step": 15315
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.42400915652636e-06,
+      "loss": 1.0202,
+      "step": 15316
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.422884325351864e-06,
+      "loss": 0.9576,
+      "step": 15317
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.421759501503654e-06,
+      "loss": 0.9918,
+      "step": 15318
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.420634684996006e-06,
+      "loss": 0.9218,
+      "step": 15319
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.419509875843197e-06,
+      "loss": 0.9909,
+      "step": 15320
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.418385074059504e-06,
+      "loss": 0.9811,
+      "step": 15321
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.417260279659211e-06,
+      "loss": 0.9309,
+      "step": 15322
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.416135492656597e-06,
+      "loss": 0.9475,
+      "step": 15323
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.415010713065936e-06,
+      "loss": 1.0462,
+      "step": 15324
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.413885940901513e-06,
+      "loss": 1.0682,
+      "step": 15325
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.412761176177599e-06,
+      "loss": 1.0314,
+      "step": 15326
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41163641890848e-06,
+      "loss": 0.984,
+      "step": 15327
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.41051166910843e-06,
+      "loss": 0.9378,
+      "step": 15328
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.409386926791728e-06,
+      "loss": 1.0192,
+      "step": 15329
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.408262191972656e-06,
+      "loss": 1.1493,
+      "step": 15330
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.40713746466548e-06,
+      "loss": 0.9695,
+      "step": 15331
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.406012744884491e-06,
+      "loss": 0.9595,
+      "step": 15332
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.404888032643966e-06,
+      "loss": 0.9005,
+      "step": 15333
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.403763327958179e-06,
+      "loss": 0.9768,
+      "step": 15334
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.402638630841404e-06,
+      "loss": 1.04,
+      "step": 15335
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.401513941307923e-06,
+      "loss": 0.9492,
+      "step": 15336
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.400389259372016e-06,
+      "loss": 0.8737,
+      "step": 15337
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.399264585047957e-06,
+      "loss": 0.9027,
+      "step": 15338
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.398139918350025e-06,
+      "loss": 0.9118,
+      "step": 15339
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.397015259292492e-06,
+      "loss": 1.0178,
+      "step": 15340
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.395890607889644e-06,
+      "loss": 1.0065,
+      "step": 15341
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.394765964155753e-06,
+      "loss": 0.944,
+      "step": 15342
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.393641328105098e-06,
+      "loss": 0.9247,
+      "step": 15343
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.392516699751953e-06,
+      "loss": 0.9502,
+      "step": 15344
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.391392079110593e-06,
+      "loss": 0.9549,
+      "step": 15345
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.3902674661953e-06,
+      "loss": 0.9315,
+      "step": 15346
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.38914286102035e-06,
+      "loss": 1.0283,
+      "step": 15347
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.388018263600016e-06,
+      "loss": 1.1208,
+      "step": 15348
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.386893673948579e-06,
+      "loss": 1.0826,
+      "step": 15349
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.385769092080307e-06,
+      "loss": 0.9691,
+      "step": 15350
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.384644518009487e-06,
+      "loss": 0.9267,
+      "step": 15351
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.383519951750387e-06,
+      "loss": 1.0144,
+      "step": 15352
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.382395393317287e-06,
+      "loss": 0.9132,
+      "step": 15353
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.381270842724459e-06,
+      "loss": 1.1251,
+      "step": 15354
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.380146299986184e-06,
+      "loss": 0.86,
+      "step": 15355
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.379021765116735e-06,
+      "loss": 0.91,
+      "step": 15356
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37789723813039e-06,
+      "loss": 0.92,
+      "step": 15357
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.376772719041419e-06,
+      "loss": 0.9916,
+      "step": 15358
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.3756482078641e-06,
+      "loss": 1.0126,
+      "step": 15359
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.374523704612708e-06,
+      "loss": 0.9782,
+      "step": 15360
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.37339920930152e-06,
+      "loss": 0.9535,
+      "step": 15361
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.372274721944811e-06,
+      "loss": 1.0245,
+      "step": 15362
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.371150242556857e-06,
+      "loss": 1.0363,
+      "step": 15363
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.370025771151924e-06,
+      "loss": 0.9268,
+      "step": 15364
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.368901307744302e-06,
+      "loss": 0.9402,
+      "step": 15365
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.367776852348251e-06,
+      "loss": 1.0481,
+      "step": 15366
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.366652404978056e-06,
+      "loss": 0.9224,
+      "step": 15367
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.365527965647983e-06,
+      "loss": 1.0546,
+      "step": 15368
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.364403534372315e-06,
+      "loss": 0.9811,
+      "step": 15369
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.36327911116532e-06,
+      "loss": 0.9611,
+      "step": 15370
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.362154696041275e-06,
+      "loss": 1.1582,
+      "step": 15371
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.361030289014453e-06,
+      "loss": 0.9257,
+      "step": 15372
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.359905890099127e-06,
+      "loss": 1.0189,
+      "step": 15373
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.358781499309572e-06,
+      "loss": 0.9339,
+      "step": 15374
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.357657116660063e-06,
+      "loss": 1.0536,
+      "step": 15375
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.356532742164873e-06,
+      "loss": 0.8741,
+      "step": 15376
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.355408375838274e-06,
+      "loss": 0.8896,
+      "step": 15377
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.354284017694538e-06,
+      "loss": 0.9974,
+      "step": 15378
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.353159667747944e-06,
+      "loss": 1.0072,
+      "step": 15379
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.35203532601276e-06,
+      "loss": 0.9766,
+      "step": 15380
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.350910992503263e-06,
+      "loss": 1.0303,
+      "step": 15381
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.34978666723372e-06,
+      "loss": 0.9485,
+      "step": 15382
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.348662350218413e-06,
+      "loss": 0.946,
+      "step": 15383
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.347538041471606e-06,
+      "loss": 0.9725,
+      "step": 15384
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.346413741007578e-06,
+      "loss": 0.9624,
+      "step": 15385
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.345289448840599e-06,
+      "loss": 1.0605,
+      "step": 15386
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.344165164984937e-06,
+      "loss": 0.8837,
+      "step": 15387
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.343040889454876e-06,
+      "loss": 1.0788,
+      "step": 15388
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.341916622264679e-06,
+      "loss": 0.9435,
+      "step": 15389
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.340792363428617e-06,
+      "loss": 0.8676,
+      "step": 15390
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.339668112960972e-06,
+      "loss": 1.0075,
+      "step": 15391
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.338543870876001e-06,
+      "loss": 0.997,
+      "step": 15392
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.33741963718799e-06,
+      "loss": 1.0216,
+      "step": 15393
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.336295411911204e-06,
+      "loss": 1.0108,
+      "step": 15394
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.335171195059918e-06,
+      "loss": 0.9367,
+      "step": 15395
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.334046986648397e-06,
+      "loss": 0.9537,
+      "step": 15396
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.332922786690917e-06,
+      "loss": 0.9816,
+      "step": 15397
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.331798595201753e-06,
+      "loss": 1.0205,
+      "step": 15398
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.330674412195171e-06,
+      "loss": 0.9852,
+      "step": 15399
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.329550237685442e-06,
+      "loss": 1.0077,
+      "step": 15400
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.328426071686836e-06,
+      "loss": 0.9426,
+      "step": 15401
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.327301914213631e-06,
+      "loss": 0.9495,
+      "step": 15402
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.326177765280093e-06,
+      "loss": 1.0431,
+      "step": 15403
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.325053624900491e-06,
+      "loss": 1.0524,
+      "step": 15404
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.323929493089098e-06,
+      "loss": 0.908,
+      "step": 15405
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.32280536986018e-06,
+      "loss": 0.8219,
+      "step": 15406
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.321681255228017e-06,
+      "loss": 0.9001,
+      "step": 15407
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.320557149206869e-06,
+      "loss": 0.9018,
+      "step": 15408
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.319433051811013e-06,
+      "loss": 1.0808,
+      "step": 15409
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.318308963054713e-06,
+      "loss": 1.0183,
+      "step": 15410
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.317184882952244e-06,
+      "loss": 0.951,
+      "step": 15411
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.316060811517877e-06,
+      "loss": 0.9631,
+      "step": 15412
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.31493674876588e-06,
+      "loss": 1.0191,
+      "step": 15413
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.313812694710518e-06,
+      "loss": 1.0613,
+      "step": 15414
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.312688649366061e-06,
+      "loss": 0.966,
+      "step": 15415
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.311564612746787e-06,
+      "loss": 1.0308,
+      "step": 15416
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.310440584866959e-06,
+      "loss": 1.0102,
+      "step": 15417
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.309316565740845e-06,
+      "loss": 0.9606,
+      "step": 15418
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.308192555382719e-06,
+      "loss": 0.94,
+      "step": 15419
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.30706855380684e-06,
+      "loss": 1.0042,
+      "step": 15420
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.305944561027487e-06,
+      "loss": 0.9745,
+      "step": 15421
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.304820577058928e-06,
+      "loss": 0.9169,
+      "step": 15422
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.303696601915427e-06,
+      "loss": 1.1674,
+      "step": 15423
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.302572635611252e-06,
+      "loss": 0.8461,
+      "step": 15424
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.301448678160675e-06,
+      "loss": 1.0289,
+      "step": 15425
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.300324729577964e-06,
+      "loss": 1.0097,
+      "step": 15426
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.299200789877386e-06,
+      "loss": 1.0392,
+      "step": 15427
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.298076859073207e-06,
+      "loss": 0.9896,
+      "step": 15428
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.296952937179695e-06,
+      "loss": 1.1313,
+      "step": 15429
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.295829024211125e-06,
+      "loss": 0.8834,
+      "step": 15430
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.294705120181759e-06,
+      "loss": 1.0333,
+      "step": 15431
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.293581225105862e-06,
+      "loss": 0.9395,
+      "step": 15432
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.292457338997709e-06,
+      "loss": 0.9682,
+      "step": 15433
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.291333461871554e-06,
+      "loss": 0.9877,
+      "step": 15434
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.290209593741679e-06,
+      "loss": 0.9504,
+      "step": 15435
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.289085734622345e-06,
+      "loss": 0.9212,
+      "step": 15436
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.287961884527821e-06,
+      "loss": 1.0727,
+      "step": 15437
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.286838043472366e-06,
+      "loss": 1.0711,
+      "step": 15438
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.285714211470256e-06,
+      "loss": 0.9845,
+      "step": 15439
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.284590388535756e-06,
+      "loss": 0.915,
+      "step": 15440
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.283466574683132e-06,
+      "loss": 0.9115,
+      "step": 15441
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.282342769926646e-06,
+      "loss": 0.9936,
+      "step": 15442
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.281218974280565e-06,
+      "loss": 0.9895,
+      "step": 15443
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.280095187759165e-06,
+      "loss": 0.9,
+      "step": 15444
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2789714103767e-06,
+      "loss": 0.9396,
+      "step": 15445
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.277847642147446e-06,
+      "loss": 1.1139,
+      "step": 15446
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.276723883085656e-06,
+      "loss": 0.9731,
+      "step": 15447
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.27560013320561e-06,
+      "loss": 0.9625,
+      "step": 15448
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.274476392521565e-06,
+      "loss": 1.0891,
+      "step": 15449
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.273352661047789e-06,
+      "loss": 0.8842,
+      "step": 15450
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.272228938798548e-06,
+      "loss": 0.9909,
+      "step": 15451
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.271105225788101e-06,
+      "loss": 0.9686,
+      "step": 15452
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.269981522030723e-06,
+      "loss": 0.9436,
+      "step": 15453
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.268857827540675e-06,
+      "loss": 1.0773,
+      "step": 15454
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.267734142332219e-06,
+      "loss": 0.897,
+      "step": 15455
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.266610466419625e-06,
+      "loss": 0.9312,
+      "step": 15456
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.26548679981715e-06,
+      "loss": 0.8629,
+      "step": 15457
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.264363142539066e-06,
+      "loss": 0.9516,
+      "step": 15458
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.263239494599636e-06,
+      "loss": 1.0988,
+      "step": 15459
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.262115856013121e-06,
+      "loss": 0.9802,
+      "step": 15460
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.260992226793786e-06,
+      "loss": 0.9465,
+      "step": 15461
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.2598686069559e-06,
+      "loss": 0.9351,
+      "step": 15462
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.258744996513722e-06,
+      "loss": 0.8877,
+      "step": 15463
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.257621395481518e-06,
+      "loss": 1.0197,
+      "step": 15464
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.256497803873553e-06,
+      "loss": 0.9139,
+      "step": 15465
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.255374221704082e-06,
+      "loss": 1.0501,
+      "step": 15466
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.254250648987378e-06,
+      "loss": 0.973,
+      "step": 15467
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.253127085737703e-06,
+      "loss": 0.8891,
+      "step": 15468
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.252003531969317e-06,
+      "loss": 0.8961,
+      "step": 15469
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.250879987696487e-06,
+      "loss": 1.0643,
+      "step": 15470
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.249756452933468e-06,
+      "loss": 0.8502,
+      "step": 15471
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.248632927694534e-06,
+      "loss": 1.0116,
+      "step": 15472
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.247509411993941e-06,
+      "loss": 0.9084,
+      "step": 15473
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.246385905845954e-06,
+      "loss": 1.0955,
+      "step": 15474
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24526240926483e-06,
+      "loss": 0.951,
+      "step": 15475
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.24413892226484e-06,
+      "loss": 0.8892,
+      "step": 15476
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.243015444860242e-06,
+      "loss": 1.0531,
+      "step": 15477
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.241891977065299e-06,
+      "loss": 1.0226,
+      "step": 15478
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.240768518894271e-06,
+      "loss": 0.8976,
+      "step": 15479
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.239645070361421e-06,
+      "loss": 0.9303,
+      "step": 15480
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23852163148101e-06,
+      "loss": 1.0173,
+      "step": 15481
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.237398202267304e-06,
+      "loss": 0.9898,
+      "step": 15482
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.236274782734558e-06,
+      "loss": 1.1057,
+      "step": 15483
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.23515137289704e-06,
+      "loss": 0.9457,
+      "step": 15484
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.234027972769004e-06,
+      "loss": 0.9721,
+      "step": 15485
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.232904582364718e-06,
+      "loss": 0.8794,
+      "step": 15486
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.231781201698438e-06,
+      "loss": 0.9895,
+      "step": 15487
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.230657830784431e-06,
+      "loss": 0.8712,
+      "step": 15488
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.229534469636948e-06,
+      "loss": 1.0958,
+      "step": 15489
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.228411118270261e-06,
+      "loss": 1.1504,
+      "step": 15490
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.227287776698623e-06,
+      "loss": 0.9779,
+      "step": 15491
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.226164444936298e-06,
+      "loss": 0.972,
+      "step": 15492
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.225041122997544e-06,
+      "loss": 0.8986,
+      "step": 15493
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.223917810896617e-06,
+      "loss": 1.0292,
+      "step": 15494
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.22279450864779e-06,
+      "loss": 1.0205,
+      "step": 15495
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.221671216265314e-06,
+      "loss": 0.9775,
+      "step": 15496
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.220547933763447e-06,
+      "loss": 1.0274,
+      "step": 15497
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.219424661156454e-06,
+      "loss": 0.9944,
+      "step": 15498
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.218301398458587e-06,
+      "loss": 0.9415,
+      "step": 15499
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.217178145684114e-06,
+      "loss": 0.9448,
+      "step": 15500
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.216054902847291e-06,
+      "loss": 1.1397,
+      "step": 15501
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.214931669962379e-06,
+      "loss": 0.9678,
+      "step": 15502
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21380844704363e-06,
+      "loss": 0.9188,
+      "step": 15503
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.212685234105312e-06,
+      "loss": 0.9748,
+      "step": 15504
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.211562031161677e-06,
+      "loss": 0.9039,
+      "step": 15505
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.21043883822699e-06,
+      "loss": 1.0182,
+      "step": 15506
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.209315655315503e-06,
+      "loss": 0.9107,
+      "step": 15507
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.208192482441475e-06,
+      "loss": 1.1535,
+      "step": 15508
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.207069319619172e-06,
+      "loss": 1.0761,
+      "step": 15509
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.205946166862846e-06,
+      "loss": 1.0201,
+      "step": 15510
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.204823024186755e-06,
+      "loss": 1.1526,
+      "step": 15511
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.203699891605158e-06,
+      "loss": 0.9203,
+      "step": 15512
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.20257676913231e-06,
+      "loss": 0.9166,
+      "step": 15513
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.201453656782474e-06,
+      "loss": 0.9348,
+      "step": 15514
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.200330554569904e-06,
+      "loss": 0.893,
+      "step": 15515
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.199207462508859e-06,
+      "loss": 1.094,
+      "step": 15516
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.198084380613591e-06,
+      "loss": 0.9249,
+      "step": 15517
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.196961308898364e-06,
+      "loss": 0.8623,
+      "step": 15518
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.195838247377436e-06,
+      "loss": 0.8981,
+      "step": 15519
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.194715196065059e-06,
+      "loss": 0.9967,
+      "step": 15520
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.193592154975489e-06,
+      "loss": 1.1496,
+      "step": 15521
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.19246912412298e-06,
+      "loss": 0.8305,
+      "step": 15522
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.191346103521802e-06,
+      "loss": 0.9986,
+      "step": 15523
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.190223093186201e-06,
+      "loss": 1.1288,
+      "step": 15524
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.189100093130432e-06,
+      "loss": 0.9022,
+      "step": 15525
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.187977103368756e-06,
+      "loss": 0.9868,
+      "step": 15526
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.186854123915422e-06,
+      "loss": 1.0894,
+      "step": 15527
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.185731154784696e-06,
+      "loss": 0.9211,
+      "step": 15528
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.184608195990827e-06,
+      "loss": 0.998,
+      "step": 15529
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.183485247548074e-06,
+      "loss": 1.0542,
+      "step": 15530
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.182362309470685e-06,
+      "loss": 0.952,
+      "step": 15531
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.181239381772923e-06,
+      "loss": 0.9373,
+      "step": 15532
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.180116464469044e-06,
+      "loss": 0.8664,
+      "step": 15533
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.178993557573297e-06,
+      "loss": 0.9859,
+      "step": 15534
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17787066109994e-06,
+      "loss": 1.0434,
+      "step": 15535
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.176747775063226e-06,
+      "loss": 1.0479,
+      "step": 15536
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.175624899477415e-06,
+      "loss": 1.0088,
+      "step": 15537
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.174502034356756e-06,
+      "loss": 1.0258,
+      "step": 15538
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.173379179715506e-06,
+      "loss": 0.9821,
+      "step": 15539
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.17225633556792e-06,
+      "loss": 1.0592,
+      "step": 15540
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.171133501928244e-06,
+      "loss": 1.0079,
+      "step": 15541
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.170010678810741e-06,
+      "loss": 0.9512,
+      "step": 15542
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.168887866229667e-06,
+      "loss": 0.8982,
+      "step": 15543
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16776506419927e-06,
+      "loss": 0.8927,
+      "step": 15544
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.166642272733797e-06,
+      "loss": 1.0062,
+      "step": 15545
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.165519491847515e-06,
+      "loss": 0.8937,
+      "step": 15546
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.164396721554673e-06,
+      "loss": 1.0993,
+      "step": 15547
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.16327396186952e-06,
+      "loss": 1.0313,
+      "step": 15548
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.162151212806312e-06,
+      "loss": 0.891,
+      "step": 15549
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.161028474379299e-06,
+      "loss": 0.9282,
+      "step": 15550
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15990574660274e-06,
+      "loss": 0.9397,
+      "step": 15551
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.158783029490881e-06,
+      "loss": 1.0204,
+      "step": 15552
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.157660323057983e-06,
+      "loss": 1.0271,
+      "step": 15553
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15653762731829e-06,
+      "loss": 0.8976,
+      "step": 15554
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.155414942286051e-06,
+      "loss": 1.0167,
+      "step": 15555
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.15429226797553e-06,
+      "loss": 0.9903,
+      "step": 15556
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.153169604400974e-06,
+      "loss": 1.026,
+      "step": 15557
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.152046951576632e-06,
+      "loss": 1.0651,
+      "step": 15558
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.150924309516754e-06,
+      "loss": 0.961,
+      "step": 15559
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.149801678235598e-06,
+      "loss": 0.9588,
+      "step": 15560
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.148679057747415e-06,
+      "loss": 1.028,
+      "step": 15561
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.147556448066453e-06,
+      "loss": 1.0188,
+      "step": 15562
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.146433849206962e-06,
+      "loss": 1.0105,
+      "step": 15563
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.145311261183192e-06,
+      "loss": 0.889,
+      "step": 15564
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.144188684009403e-06,
+      "loss": 0.8888,
+      "step": 15565
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.143066117699837e-06,
+      "loss": 0.9742,
+      "step": 15566
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.141943562268748e-06,
+      "loss": 0.9853,
+      "step": 15567
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.140821017730386e-06,
+      "loss": 1.0475,
+      "step": 15568
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.139698484098997e-06,
+      "loss": 1.0926,
+      "step": 15569
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.138575961388837e-06,
+      "loss": 0.9868,
+      "step": 15570
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.137453449614157e-06,
+      "loss": 0.9615,
+      "step": 15571
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.136330948789202e-06,
+      "loss": 1.013,
+      "step": 15572
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.135208458928221e-06,
+      "loss": 1.0352,
+      "step": 15573
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.134085980045467e-06,
+      "loss": 0.8752,
+      "step": 15574
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.132963512155193e-06,
+      "loss": 0.9821,
+      "step": 15575
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.13184105527164e-06,
+      "loss": 0.9503,
+      "step": 15576
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.130718609409065e-06,
+      "loss": 1.0206,
+      "step": 15577
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.129596174581709e-06,
+      "loss": 1.1441,
+      "step": 15578
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12847375080383e-06,
+      "loss": 0.8952,
+      "step": 15579
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.127351338089668e-06,
+      "loss": 0.9599,
+      "step": 15580
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.12622893645348e-06,
+      "loss": 0.9475,
+      "step": 15581
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.125106545909509e-06,
+      "loss": 0.9597,
+      "step": 15582
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.123984166472e-06,
+      "loss": 0.9786,
+      "step": 15583
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.122861798155208e-06,
+      "loss": 1.0153,
+      "step": 15584
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.121739440973382e-06,
+      "loss": 1.0424,
+      "step": 15585
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.120617094940764e-06,
+      "loss": 0.8407,
+      "step": 15586
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.119494760071605e-06,
+      "loss": 0.922,
+      "step": 15587
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.118372436380152e-06,
+      "loss": 0.9657,
+      "step": 15588
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.117250123880655e-06,
+      "loss": 1.0407,
+      "step": 15589
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.116127822587357e-06,
+      "loss": 1.0982,
+      "step": 15590
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11500553251451e-06,
+      "loss": 0.9995,
+      "step": 15591
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.113883253676354e-06,
+      "loss": 0.9555,
+      "step": 15592
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.112760986087144e-06,
+      "loss": 0.8901,
+      "step": 15593
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.11163872976112e-06,
+      "loss": 1.1223,
+      "step": 15594
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.110516484712536e-06,
+      "loss": 1.015,
+      "step": 15595
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.109394250955628e-06,
+      "loss": 1.0035,
+      "step": 15596
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.108272028504655e-06,
+      "loss": 1.0,
+      "step": 15597
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.107149817373855e-06,
+      "loss": 0.87,
+      "step": 15598
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.106027617577476e-06,
+      "loss": 0.9146,
+      "step": 15599
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.104905429129763e-06,
+      "loss": 1.1924,
+      "step": 15600
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.103783252044961e-06,
+      "loss": 0.962,
+      "step": 15601
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.102661086337321e-06,
+      "loss": 0.968,
+      "step": 15602
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.101538932021084e-06,
+      "loss": 1.0529,
+      "step": 15603
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.100416789110495e-06,
+      "loss": 1.0152,
+      "step": 15604
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.099294657619804e-06,
+      "loss": 1.0345,
+      "step": 15605
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.098172537563245e-06,
+      "loss": 0.9984,
+      "step": 15606
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.097050428955077e-06,
+      "loss": 0.9351,
+      "step": 15607
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.095928331809536e-06,
+      "loss": 0.9327,
+      "step": 15608
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09480624614087e-06,
+      "loss": 0.9296,
+      "step": 15609
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.09368417196332e-06,
+      "loss": 1.016,
+      "step": 15610
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.092562109291134e-06,
+      "loss": 1.0742,
+      "step": 15611
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.091440058138556e-06,
+      "loss": 0.9851,
+      "step": 15612
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.090318018519827e-06,
+      "loss": 0.9828,
+      "step": 15613
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.089195990449194e-06,
+      "loss": 0.992,
+      "step": 15614
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.088073973940897e-06,
+      "loss": 1.0088,
+      "step": 15615
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.086951969009185e-06,
+      "loss": 0.9533,
+      "step": 15616
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.085829975668298e-06,
+      "loss": 1.0447,
+      "step": 15617
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08470799393248e-06,
+      "loss": 0.9612,
+      "step": 15618
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.083586023815975e-06,
+      "loss": 1.0296,
+      "step": 15619
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.082464065333021e-06,
+      "loss": 0.9997,
+      "step": 15620
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.08134211849787e-06,
+      "loss": 0.9832,
+      "step": 15621
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.080220183324757e-06,
+      "loss": 0.9789,
+      "step": 15622
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.07909825982793e-06,
+      "loss": 1.0144,
+      "step": 15623
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.077976348021623e-06,
+      "loss": 0.8725,
+      "step": 15624
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.076854447920088e-06,
+      "loss": 1.152,
+      "step": 15625
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.075732559537564e-06,
+      "loss": 0.9625,
+      "step": 15626
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.074610682888293e-06,
+      "loss": 1.0173,
+      "step": 15627
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.073488817986511e-06,
+      "loss": 0.8779,
+      "step": 15628
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.072366964846465e-06,
+      "loss": 1.0834,
+      "step": 15629
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0712451234824e-06,
+      "loss": 1.02,
+      "step": 15630
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.070123293908553e-06,
+      "loss": 0.9195,
+      "step": 15631
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.069001476139165e-06,
+      "loss": 0.908,
+      "step": 15632
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06787967018848e-06,
+      "loss": 1.0253,
+      "step": 15633
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06675787607073e-06,
+      "loss": 1.0323,
+      "step": 15634
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.065636093800168e-06,
+      "loss": 0.9619,
+      "step": 15635
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.064514323391027e-06,
+      "loss": 1.0844,
+      "step": 15636
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.063392564857553e-06,
+      "loss": 0.917,
+      "step": 15637
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.062270818213977e-06,
+      "loss": 0.7866,
+      "step": 15638
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.06114908347455e-06,
+      "loss": 0.9717,
+      "step": 15639
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.060027360653507e-06,
+      "loss": 1.0055,
+      "step": 15640
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.058905649765089e-06,
+      "loss": 1.0535,
+      "step": 15641
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.057783950823532e-06,
+      "loss": 0.962,
+      "step": 15642
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.056662263843077e-06,
+      "loss": 0.9362,
+      "step": 15643
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.055540588837968e-06,
+      "loss": 0.9798,
+      "step": 15644
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.054418925822443e-06,
+      "loss": 1.0781,
+      "step": 15645
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.053297274810734e-06,
+      "loss": 1.0557,
+      "step": 15646
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.05217563581709e-06,
+      "loss": 1.0419,
+      "step": 15647
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.051054008855741e-06,
+      "loss": 0.9617,
+      "step": 15648
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.049932393940931e-06,
+      "loss": 0.9829,
+      "step": 15649
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0488107910869e-06,
+      "loss": 1.0348,
+      "step": 15650
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.047689200307882e-06,
+      "loss": 0.9216,
+      "step": 15651
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.046567621618114e-06,
+      "loss": 1.0663,
+      "step": 15652
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.045446055031836e-06,
+      "loss": 1.0571,
+      "step": 15653
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.044324500563292e-06,
+      "loss": 0.8922,
+      "step": 15654
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.043202958226713e-06,
+      "loss": 1.0326,
+      "step": 15655
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.042081428036337e-06,
+      "loss": 1.0704,
+      "step": 15656
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.0409599100064e-06,
+      "loss": 0.9283,
+      "step": 15657
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.039838404151144e-06,
+      "loss": 0.9831,
+      "step": 15658
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.038716910484804e-06,
+      "loss": 1.0549,
+      "step": 15659
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.037595429021618e-06,
+      "loss": 0.9736,
+      "step": 15660
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.036473959775823e-06,
+      "loss": 1.0999,
+      "step": 15661
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.035352502761648e-06,
+      "loss": 0.8912,
+      "step": 15662
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.03423105799334e-06,
+      "loss": 1.0483,
+      "step": 15663
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.03310962548513e-06,
+      "loss": 1.1103,
+      "step": 15664
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.031988205251258e-06,
+      "loss": 0.936,
+      "step": 15665
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.030866797305952e-06,
+      "loss": 0.897,
+      "step": 15666
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.029745401663456e-06,
+      "loss": 0.981,
+      "step": 15667
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.028624018338005e-06,
+      "loss": 0.9601,
+      "step": 15668
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.027502647343834e-06,
+      "loss": 1.1513,
+      "step": 15669
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.026381288695171e-06,
+      "loss": 0.9343,
+      "step": 15670
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02525994240626e-06,
+      "loss": 0.868,
+      "step": 15671
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.024138608491333e-06,
+      "loss": 1.0662,
+      "step": 15672
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.023017286964627e-06,
+      "loss": 0.9747,
+      "step": 15673
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.021895977840375e-06,
+      "loss": 1.0624,
+      "step": 15674
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.02077468113281e-06,
+      "loss": 0.9465,
+      "step": 15675
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.019653396856167e-06,
+      "loss": 0.9393,
+      "step": 15676
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.018532125024683e-06,
+      "loss": 1.0189,
+      "step": 15677
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.017410865652593e-06,
+      "loss": 0.9879,
+      "step": 15678
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.016289618754128e-06,
+      "loss": 0.8207,
+      "step": 15679
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.015168384343518e-06,
+      "loss": 0.9616,
+      "step": 15680
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.014047162435002e-06,
+      "loss": 1.0399,
+      "step": 15681
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.012925953042817e-06,
+      "loss": 0.958,
+      "step": 15682
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.01180475618119e-06,
+      "loss": 0.9811,
+      "step": 15683
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.010683571864355e-06,
+      "loss": 0.9066,
+      "step": 15684
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.009562400106547e-06,
+      "loss": 1.0712,
+      "step": 15685
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.008441240921998e-06,
+      "loss": 1.0026,
+      "step": 15686
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.007320094324941e-06,
+      "loss": 0.9842,
+      "step": 15687
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.00619896032961e-06,
+      "loss": 1.005,
+      "step": 15688
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.005077838950235e-06,
+      "loss": 1.0461,
+      "step": 15689
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.003956730201046e-06,
+      "loss": 0.9489,
+      "step": 15690
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.002835634096281e-06,
+      "loss": 0.999,
+      "step": 15691
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.001714550650168e-06,
+      "loss": 0.9543,
+      "step": 15692
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.000593479876944e-06,
+      "loss": 0.9113,
+      "step": 15693
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.99947242179083e-06,
+      "loss": 0.8364,
+      "step": 15694
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.998351376406066e-06,
+      "loss": 0.9896,
+      "step": 15695
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.997230343736883e-06,
+      "loss": 0.933,
+      "step": 15696
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.996109323797509e-06,
+      "loss": 1.0611,
+      "step": 15697
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.994988316602178e-06,
+      "loss": 1.0383,
+      "step": 15698
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.993867322165116e-06,
+      "loss": 1.0512,
+      "step": 15699
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.992746340500559e-06,
+      "loss": 0.9501,
+      "step": 15700
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.991625371622733e-06,
+      "loss": 0.9241,
+      "step": 15701
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.990504415545874e-06,
+      "loss": 1.1549,
+      "step": 15702
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.989383472284208e-06,
+      "loss": 0.9771,
+      "step": 15703
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.988262541851961e-06,
+      "loss": 1.0349,
+      "step": 15704
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98714162426337e-06,
+      "loss": 0.9731,
+      "step": 15705
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.986020719532664e-06,
+      "loss": 1.1147,
+      "step": 15706
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.984899827674067e-06,
+      "loss": 0.9384,
+      "step": 15707
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98377894870181e-06,
+      "loss": 0.9723,
+      "step": 15708
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98265808263013e-06,
+      "loss": 1.0089,
+      "step": 15709
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.98153722947325e-06,
+      "loss": 1.0508,
+      "step": 15710
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.980416389245397e-06,
+      "loss": 0.9861,
+      "step": 15711
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.979295561960802e-06,
+      "loss": 0.9951,
+      "step": 15712
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.97817474763369e-06,
+      "loss": 1.0182,
+      "step": 15713
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.977053946278297e-06,
+      "loss": 0.971,
+      "step": 15714
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.975933157908844e-06,
+      "loss": 1.0811,
+      "step": 15715
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.974812382539563e-06,
+      "loss": 1.0422,
+      "step": 15716
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.973691620184678e-06,
+      "loss": 1.0361,
+      "step": 15717
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.972570870858421e-06,
+      "loss": 1.0153,
+      "step": 15718
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.971450134575016e-06,
+      "loss": 0.8761,
+      "step": 15719
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.970329411348695e-06,
+      "loss": 1.0798,
+      "step": 15720
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.96920870119368e-06,
+      "loss": 1.1223,
+      "step": 15721
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.968088004124199e-06,
+      "loss": 0.967,
+      "step": 15722
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.966967320154483e-06,
+      "loss": 1.0444,
+      "step": 15723
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.965846649298756e-06,
+      "loss": 0.9143,
+      "step": 15724
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.964725991571243e-06,
+      "loss": 0.9672,
+      "step": 15725
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.963605346986173e-06,
+      "loss": 0.9882,
+      "step": 15726
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.962484715557766e-06,
+      "loss": 0.9994,
+      "step": 15727
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.961364097300259e-06,
+      "loss": 0.8715,
+      "step": 15728
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.960243492227868e-06,
+      "loss": 0.901,
+      "step": 15729
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.959122900354824e-06,
+      "loss": 1.0047,
+      "step": 15730
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.958002321695348e-06,
+      "loss": 0.9999,
+      "step": 15731
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95688175626367e-06,
+      "loss": 0.9466,
+      "step": 15732
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.955761204074016e-06,
+      "loss": 0.8735,
+      "step": 15733
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.95464066514061e-06,
+      "loss": 0.9021,
+      "step": 15734
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.953520139477671e-06,
+      "loss": 0.9653,
+      "step": 15735
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.952399627099426e-06,
+      "loss": 0.9872,
+      "step": 15736
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.951279128020107e-06,
+      "loss": 0.8978,
+      "step": 15737
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.950158642253933e-06,
+      "loss": 0.9436,
+      "step": 15738
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.949038169815124e-06,
+      "loss": 0.9076,
+      "step": 15739
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.947917710717912e-06,
+      "loss": 0.9414,
+      "step": 15740
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.946797264976511e-06,
+      "loss": 0.8791,
+      "step": 15741
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.945676832605156e-06,
+      "loss": 0.9702,
+      "step": 15742
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.944556413618062e-06,
+      "loss": 0.9288,
+      "step": 15743
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.943436008029459e-06,
+      "loss": 1.0315,
+      "step": 15744
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.942315615853561e-06,
+      "loss": 0.978,
+      "step": 15745
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.941195237104601e-06,
+      "loss": 0.9821,
+      "step": 15746
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.940074871796797e-06,
+      "loss": 0.8476,
+      "step": 15747
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.938954519944374e-06,
+      "loss": 0.8839,
+      "step": 15748
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.937834181561549e-06,
+      "loss": 0.9018,
+      "step": 15749
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.936713856662546e-06,
+      "loss": 1.0798,
+      "step": 15750
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.935593545261592e-06,
+      "loss": 0.9326,
+      "step": 15751
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.934473247372908e-06,
+      "loss": 1.0395,
+      "step": 15752
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.933352963010711e-06,
+      "loss": 1.0034,
+      "step": 15753
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.932232692189227e-06,
+      "loss": 1.1393,
+      "step": 15754
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.931112434922671e-06,
+      "loss": 1.005,
+      "step": 15755
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.929992191225271e-06,
+      "loss": 0.9416,
+      "step": 15756
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.928871961111249e-06,
+      "loss": 0.9684,
+      "step": 15757
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.927751744594824e-06,
+      "loss": 1.056,
+      "step": 15758
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.926631541690211e-06,
+      "loss": 0.9925,
+      "step": 15759
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.925511352411637e-06,
+      "loss": 0.9876,
+      "step": 15760
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.924391176773322e-06,
+      "loss": 0.9732,
+      "step": 15761
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.923271014789487e-06,
+      "loss": 0.997,
+      "step": 15762
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.922150866474346e-06,
+      "loss": 0.8611,
+      "step": 15763
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.921030731842121e-06,
+      "loss": 1.0418,
+      "step": 15764
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.91991061090704e-06,
+      "loss": 1.0689,
+      "step": 15765
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.918790503683315e-06,
+      "loss": 0.9034,
+      "step": 15766
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.917670410185165e-06,
+      "loss": 0.9669,
+      "step": 15767
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.916550330426813e-06,
+      "loss": 0.9472,
+      "step": 15768
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.915430264422471e-06,
+      "loss": 0.967,
+      "step": 15769
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.914310212186364e-06,
+      "loss": 1.0328,
+      "step": 15770
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.913190173732713e-06,
+      "loss": 1.0943,
+      "step": 15771
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.912070149075734e-06,
+      "loss": 1.0001,
+      "step": 15772
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.910950138229637e-06,
+      "loss": 0.9349,
+      "step": 15773
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90983014120865e-06,
+      "loss": 1.048,
+      "step": 15774
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.908710158026992e-06,
+      "loss": 0.9073,
+      "step": 15775
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.907590188698876e-06,
+      "loss": 1.0539,
+      "step": 15776
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.90647023323852e-06,
+      "loss": 1.0067,
+      "step": 15777
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.905350291660138e-06,
+      "loss": 1.0374,
+      "step": 15778
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.904230363977957e-06,
+      "loss": 1.0878,
+      "step": 15779
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.903110450206186e-06,
+      "loss": 0.8328,
+      "step": 15780
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.901990550359046e-06,
+      "loss": 1.0641,
+      "step": 15781
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.900870664450753e-06,
+      "loss": 0.9748,
+      "step": 15782
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.899750792495519e-06,
+      "loss": 0.967,
+      "step": 15783
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.898630934507565e-06,
+      "loss": 0.9893,
+      "step": 15784
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.897511090501109e-06,
+      "loss": 0.9477,
+      "step": 15785
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.896391260490365e-06,
+      "loss": 0.9817,
+      "step": 15786
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.895271444489543e-06,
+      "loss": 0.9888,
+      "step": 15787
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.894151642512864e-06,
+      "loss": 1.0734,
+      "step": 15788
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.893031854574548e-06,
+      "loss": 1.0527,
+      "step": 15789
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.891912080688806e-06,
+      "loss": 0.9981,
+      "step": 15790
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.890792320869851e-06,
+      "loss": 0.9796,
+      "step": 15791
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.889672575131896e-06,
+      "loss": 0.9408,
+      "step": 15792
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.888552843489165e-06,
+      "loss": 0.9199,
+      "step": 15793
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.887433125955864e-06,
+      "loss": 1.1489,
+      "step": 15794
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.886313422546213e-06,
+      "loss": 0.9733,
+      "step": 15795
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.885193733274424e-06,
+      "loss": 0.9481,
+      "step": 15796
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.884074058154706e-06,
+      "loss": 0.8498,
+      "step": 15797
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.882954397201283e-06,
+      "loss": 1.1008,
+      "step": 15798
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.881834750428362e-06,
+      "loss": 0.9348,
+      "step": 15799
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.88071511785016e-06,
+      "loss": 0.9249,
+      "step": 15800
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.879595499480884e-06,
+      "loss": 0.9493,
+      "step": 15801
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.878475895334752e-06,
+      "loss": 0.9557,
+      "step": 15802
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.87735630542598e-06,
+      "loss": 1.037,
+      "step": 15803
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.876236729768774e-06,
+      "loss": 1.021,
+      "step": 15804
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.875117168377354e-06,
+      "loss": 1.0751,
+      "step": 15805
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.873997621265922e-06,
+      "loss": 0.8684,
+      "step": 15806
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.872878088448702e-06,
+      "loss": 0.9751,
+      "step": 15807
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.871758569939899e-06,
+      "loss": 0.9617,
+      "step": 15808
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.870639065753727e-06,
+      "loss": 1.046,
+      "step": 15809
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.869519575904398e-06,
+      "loss": 1.0099,
+      "step": 15810
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.86840010040612e-06,
+      "loss": 0.995,
+      "step": 15811
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.867280639273108e-06,
+      "loss": 1.0162,
+      "step": 15812
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.866161192519574e-06,
+      "loss": 1.061,
+      "step": 15813
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.865041760159726e-06,
+      "loss": 0.9786,
+      "step": 15814
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.863922342207775e-06,
+      "loss": 0.9233,
+      "step": 15815
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.862802938677936e-06,
+      "loss": 1.0788,
+      "step": 15816
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.861683549584416e-06,
+      "loss": 0.8972,
+      "step": 15817
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.860564174941422e-06,
+      "loss": 0.9201,
+      "step": 15818
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.859444814763171e-06,
+      "loss": 0.9614,
+      "step": 15819
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.858325469063866e-06,
+      "loss": 1.0058,
+      "step": 15820
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.857206137857724e-06,
+      "loss": 1.0428,
+      "step": 15821
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.856086821158948e-06,
+      "loss": 0.9118,
+      "step": 15822
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.854967518981753e-06,
+      "loss": 1.0532,
+      "step": 15823
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.853848231340344e-06,
+      "loss": 1.0896,
+      "step": 15824
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.852728958248928e-06,
+      "loss": 0.9359,
+      "step": 15825
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.851609699721718e-06,
+      "loss": 1.0599,
+      "step": 15826
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.850490455772923e-06,
+      "loss": 1.0111,
+      "step": 15827
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.84937122641675e-06,
+      "loss": 0.8684,
+      "step": 15828
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.848252011667403e-06,
+      "loss": 0.9858,
+      "step": 15829
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.847132811539098e-06,
+      "loss": 1.0264,
+      "step": 15830
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.846013626046041e-06,
+      "loss": 0.9192,
+      "step": 15831
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.844894455202434e-06,
+      "loss": 0.9407,
+      "step": 15832
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.843775299022491e-06,
+      "loss": 0.9983,
+      "step": 15833
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.842656157520412e-06,
+      "loss": 1.0074,
+      "step": 15834
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.841537030710412e-06,
+      "loss": 0.8477,
+      "step": 15835
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.840417918606693e-06,
+      "loss": 0.9108,
+      "step": 15836
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.839298821223465e-06,
+      "loss": 1.0886,
+      "step": 15837
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.838179738574932e-06,
+      "loss": 1.0807,
+      "step": 15838
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.837060670675298e-06,
+      "loss": 0.9663,
+      "step": 15839
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.835941617538777e-06,
+      "loss": 1.014,
+      "step": 15840
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83482257917957e-06,
+      "loss": 0.842,
+      "step": 15841
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83370355561188e-06,
+      "loss": 1.0205,
+      "step": 15842
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.832584546849914e-06,
+      "loss": 0.8734,
+      "step": 15843
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.831465552907884e-06,
+      "loss": 1.1587,
+      "step": 15844
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.83034657379999e-06,
+      "loss": 1.0154,
+      "step": 15845
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.829227609540436e-06,
+      "loss": 1.0897,
+      "step": 15846
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82810866014343e-06,
+      "loss": 0.9646,
+      "step": 15847
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82698972562317e-06,
+      "loss": 0.8975,
+      "step": 15848
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82587080599387e-06,
+      "loss": 0.9848,
+      "step": 15849
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.824751901269729e-06,
+      "loss": 1.0488,
+      "step": 15850
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.823633011464952e-06,
+      "loss": 1.0209,
+      "step": 15851
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.82251413659374e-06,
+      "loss": 0.9873,
+      "step": 15852
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.8213952766703e-06,
+      "loss": 0.9808,
+      "step": 15853
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.820276431708838e-06,
+      "loss": 1.0131,
+      "step": 15854
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.819157601723553e-06,
+      "loss": 1.0299,
+      "step": 15855
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.818038786728648e-06,
+      "loss": 0.9913,
+      "step": 15856
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.816919986738327e-06,
+      "loss": 0.9316,
+      "step": 15857
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.815801201766796e-06,
+      "loss": 0.9679,
+      "step": 15858
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.814682431828257e-06,
+      "loss": 0.8865,
+      "step": 15859
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.813563676936905e-06,
+      "loss": 1.1005,
+      "step": 15860
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.81244493710695e-06,
+      "loss": 0.9202,
+      "step": 15861
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.811326212352589e-06,
+      "loss": 1.0684,
+      "step": 15862
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.810207502688028e-06,
+      "loss": 1.0501,
+      "step": 15863
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.809088808127468e-06,
+      "loss": 0.9296,
+      "step": 15864
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80797012868511e-06,
+      "loss": 1.0052,
+      "step": 15865
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.806851464375152e-06,
+      "loss": 0.9513,
+      "step": 15866
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.805732815211797e-06,
+      "loss": 1.0433,
+      "step": 15867
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.804614181209249e-06,
+      "loss": 0.9404,
+      "step": 15868
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.803495562381706e-06,
+      "loss": 1.0303,
+      "step": 15869
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.802376958743366e-06,
+      "loss": 0.8777,
+      "step": 15870
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.80125837030843e-06,
+      "loss": 1.0664,
+      "step": 15871
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.800139797091106e-06,
+      "loss": 1.0424,
+      "step": 15872
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.799021239105587e-06,
+      "loss": 0.9722,
+      "step": 15873
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.797902696366072e-06,
+      "loss": 0.9685,
+      "step": 15874
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.796784168886763e-06,
+      "loss": 1.0435,
+      "step": 15875
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.795665656681856e-06,
+      "loss": 1.1861,
+      "step": 15876
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.794547159765553e-06,
+      "loss": 0.9182,
+      "step": 15877
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.793428678152055e-06,
+      "loss": 1.0366,
+      "step": 15878
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.792310211855557e-06,
+      "loss": 0.9297,
+      "step": 15879
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.791191760890256e-06,
+      "loss": 0.9401,
+      "step": 15880
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.790073325270356e-06,
+      "loss": 0.8551,
+      "step": 15881
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.788954905010051e-06,
+      "loss": 0.9918,
+      "step": 15882
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.787836500123544e-06,
+      "loss": 1.0888,
+      "step": 15883
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.786718110625025e-06,
+      "loss": 0.9779,
+      "step": 15884
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.785599736528694e-06,
+      "loss": 0.9507,
+      "step": 15885
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.784481377848753e-06,
+      "loss": 0.9444,
+      "step": 15886
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.783363034599395e-06,
+      "loss": 0.9946,
+      "step": 15887
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.782244706794819e-06,
+      "loss": 1.1011,
+      "step": 15888
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.781126394449223e-06,
+      "loss": 0.9526,
+      "step": 15889
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.780008097576794e-06,
+      "loss": 1.0033,
+      "step": 15890
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77888981619174e-06,
+      "loss": 0.9551,
+      "step": 15891
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.777771550308254e-06,
+      "loss": 0.9678,
+      "step": 15892
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.776653299940531e-06,
+      "loss": 0.9559,
+      "step": 15893
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.775535065102764e-06,
+      "loss": 0.9205,
+      "step": 15894
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.774416845809153e-06,
+      "loss": 0.9691,
+      "step": 15895
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.773298642073894e-06,
+      "loss": 1.0136,
+      "step": 15896
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.77218045391118e-06,
+      "loss": 1.0515,
+      "step": 15897
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.771062281335203e-06,
+      "loss": 0.889,
+      "step": 15898
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.769944124360158e-06,
+      "loss": 1.0291,
+      "step": 15899
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.768825983000248e-06,
+      "loss": 1.0741,
+      "step": 15900
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.767707857269659e-06,
+      "loss": 0.89,
+      "step": 15901
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.76658974718259e-06,
+      "loss": 1.0063,
+      "step": 15902
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.765471652753234e-06,
+      "loss": 0.861,
+      "step": 15903
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.764353573995778e-06,
+      "loss": 1.0572,
+      "step": 15904
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.763235510924424e-06,
+      "loss": 1.0497,
+      "step": 15905
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.762117463553366e-06,
+      "loss": 0.8962,
+      "step": 15906
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.760999431896794e-06,
+      "loss": 0.9362,
+      "step": 15907
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.759881415968897e-06,
+      "loss": 0.9934,
+      "step": 15908
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.758763415783875e-06,
+      "loss": 1.0205,
+      "step": 15909
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.757645431355918e-06,
+      "loss": 0.9256,
+      "step": 15910
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.756527462699215e-06,
+      "loss": 0.9446,
+      "step": 15911
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.755409509827966e-06,
+      "loss": 1.0126,
+      "step": 15912
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.754291572756353e-06,
+      "loss": 1.0793,
+      "step": 15913
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.753173651498577e-06,
+      "loss": 0.934,
+      "step": 15914
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.752055746068823e-06,
+      "loss": 1.01,
+      "step": 15915
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.750937856481287e-06,
+      "loss": 0.8721,
+      "step": 15916
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.749819982750159e-06,
+      "loss": 1.116,
+      "step": 15917
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.748702124889625e-06,
+      "loss": 0.9956,
+      "step": 15918
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.747584282913884e-06,
+      "loss": 0.9292,
+      "step": 15919
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.746466456837122e-06,
+      "loss": 0.9318,
+      "step": 15920
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.74534864667353e-06,
+      "loss": 0.8689,
+      "step": 15921
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.744230852437297e-06,
+      "loss": 1.0597,
+      "step": 15922
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.743113074142615e-06,
+      "loss": 0.8645,
+      "step": 15923
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.741995311803677e-06,
+      "loss": 0.9512,
+      "step": 15924
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.740877565434664e-06,
+      "loss": 1.0036,
+      "step": 15925
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.739759835049774e-06,
+      "loss": 0.9944,
+      "step": 15926
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.738642120663187e-06,
+      "loss": 1.0784,
+      "step": 15927
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.737524422289104e-06,
+      "loss": 0.8978,
+      "step": 15928
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.736406739941702e-06,
+      "loss": 0.8479,
+      "step": 15929
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.735289073635179e-06,
+      "loss": 0.9462,
+      "step": 15930
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.734171423383717e-06,
+      "loss": 1.0255,
+      "step": 15931
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.733053789201504e-06,
+      "loss": 0.9621,
+      "step": 15932
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.731936171102731e-06,
+      "loss": 0.9124,
+      "step": 15933
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.730818569101588e-06,
+      "loss": 0.9367,
+      "step": 15934
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.729700983212257e-06,
+      "loss": 1.0735,
+      "step": 15935
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.728583413448928e-06,
+      "loss": 1.0851,
+      "step": 15936
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72746585982579e-06,
+      "loss": 1.0362,
+      "step": 15937
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72634832235703e-06,
+      "loss": 0.9104,
+      "step": 15938
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.725230801056828e-06,
+      "loss": 0.967,
+      "step": 15939
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.72411329593938e-06,
+      "loss": 0.9373,
+      "step": 15940
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.722995807018864e-06,
+      "loss": 0.9586,
+      "step": 15941
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.721878334309471e-06,
+      "loss": 0.9632,
+      "step": 15942
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.720760877825387e-06,
+      "loss": 0.9683,
+      "step": 15943
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.719643437580796e-06,
+      "loss": 0.9441,
+      "step": 15944
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.718526013589884e-06,
+      "loss": 0.8893,
+      "step": 15945
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.717408605866835e-06,
+      "loss": 1.1126,
+      "step": 15946
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.716291214425835e-06,
+      "loss": 0.9223,
+      "step": 15947
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.715173839281073e-06,
+      "loss": 0.9603,
+      "step": 15948
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.714056480446727e-06,
+      "loss": 1.0577,
+      "step": 15949
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.712939137936982e-06,
+      "loss": 0.9274,
+      "step": 15950
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.71182181176603e-06,
+      "loss": 0.9904,
+      "step": 15951
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.710704501948049e-06,
+      "loss": 1.0008,
+      "step": 15952
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.709587208497223e-06,
+      "loss": 1.0468,
+      "step": 15953
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 8.708469931427737e-06,
+      "loss": 0.9961,
+      "step": 15954
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70735267075377e-06,
+      "loss": 0.9767,
+      "step": 15955
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.706235426489514e-06,
+      "loss": 0.9729,
+      "step": 15956
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.705118198649143e-06,
+      "loss": 0.9657,
+      "step": 15957
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.704000987246848e-06,
+      "loss": 0.9915,
+      "step": 15958
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.702883792296804e-06,
+      "loss": 0.9928,
+      "step": 15959
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.701766613813195e-06,
+      "loss": 1.037,
+      "step": 15960
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.70064945181021e-06,
+      "loss": 0.8915,
+      "step": 15961
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.699532306302026e-06,
+      "loss": 0.8916,
+      "step": 15962
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.698415177302821e-06,
+      "loss": 1.1416,
+      "step": 15963
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69729806482678e-06,
+      "loss": 0.9867,
+      "step": 15964
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.696180968888088e-06,
+      "loss": 0.8692,
+      "step": 15965
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.695063889500922e-06,
+      "loss": 1.0006,
+      "step": 15966
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.693946826679465e-06,
+      "loss": 1.009,
+      "step": 15967
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.692829780437897e-06,
+      "loss": 0.9405,
+      "step": 15968
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.691712750790394e-06,
+      "loss": 1.0732,
+      "step": 15969
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.69059573775114e-06,
+      "loss": 0.9469,
+      "step": 15970
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68947874133432e-06,
+      "loss": 1.029,
+      "step": 15971
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.688361761554109e-06,
+      "loss": 0.9194,
+      "step": 15972
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.687244798424685e-06,
+      "loss": 0.8403,
+      "step": 15973
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.686127851960226e-06,
+      "loss": 1.0083,
+      "step": 15974
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68501092217492e-06,
+      "loss": 0.984,
+      "step": 15975
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.683894009082941e-06,
+      "loss": 0.9989,
+      "step": 15976
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.682777112698463e-06,
+      "loss": 0.8799,
+      "step": 15977
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.68166023303567e-06,
+      "loss": 1.0159,
+      "step": 15978
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.680543370108743e-06,
+      "loss": 1.0041,
+      "step": 15979
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.679426523931857e-06,
+      "loss": 0.9701,
+      "step": 15980
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.678309694519187e-06,
+      "loss": 0.8836,
+      "step": 15981
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.677192881884914e-06,
+      "loss": 0.8437,
+      "step": 15982
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.676076086043214e-06,
+      "loss": 0.8718,
+      "step": 15983
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.674959307008264e-06,
+      "loss": 1.1834,
+      "step": 15984
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.673842544794248e-06,
+      "loss": 1.0638,
+      "step": 15985
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.672725799415336e-06,
+      "loss": 0.9478,
+      "step": 15986
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.671609070885702e-06,
+      "loss": 0.9056,
+      "step": 15987
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.670492359219527e-06,
+      "loss": 1.0072,
+      "step": 15988
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.669375664430991e-06,
+      "loss": 1.0401,
+      "step": 15989
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.668258986534264e-06,
+      "loss": 1.1066,
+      "step": 15990
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.667142325543522e-06,
+      "loss": 1.0072,
+      "step": 15991
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.66602568147294e-06,
+      "loss": 1.0096,
+      "step": 15992
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.664909054336701e-06,
+      "loss": 0.9726,
+      "step": 15993
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.663792444148973e-06,
+      "loss": 0.8707,
+      "step": 15994
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.662675850923935e-06,
+      "loss": 1.0471,
+      "step": 15995
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.661559274675757e-06,
+      "loss": 1.1769,
+      "step": 15996
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.660442715418616e-06,
+      "loss": 0.8986,
+      "step": 15997
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.659326173166685e-06,
+      "loss": 0.9573,
+      "step": 15998
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.658209647934144e-06,
+      "loss": 0.9475,
+      "step": 15999
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.657093139735162e-06,
+      "loss": 0.9981,
+      "step": 16000
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.65597664858391e-06,
+      "loss": 1.1466,
+      "step": 16001
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.654860174494564e-06,
+      "loss": 0.9885,
+      "step": 16002
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.653743717481302e-06,
+      "loss": 1.0807,
+      "step": 16003
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.652627277558292e-06,
+      "loss": 0.9147,
+      "step": 16004
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.651510854739706e-06,
+      "loss": 0.8695,
+      "step": 16005
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.650394449039716e-06,
+      "loss": 1.0302,
+      "step": 16006
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.649278060472501e-06,
+      "loss": 0.9196,
+      "step": 16007
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.648161689052229e-06,
+      "loss": 1.0493,
+      "step": 16008
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.647045334793071e-06,
+      "loss": 1.0542,
+      "step": 16009
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.6459289977092e-06,
+      "loss": 1.0661,
+      "step": 16010
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.644812677814784e-06,
+      "loss": 1.0343,
+      "step": 16011
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.643696375123998e-06,
+      "loss": 0.9717,
+      "step": 16012
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.642580089651015e-06,
+      "loss": 1.0763,
+      "step": 16013
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.641463821410005e-06,
+      "loss": 1.0552,
+      "step": 16014
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.640347570415132e-06,
+      "loss": 0.9996,
+      "step": 16015
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.639231336680572e-06,
+      "loss": 0.9497,
+      "step": 16016
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.638115120220498e-06,
+      "loss": 0.9538,
+      "step": 16017
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.636998921049075e-06,
+      "loss": 0.9037,
+      "step": 16018
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.635882739180475e-06,
+      "loss": 0.9899,
+      "step": 16019
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.634766574628862e-06,
+      "loss": 0.969,
+      "step": 16020
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.633650427408415e-06,
+      "loss": 0.9528,
+      "step": 16021
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.632534297533296e-06,
+      "loss": 0.98,
+      "step": 16022
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63141818501768e-06,
+      "loss": 0.9892,
+      "step": 16023
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.63030208987573e-06,
+      "loss": 1.1008,
+      "step": 16024
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.629186012121613e-06,
+      "loss": 0.974,
+      "step": 16025
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.628069951769503e-06,
+      "loss": 1.0803,
+      "step": 16026
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.626953908833566e-06,
+      "loss": 1.0547,
+      "step": 16027
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.625837883327968e-06,
+      "loss": 0.8595,
+      "step": 16028
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.624721875266878e-06,
+      "loss": 1.0563,
+      "step": 16029
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.623605884664461e-06,
+      "loss": 0.9176,
+      "step": 16030
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62248991153489e-06,
+      "loss": 0.9577,
+      "step": 16031
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.621373955892325e-06,
+      "loss": 1.0145,
+      "step": 16032
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.62025801775094e-06,
+      "loss": 1.0226,
+      "step": 16033
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61914209712489e-06,
+      "loss": 1.0632,
+      "step": 16034
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.618026194028356e-06,
+      "loss": 1.0525,
+      "step": 16035
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61691030847549e-06,
+      "loss": 1.0939,
+      "step": 16036
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61579444048047e-06,
+      "loss": 1.046,
+      "step": 16037
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.614678590057455e-06,
+      "loss": 0.9239,
+      "step": 16038
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.613562757220607e-06,
+      "loss": 0.96,
+      "step": 16039
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.612446941984095e-06,
+      "loss": 0.909,
+      "step": 16040
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.61133114436209e-06,
+      "loss": 1.0447,
+      "step": 16041
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.610215364368746e-06,
+      "loss": 1.0392,
+      "step": 16042
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.609099602018231e-06,
+      "loss": 1.0596,
+      "step": 16043
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.607983857324712e-06,
+      "loss": 0.9873,
+      "step": 16044
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.606868130302354e-06,
+      "loss": 0.9379,
+      "step": 16045
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.605752420965317e-06,
+      "loss": 1.1446,
+      "step": 16046
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.604636729327767e-06,
+      "loss": 0.9471,
+      "step": 16047
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.60352105540386e-06,
+      "loss": 0.8874,
+      "step": 16048
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.602405399207771e-06,
+      "loss": 1.0901,
+      "step": 16049
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.601289760753655e-06,
+      "loss": 0.9267,
+      "step": 16050
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.600174140055676e-06,
+      "loss": 0.9238,
+      "step": 16051
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.599058537127997e-06,
+      "loss": 1.001,
+      "step": 16052
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.59794295198478e-06,
+      "loss": 0.9898,
+      "step": 16053
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.596827384640186e-06,
+      "loss": 0.9074,
+      "step": 16054
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.595711835108383e-06,
+      "loss": 0.9178,
+      "step": 16055
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.594596303403524e-06,
+      "loss": 0.9278,
+      "step": 16056
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.593480789539771e-06,
+      "loss": 1.0684,
+      "step": 16057
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.592365293531292e-06,
+      "loss": 1.0048,
+      "step": 16058
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.591249815392242e-06,
+      "loss": 1.0102,
+      "step": 16059
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.590134355136785e-06,
+      "loss": 1.0671,
+      "step": 16060
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.589018912779078e-06,
+      "loss": 0.9463,
+      "step": 16061
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.58790348833328e-06,
+      "loss": 0.95,
+      "step": 16062
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.586788081813558e-06,
+      "loss": 0.9757,
+      "step": 16063
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.585672693234065e-06,
+      "loss": 1.0103,
+      "step": 16064
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.584557322608966e-06,
+      "loss": 0.9765,
+      "step": 16065
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.583441969952412e-06,
+      "loss": 1.1583,
+      "step": 16066
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.582326635278566e-06,
+      "loss": 0.9906,
+      "step": 16067
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.581211318601594e-06,
+      "loss": 1.004,
+      "step": 16068
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.580096019935647e-06,
+      "loss": 0.9123,
+      "step": 16069
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.578980739294884e-06,
+      "loss": 0.8956,
+      "step": 16070
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.57786547669346e-06,
+      "loss": 0.9825,
+      "step": 16071
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.576750232145541e-06,
+      "loss": 0.8097,
+      "step": 16072
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.575635005665282e-06,
+      "loss": 0.9276,
+      "step": 16073
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.574519797266834e-06,
+      "loss": 0.9273,
+      "step": 16074
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.573404606964363e-06,
+      "loss": 0.9384,
+      "step": 16075
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.572289434772016e-06,
+      "loss": 1.007,
+      "step": 16076
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.571174280703962e-06,
+      "loss": 1.0378,
+      "step": 16077
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.570059144774347e-06,
+      "loss": 0.937,
+      "step": 16078
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.568944026997334e-06,
+      "loss": 0.8182,
+      "step": 16079
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.567828927387075e-06,
+      "loss": 1.0355,
+      "step": 16080
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.566713845957725e-06,
+      "loss": 0.9633,
+      "step": 16081
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.565598782723446e-06,
+      "loss": 0.9963,
+      "step": 16082
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.564483737698388e-06,
+      "loss": 0.9099,
+      "step": 16083
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.563368710896705e-06,
+      "loss": 0.9795,
+      "step": 16084
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.562253702332553e-06,
+      "loss": 0.9986,
+      "step": 16085
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.561138712020092e-06,
+      "loss": 1.1426,
+      "step": 16086
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.56002373997347e-06,
+      "loss": 1.023,
+      "step": 16087
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.558908786206844e-06,
+      "loss": 0.9967,
+      "step": 16088
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.557793850734368e-06,
+      "loss": 0.8877,
+      "step": 16089
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.556678933570192e-06,
+      "loss": 0.8698,
+      "step": 16090
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.555564034728473e-06,
+      "loss": 1.0169,
+      "step": 16091
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.554449154223366e-06,
+      "loss": 0.8701,
+      "step": 16092
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.553334292069022e-06,
+      "loss": 1.0031,
+      "step": 16093
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.552219448279591e-06,
+      "loss": 0.9943,
+      "step": 16094
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.551104622869228e-06,
+      "loss": 0.7788,
+      "step": 16095
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.549989815852087e-06,
+      "loss": 0.9797,
+      "step": 16096
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.54887502724232e-06,
+      "loss": 1.1465,
+      "step": 16097
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.547760257054074e-06,
+      "loss": 0.9074,
+      "step": 16098
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.546645505301503e-06,
+      "loss": 0.9887,
+      "step": 16099
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.545530771998762e-06,
+      "loss": 1.0528,
+      "step": 16100
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.544416057159999e-06,
+      "loss": 0.9665,
+      "step": 16101
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.543301360799367e-06,
+      "loss": 0.9024,
+      "step": 16102
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.542186682931014e-06,
+      "loss": 0.9272,
+      "step": 16103
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.541072023569089e-06,
+      "loss": 1.0189,
+      "step": 16104
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.539957382727746e-06,
+      "loss": 0.9975,
+      "step": 16105
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.538842760421135e-06,
+      "loss": 1.0378,
+      "step": 16106
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.537728156663405e-06,
+      "loss": 0.9815,
+      "step": 16107
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.536613571468704e-06,
+      "loss": 1.1463,
+      "step": 16108
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.535499004851178e-06,
+      "loss": 0.9619,
+      "step": 16109
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.534384456824988e-06,
+      "loss": 0.9519,
+      "step": 16110
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.53326992740427e-06,
+      "loss": 0.9823,
+      "step": 16111
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.532155416603178e-06,
+      "loss": 0.8873,
+      "step": 16112
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.531040924435858e-06,
+      "loss": 1.0014,
+      "step": 16113
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.529926450916465e-06,
+      "loss": 1.0398,
+      "step": 16114
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.528811996059138e-06,
+      "loss": 0.9244,
+      "step": 16115
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.527697559878032e-06,
+      "loss": 1.0461,
+      "step": 16116
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.52658314238729e-06,
+      "loss": 0.9875,
+      "step": 16117
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.525468743601057e-06,
+      "loss": 1.0624,
+      "step": 16118
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.524354363533483e-06,
+      "loss": 0.8572,
+      "step": 16119
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.523240002198719e-06,
+      "loss": 0.8778,
+      "step": 16120
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.522125659610905e-06,
+      "loss": 1.0456,
+      "step": 16121
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.521011335784188e-06,
+      "loss": 0.9519,
+      "step": 16122
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.519897030732713e-06,
+      "loss": 0.8783,
+      "step": 16123
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.518782744470632e-06,
+      "loss": 0.843,
+      "step": 16124
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.517668477012085e-06,
+      "loss": 0.9334,
+      "step": 16125
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51655422837122e-06,
+      "loss": 0.9619,
+      "step": 16126
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.515439998562174e-06,
+      "loss": 0.8952,
+      "step": 16127
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.514325787599106e-06,
+      "loss": 1.0384,
+      "step": 16128
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.51321159549615e-06,
+      "loss": 1.0144,
+      "step": 16129
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.512097422267454e-06,
+      "loss": 1.1351,
+      "step": 16130
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.510983267927162e-06,
+      "loss": 0.9561,
+      "step": 16131
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.509869132489412e-06,
+      "loss": 0.9317,
+      "step": 16132
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.508755015968356e-06,
+      "loss": 1.161,
+      "step": 16133
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.507640918378133e-06,
+      "loss": 1.1158,
+      "step": 16134
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.506526839732889e-06,
+      "loss": 1.0527,
+      "step": 16135
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.505412780046764e-06,
+      "loss": 1.0071,
+      "step": 16136
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.504298739333898e-06,
+      "loss": 0.9674,
+      "step": 16137
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.503184717608439e-06,
+      "loss": 1.0585,
+      "step": 16138
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.502070714884527e-06,
+      "loss": 0.8601,
+      "step": 16139
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.500956731176306e-06,
+      "loss": 1.0526,
+      "step": 16140
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49984276649791e-06,
+      "loss": 1.0367,
+      "step": 16141
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49872882086349e-06,
+      "loss": 0.9774,
+      "step": 16142
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49761489428718e-06,
+      "loss": 0.7961,
+      "step": 16143
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.496500986783128e-06,
+      "loss": 0.9905,
+      "step": 16144
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49538709836547e-06,
+      "loss": 0.9421,
+      "step": 16145
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.494273229048341e-06,
+      "loss": 0.9954,
+      "step": 16146
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.49315937884589e-06,
+      "loss": 0.9676,
+      "step": 16147
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.492045547772257e-06,
+      "loss": 0.8713,
+      "step": 16148
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.490931735841577e-06,
+      "loss": 0.9035,
+      "step": 16149
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.489817943067989e-06,
+      "loss": 0.9159,
+      "step": 16150
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.488704169465636e-06,
+      "loss": 1.0076,
+      "step": 16151
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.487590415048657e-06,
+      "loss": 0.9127,
+      "step": 16152
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.486476679831186e-06,
+      "loss": 0.8514,
+      "step": 16153
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.485362963827367e-06,
+      "loss": 0.9877,
+      "step": 16154
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.484249267051333e-06,
+      "loss": 1.0172,
+      "step": 16155
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.483135589517226e-06,
+      "loss": 0.9705,
+      "step": 16156
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.482021931239183e-06,
+      "loss": 1.0421,
+      "step": 16157
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.480908292231343e-06,
+      "loss": 0.9697,
+      "step": 16158
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.479794672507838e-06,
+      "loss": 1.0991,
+      "step": 16159
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.478681072082809e-06,
+      "loss": 0.9223,
+      "step": 16160
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.477567490970391e-06,
+      "loss": 1.0312,
+      "step": 16161
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.476453929184725e-06,
+      "loss": 1.0148,
+      "step": 16162
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47534038673994e-06,
+      "loss": 0.916,
+      "step": 16163
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.474226863650176e-06,
+      "loss": 0.9823,
+      "step": 16164
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.47311335992957e-06,
+      "loss": 1.0325,
+      "step": 16165
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.471999875592259e-06,
+      "loss": 0.9392,
+      "step": 16166
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.470886410652373e-06,
+      "loss": 0.9375,
+      "step": 16167
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.469772965124051e-06,
+      "loss": 0.9849,
+      "step": 16168
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.468659539021422e-06,
+      "loss": 0.9342,
+      "step": 16169
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46754613235863e-06,
+      "loss": 0.9461,
+      "step": 16170
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.4664327451498e-06,
+      "loss": 1.0883,
+      "step": 16171
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.465319377409074e-06,
+      "loss": 1.1911,
+      "step": 16172
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46420602915058e-06,
+      "loss": 0.9613,
+      "step": 16173
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.46309270038845e-06,
+      "loss": 1.0088,
+      "step": 16174
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.461979391136827e-06,
+      "loss": 0.9859,
+      "step": 16175
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.460866101409838e-06,
+      "loss": 1.0188,
+      "step": 16176
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.459752831221612e-06,
+      "loss": 1.0033,
+      "step": 16177
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.458639580586285e-06,
+      "loss": 0.9717,
+      "step": 16178
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.457526349517994e-06,
+      "loss": 1.0587,
+      "step": 16179
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.456413138030866e-06,
+      "loss": 1.0909,
+      "step": 16180
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.455299946139032e-06,
+      "loss": 1.0842,
+      "step": 16181
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.454186773856627e-06,
+      "loss": 1.0042,
+      "step": 16182
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.453073621197777e-06,
+      "loss": 0.8576,
+      "step": 16183
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.451960488176621e-06,
+      "loss": 0.8813,
+      "step": 16184
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.450847374807282e-06,
+      "loss": 1.0674,
+      "step": 16185
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.449734281103897e-06,
+      "loss": 0.8939,
+      "step": 16186
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.448621207080592e-06,
+      "loss": 1.0011,
+      "step": 16187
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.447508152751496e-06,
+      "loss": 1.0134,
+      "step": 16188
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.446395118130744e-06,
+      "loss": 1.0172,
+      "step": 16189
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.445282103232465e-06,
+      "loss": 1.0057,
+      "step": 16190
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.444169108070784e-06,
+      "loss": 0.9239,
+      "step": 16191
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.443056132659829e-06,
+      "loss": 1.0148,
+      "step": 16192
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.441943177013737e-06,
+      "loss": 0.9584,
+      "step": 16193
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.440830241146632e-06,
+      "loss": 1.0288,
+      "step": 16194
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.43971732507264e-06,
+      "loss": 0.9807,
+      "step": 16195
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.438604428805893e-06,
+      "loss": 0.927,
+      "step": 16196
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.437491552360512e-06,
+      "loss": 1.0481,
+      "step": 16197
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.436378695750634e-06,
+      "loss": 0.9869,
+      "step": 16198
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.435265858990382e-06,
+      "loss": 1.0774,
+      "step": 16199
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.434153042093884e-06,
+      "loss": 0.8862,
+      "step": 16200
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.433040245075263e-06,
+      "loss": 0.9596,
+      "step": 16201
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.431927467948648e-06,
+      "loss": 1.0169,
+      "step": 16202
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.430814710728167e-06,
+      "loss": 1.0451,
+      "step": 16203
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.429701973427947e-06,
+      "loss": 0.8389,
+      "step": 16204
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42858925606211e-06,
+      "loss": 1.027,
+      "step": 16205
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42747655864478e-06,
+      "loss": 0.9108,
+      "step": 16206
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.426363881190088e-06,
+      "loss": 0.994,
+      "step": 16207
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.425251223712158e-06,
+      "loss": 1.0524,
+      "step": 16208
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.424138586225111e-06,
+      "loss": 0.9651,
+      "step": 16209
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.423025968743076e-06,
+      "loss": 0.8372,
+      "step": 16210
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.42191337128017e-06,
+      "loss": 1.0486,
+      "step": 16211
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.420800793850526e-06,
+      "loss": 1.1016,
+      "step": 16212
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.419688236468267e-06,
+      "loss": 1.0598,
+      "step": 16213
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.418575699147511e-06,
+      "loss": 1.0211,
+      "step": 16214
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.417463181902383e-06,
+      "loss": 1.1214,
+      "step": 16215
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.416350684747004e-06,
+      "loss": 0.9689,
+      "step": 16216
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.415238207695504e-06,
+      "loss": 0.9302,
+      "step": 16217
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.414125750762e-06,
+      "loss": 1.0129,
+      "step": 16218
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.413013313960613e-06,
+      "loss": 1.14,
+      "step": 16219
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.411900897305466e-06,
+      "loss": 1.0061,
+      "step": 16220
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.410788500810688e-06,
+      "loss": 0.9868,
+      "step": 16221
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.409676124490388e-06,
+      "loss": 0.8815,
+      "step": 16222
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.408563768358699e-06,
+      "loss": 0.9723,
+      "step": 16223
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.407451432429737e-06,
+      "loss": 0.904,
+      "step": 16224
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.406339116717618e-06,
+      "loss": 1.0786,
+      "step": 16225
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.405226821236467e-06,
+      "loss": 0.9589,
+      "step": 16226
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.404114546000409e-06,
+      "loss": 0.9097,
+      "step": 16227
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.403002291023557e-06,
+      "loss": 0.9537,
+      "step": 16228
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.401890056320031e-06,
+      "loss": 0.9453,
+      "step": 16229
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.40077784190395e-06,
+      "loss": 0.9375,
+      "step": 16230
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39966564778944e-06,
+      "loss": 0.9669,
+      "step": 16231
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.398553473990612e-06,
+      "loss": 1.0012,
+      "step": 16232
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.397441320521591e-06,
+      "loss": 0.8728,
+      "step": 16233
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.396329187396489e-06,
+      "loss": 0.987,
+      "step": 16234
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.39521707462943e-06,
+      "loss": 0.9824,
+      "step": 16235
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.394104982234528e-06,
+      "loss": 1.0384,
+      "step": 16236
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.392992910225904e-06,
+      "loss": 0.9355,
+      "step": 16237
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.391880858617673e-06,
+      "loss": 0.9522,
+      "step": 16238
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.390768827423948e-06,
+      "loss": 0.9974,
+      "step": 16239
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.389656816658854e-06,
+      "loss": 1.018,
+      "step": 16240
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 8.388544826336503e-06,
+      "loss": 1.0874,
+      "step": 16241
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.387432856471013e-06,
+      "loss": 1.0626,
+      "step": 16242
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.3863209070765e-06,
+      "loss": 1.0769,
+      "step": 16243
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.385208978167074e-06,
+      "loss": 0.8739,
+      "step": 16244
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.38409706975686e-06,
+      "loss": 0.9062,
+      "step": 16245
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.382985181859968e-06,
+      "loss": 0.9664,
+      "step": 16246
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.381873314490516e-06,
+      "loss": 0.9807,
+      "step": 16247
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.380761467662612e-06,
+      "loss": 0.8764,
+      "step": 16248
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.379649641390379e-06,
+      "loss": 0.8329,
+      "step": 16249
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.378537835687926e-06,
+      "loss": 0.9812,
+      "step": 16250
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37742605056937e-06,
+      "loss": 0.8952,
+      "step": 16251
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.376314286048823e-06,
+      "loss": 0.9901,
+      "step": 16252
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.375202542140395e-06,
+      "loss": 0.8953,
+      "step": 16253
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.374090818858203e-06,
+      "loss": 0.9213,
+      "step": 16254
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.372979116216365e-06,
+      "loss": 0.97,
+      "step": 16255
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.371867434228984e-06,
+      "loss": 0.9211,
+      "step": 16256
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.37075577291018e-06,
+      "loss": 0.9875,
+      "step": 16257
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.369644132274057e-06,
+      "loss": 0.886,
+      "step": 16258
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.368532512334738e-06,
+      "loss": 1.0139,
+      "step": 16259
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.367420913106323e-06,
+      "loss": 1.0074,
+      "step": 16260
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.366309334602933e-06,
+      "loss": 0.9655,
+      "step": 16261
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.36519777683867e-06,
+      "loss": 0.9789,
+      "step": 16262
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.364086239827655e-06,
+      "loss": 0.9646,
+      "step": 16263
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.362974723583992e-06,
+      "loss": 1.0183,
+      "step": 16264
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.361863228121794e-06,
+      "loss": 0.8873,
+      "step": 16265
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.360751753455169e-06,
+      "loss": 1.0464,
+      "step": 16266
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.359640299598223e-06,
+      "loss": 0.9412,
+      "step": 16267
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.358528866565073e-06,
+      "loss": 0.9702,
+      "step": 16268
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.357417454369827e-06,
+      "loss": 0.9034,
+      "step": 16269
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.356306063026591e-06,
+      "loss": 0.9231,
+      "step": 16270
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.355194692549477e-06,
+      "loss": 0.9542,
+      "step": 16271
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.354083342952585e-06,
+      "loss": 1.0572,
+      "step": 16272
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.352972014250036e-06,
+      "loss": 0.8681,
+      "step": 16273
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.35186070645593e-06,
+      "loss": 0.9969,
+      "step": 16274
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.350749419584378e-06,
+      "loss": 1.0787,
+      "step": 16275
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.349638153649479e-06,
+      "loss": 0.9887,
+      "step": 16276
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.348526908665354e-06,
+      "loss": 1.0103,
+      "step": 16277
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.3474156846461e-06,
+      "loss": 0.8848,
+      "step": 16278
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.346304481605826e-06,
+      "loss": 0.8059,
+      "step": 16279
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.34519329955864e-06,
+      "loss": 1.0086,
+      "step": 16280
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.344082138518645e-06,
+      "loss": 1.0604,
+      "step": 16281
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.342970998499947e-06,
+      "loss": 0.9075,
+      "step": 16282
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.341859879516657e-06,
+      "loss": 0.9343,
+      "step": 16283
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.340748781582873e-06,
+      "loss": 0.989,
+      "step": 16284
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.339637704712706e-06,
+      "loss": 1.0388,
+      "step": 16285
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.338526648920255e-06,
+      "loss": 0.9454,
+      "step": 16286
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33741561421963e-06,
+      "loss": 0.8178,
+      "step": 16287
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33630460062493e-06,
+      "loss": 1.0324,
+      "step": 16288
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.335193608150264e-06,
+      "loss": 0.9822,
+      "step": 16289
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33408263680973e-06,
+      "loss": 0.911,
+      "step": 16290
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.33297168661744e-06,
+      "loss": 1.0208,
+      "step": 16291
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.331860757587487e-06,
+      "loss": 0.948,
+      "step": 16292
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.330749849733982e-06,
+      "loss": 0.8939,
+      "step": 16293
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.329638963071021e-06,
+      "loss": 0.9295,
+      "step": 16294
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.328528097612709e-06,
+      "loss": 1.0696,
+      "step": 16295
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.327417253373153e-06,
+      "loss": 1.0027,
+      "step": 16296
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.32630643036645e-06,
+      "loss": 0.9438,
+      "step": 16297
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.325195628606699e-06,
+      "loss": 0.9524,
+      "step": 16298
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.324084848108e-06,
+      "loss": 1.1126,
+      "step": 16299
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.322974088884467e-06,
+      "loss": 1.1205,
+      "step": 16300
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.321863350950188e-06,
+      "loss": 0.9926,
+      "step": 16301
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.320752634319266e-06,
+      "loss": 0.9417,
+      "step": 16302
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.319641939005805e-06,
+      "loss": 1.0731,
+      "step": 16303
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.318531265023897e-06,
+      "loss": 0.9262,
+      "step": 16304
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31742061238765e-06,
+      "loss": 0.8833,
+      "step": 16305
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.316309981111161e-06,
+      "loss": 0.9712,
+      "step": 16306
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.31519937120853e-06,
+      "loss": 1.0212,
+      "step": 16307
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.314088782693852e-06,
+      "loss": 1.0665,
+      "step": 16308
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.312978215581225e-06,
+      "loss": 0.9642,
+      "step": 16309
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.311867669884753e-06,
+      "loss": 0.9111,
+      "step": 16310
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.310757145618532e-06,
+      "loss": 1.0117,
+      "step": 16311
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.309646642796656e-06,
+      "loss": 0.8541,
+      "step": 16312
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.308536161433223e-06,
+      "loss": 0.939,
+      "step": 16313
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.307425701542336e-06,
+      "loss": 0.9076,
+      "step": 16314
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30631526313809e-06,
+      "loss": 0.996,
+      "step": 16315
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.305204846234575e-06,
+      "loss": 0.941,
+      "step": 16316
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.304094450845896e-06,
+      "loss": 0.9409,
+      "step": 16317
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.30298407698614e-06,
+      "loss": 0.9619,
+      "step": 16318
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.301873724669411e-06,
+      "loss": 0.8958,
+      "step": 16319
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.300763393909803e-06,
+      "loss": 1.1827,
+      "step": 16320
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29965308472141e-06,
+      "loss": 0.9494,
+      "step": 16321
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.298542797118324e-06,
+      "loss": 0.9364,
+      "step": 16322
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.29743253111464e-06,
+      "loss": 0.9973,
+      "step": 16323
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.296322286724462e-06,
+      "loss": 1.0148,
+      "step": 16324
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.295212063961875e-06,
+      "loss": 1.0145,
+      "step": 16325
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.294101862840973e-06,
+      "loss": 1.0,
+      "step": 16326
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.292991683375848e-06,
+      "loss": 0.8982,
+      "step": 16327
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.291881525580602e-06,
+      "loss": 1.0538,
+      "step": 16328
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.290771389469323e-06,
+      "loss": 1.0389,
+      "step": 16329
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.289661275056104e-06,
+      "loss": 1.0263,
+      "step": 16330
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.288551182355038e-06,
+      "loss": 1.0272,
+      "step": 16331
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.287441111380211e-06,
+      "loss": 1.0412,
+      "step": 16332
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.286331062145724e-06,
+      "loss": 0.9867,
+      "step": 16333
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.285221034665668e-06,
+      "loss": 0.9001,
+      "step": 16334
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28411102895413e-06,
+      "loss": 0.9722,
+      "step": 16335
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.2830010450252e-06,
+      "loss": 1.0369,
+      "step": 16336
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.28189108289297e-06,
+      "loss": 0.8994,
+      "step": 16337
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.280781142571537e-06,
+      "loss": 1.0443,
+      "step": 16338
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.279671224074987e-06,
+      "loss": 1.0057,
+      "step": 16339
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.278561327417407e-06,
+      "loss": 0.9223,
+      "step": 16340
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.277451452612886e-06,
+      "loss": 0.9831,
+      "step": 16341
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.276341599675521e-06,
+      "loss": 0.8769,
+      "step": 16342
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.275231768619395e-06,
+      "loss": 0.9497,
+      "step": 16343
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.274121959458604e-06,
+      "loss": 0.9143,
+      "step": 16344
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.273012172207227e-06,
+      "loss": 0.9103,
+      "step": 16345
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.271902406879356e-06,
+      "loss": 0.9067,
+      "step": 16346
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.27079266348908e-06,
+      "loss": 0.991,
+      "step": 16347
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26968294205049e-06,
+      "loss": 1.001,
+      "step": 16348
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.268573242577668e-06,
+      "loss": 0.9785,
+      "step": 16349
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.267463565084704e-06,
+      "loss": 0.9928,
+      "step": 16350
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.26635390958568e-06,
+      "loss": 0.9102,
+      "step": 16351
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.265244276094694e-06,
+      "loss": 0.9755,
+      "step": 16352
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.264134664625822e-06,
+      "loss": 0.9726,
+      "step": 16353
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.263025075193156e-06,
+      "loss": 0.8983,
+      "step": 16354
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.261915507810776e-06,
+      "loss": 0.8848,
+      "step": 16355
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.260805962492773e-06,
+      "loss": 0.9893,
+      "step": 16356
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.25969643925323e-06,
+      "loss": 1.0098,
+      "step": 16357
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.258586938106236e-06,
+      "loss": 0.9241,
+      "step": 16358
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.257477459065871e-06,
+      "loss": 1.0176,
+      "step": 16359
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.256368002146214e-06,
+      "loss": 1.099,
+      "step": 16360
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.255258567361361e-06,
+      "loss": 1.0469,
+      "step": 16361
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.254149154725391e-06,
+      "loss": 1.1258,
+      "step": 16362
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.253039764252386e-06,
+      "loss": 0.9957,
+      "step": 16363
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.251930395956432e-06,
+      "loss": 1.101,
+      "step": 16364
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.250821049851607e-06,
+      "loss": 0.859,
+      "step": 16365
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.249711725952002e-06,
+      "loss": 1.0568,
+      "step": 16366
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.248602424271691e-06,
+      "loss": 0.9768,
+      "step": 16367
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.247493144824763e-06,
+      "loss": 1.0239,
+      "step": 16368
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.246383887625293e-06,
+      "loss": 1.1097,
+      "step": 16369
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.245274652687371e-06,
+      "loss": 1.1004,
+      "step": 16370
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.244165440025072e-06,
+      "loss": 0.8957,
+      "step": 16371
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.24305624965248e-06,
+      "loss": 0.9804,
+      "step": 16372
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.241947081583676e-06,
+      "loss": 1.0018,
+      "step": 16373
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.240837935832734e-06,
+      "loss": 1.0113,
+      "step": 16374
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.239728812413742e-06,
+      "loss": 0.8972,
+      "step": 16375
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.238619711340781e-06,
+      "loss": 0.9457,
+      "step": 16376
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.237510632627923e-06,
+      "loss": 0.9738,
+      "step": 16377
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.236401576289254e-06,
+      "loss": 1.0004,
+      "step": 16378
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.235292542338847e-06,
+      "loss": 0.8464,
+      "step": 16379
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.234183530790787e-06,
+      "loss": 1.0737,
+      "step": 16380
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.23307454165915e-06,
+      "loss": 0.9998,
+      "step": 16381
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.231965574958015e-06,
+      "loss": 0.8569,
+      "step": 16382
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.230856630701455e-06,
+      "loss": 0.965,
+      "step": 16383
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.229747708903556e-06,
+      "loss": 0.9299,
+      "step": 16384
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.228638809578389e-06,
+      "loss": 0.9723,
+      "step": 16385
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.227529932740035e-06,
+      "loss": 0.7845,
+      "step": 16386
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.226421078402566e-06,
+      "loss": 1.013,
+      "step": 16387
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.225312246580065e-06,
+      "loss": 1.1548,
+      "step": 16388
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.224203437286602e-06,
+      "loss": 0.8549,
+      "step": 16389
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.223094650536257e-06,
+      "loss": 1.2545,
+      "step": 16390
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.221985886343102e-06,
+      "loss": 0.9577,
+      "step": 16391
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.220877144721219e-06,
+      "loss": 1.0548,
+      "step": 16392
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.219768425684673e-06,
+      "loss": 1.0696,
+      "step": 16393
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.218659729247548e-06,
+      "loss": 0.9923,
+      "step": 16394
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.217551055423915e-06,
+      "loss": 0.9605,
+      "step": 16395
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.216442404227851e-06,
+      "loss": 0.9913,
+      "step": 16396
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.215333775673421e-06,
+      "loss": 0.9915,
+      "step": 16397
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21422516977471e-06,
+      "loss": 0.9107,
+      "step": 16398
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.213116586545784e-06,
+      "loss": 1.0265,
+      "step": 16399
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.21200802600072e-06,
+      "loss": 0.9205,
+      "step": 16400
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.210899488153587e-06,
+      "loss": 0.8583,
+      "step": 16401
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20979097301846e-06,
+      "loss": 0.997,
+      "step": 16402
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.208682480609412e-06,
+      "loss": 1.0504,
+      "step": 16403
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.207574010940516e-06,
+      "loss": 0.8996,
+      "step": 16404
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.206465564025837e-06,
+      "loss": 0.882,
+      "step": 16405
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.205357139879455e-06,
+      "loss": 1.0161,
+      "step": 16406
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.204248738515431e-06,
+      "loss": 1.0054,
+      "step": 16407
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.203140359947846e-06,
+      "loss": 0.9664,
+      "step": 16408
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.202032004190765e-06,
+      "loss": 0.927,
+      "step": 16409
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.20092367125826e-06,
+      "loss": 0.9722,
+      "step": 16410
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.199815361164396e-06,
+      "loss": 0.9828,
+      "step": 16411
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.198707073923253e-06,
+      "loss": 0.9925,
+      "step": 16412
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19759880954889e-06,
+      "loss": 0.9328,
+      "step": 16413
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.196490568055385e-06,
+      "loss": 0.9329,
+      "step": 16414
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.195382349456797e-06,
+      "loss": 1.0134,
+      "step": 16415
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.194274153767196e-06,
+      "loss": 0.9716,
+      "step": 16416
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.193165981000661e-06,
+      "loss": 0.9286,
+      "step": 16417
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.19205783117125e-06,
+      "loss": 0.9187,
+      "step": 16418
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.190949704293032e-06,
+      "loss": 1.1636,
+      "step": 16419
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.189841600380078e-06,
+      "loss": 0.9433,
+      "step": 16420
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.188733519446445e-06,
+      "loss": 1.0038,
+      "step": 16421
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.187625461506214e-06,
+      "loss": 1.038,
+      "step": 16422
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.186517426573441e-06,
+      "loss": 1.0378,
+      "step": 16423
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.185409414662197e-06,
+      "loss": 0.8253,
+      "step": 16424
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.184301425786542e-06,
+      "loss": 0.9897,
+      "step": 16425
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.18319345996055e-06,
+      "loss": 1.0945,
+      "step": 16426
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.182085517198282e-06,
+      "loss": 0.9841,
+      "step": 16427
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.180977597513804e-06,
+      "loss": 0.9524,
+      "step": 16428
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.179869700921176e-06,
+      "loss": 0.8447,
+      "step": 16429
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.178761827434465e-06,
+      "loss": 0.9092,
+      "step": 16430
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17765397706774e-06,
+      "loss": 1.0089,
+      "step": 16431
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.176546149835061e-06,
+      "loss": 0.9883,
+      "step": 16432
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.175438345750487e-06,
+      "loss": 1.0811,
+      "step": 16433
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.174330564828088e-06,
+      "loss": 0.9984,
+      "step": 16434
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.173222807081925e-06,
+      "loss": 1.0029,
+      "step": 16435
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.17211507252606e-06,
+      "loss": 1.0575,
+      "step": 16436
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.171007361174555e-06,
+      "loss": 1.0552,
+      "step": 16437
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.169899673041472e-06,
+      "loss": 1.0312,
+      "step": 16438
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.168792008140871e-06,
+      "loss": 1.1162,
+      "step": 16439
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.167684366486817e-06,
+      "loss": 1.0213,
+      "step": 16440
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.166576748093372e-06,
+      "loss": 1.0434,
+      "step": 16441
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.165469152974593e-06,
+      "loss": 1.1136,
+      "step": 16442
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.164361581144539e-06,
+      "loss": 0.984,
+      "step": 16443
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.163254032617273e-06,
+      "loss": 0.9528,
+      "step": 16444
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.162146507406859e-06,
+      "loss": 1.015,
+      "step": 16445
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.161039005527351e-06,
+      "loss": 0.9686,
+      "step": 16446
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15993152699281e-06,
+      "loss": 0.9902,
+      "step": 16447
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.15882407181729e-06,
+      "loss": 0.8512,
+      "step": 16448
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.157716640014863e-06,
+      "loss": 1.0388,
+      "step": 16449
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.156609231599574e-06,
+      "loss": 1.0027,
+      "step": 16450
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.155501846585491e-06,
+      "loss": 1.0234,
+      "step": 16451
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.154394484986666e-06,
+      "loss": 1.019,
+      "step": 16452
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.153287146817154e-06,
+      "loss": 1.0797,
+      "step": 16453
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.152179832091019e-06,
+      "loss": 0.9148,
+      "step": 16454
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.151072540822316e-06,
+      "loss": 0.9786,
+      "step": 16455
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.149965273025101e-06,
+      "loss": 1.0551,
+      "step": 16456
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.148858028713429e-06,
+      "loss": 1.0837,
+      "step": 16457
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.147750807901353e-06,
+      "loss": 0.962,
+      "step": 16458
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14664361060294e-06,
+      "loss": 0.9538,
+      "step": 16459
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.145536436832235e-06,
+      "loss": 1.0506,
+      "step": 16460
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.144429286603298e-06,
+      "loss": 0.8438,
+      "step": 16461
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14332215993018e-06,
+      "loss": 0.969,
+      "step": 16462
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.142215056826943e-06,
+      "loss": 0.9811,
+      "step": 16463
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.141107977307634e-06,
+      "loss": 0.976,
+      "step": 16464
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.14000092138631e-06,
+      "loss": 1.0254,
+      "step": 16465
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.138893889077025e-06,
+      "loss": 1.0004,
+      "step": 16466
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.137786880393828e-06,
+      "loss": 0.9606,
+      "step": 16467
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.136679895350777e-06,
+      "loss": 0.9451,
+      "step": 16468
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.135572933961927e-06,
+      "loss": 0.941,
+      "step": 16469
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.134465996241325e-06,
+      "loss": 0.9304,
+      "step": 16470
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.133359082203023e-06,
+      "loss": 1.0017,
+      "step": 16471
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.132252191861074e-06,
+      "loss": 1.0297,
+      "step": 16472
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.131145325229534e-06,
+      "loss": 1.016,
+      "step": 16473
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.130038482322448e-06,
+      "loss": 0.9755,
+      "step": 16474
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.128931663153874e-06,
+      "loss": 0.9749,
+      "step": 16475
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.127824867737854e-06,
+      "loss": 0.8777,
+      "step": 16476
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.126718096088446e-06,
+      "loss": 0.9199,
+      "step": 16477
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.125611348219696e-06,
+      "loss": 1.0712,
+      "step": 16478
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.124504624145656e-06,
+      "loss": 0.9692,
+      "step": 16479
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.123397923880373e-06,
+      "loss": 1.0134,
+      "step": 16480
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.122291247437896e-06,
+      "loss": 0.9187,
+      "step": 16481
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.121184594832276e-06,
+      "loss": 0.9262,
+      "step": 16482
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.120077966077562e-06,
+      "loss": 1.0048,
+      "step": 16483
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.1189713611878e-06,
+      "loss": 0.9626,
+      "step": 16484
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.117864780177042e-06,
+      "loss": 0.8654,
+      "step": 16485
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.116758223059328e-06,
+      "loss": 1.0646,
+      "step": 16486
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.115651689848713e-06,
+      "loss": 0.8762,
+      "step": 16487
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.114545180559242e-06,
+      "loss": 1.025,
+      "step": 16488
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.113438695204961e-06,
+      "loss": 0.9065,
+      "step": 16489
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11233223379991e-06,
+      "loss": 1.0093,
+      "step": 16490
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.11122579635815e-06,
+      "loss": 0.8979,
+      "step": 16491
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.110119382893714e-06,
+      "loss": 0.9298,
+      "step": 16492
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.109012993420653e-06,
+      "loss": 0.978,
+      "step": 16493
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10790662795301e-06,
+      "loss": 1.0582,
+      "step": 16494
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.106800286504832e-06,
+      "loss": 0.9807,
+      "step": 16495
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.105693969090161e-06,
+      "loss": 0.962,
+      "step": 16496
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.104587675723047e-06,
+      "loss": 1.147,
+      "step": 16497
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.103481406417526e-06,
+      "loss": 1.0959,
+      "step": 16498
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10237516118765e-06,
+      "loss": 1.0083,
+      "step": 16499
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.10126894004745e-06,
+      "loss": 0.8786,
+      "step": 16500
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.100162743010984e-06,
+      "loss": 0.926,
+      "step": 16501
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.099056570092286e-06,
+      "loss": 0.9681,
+      "step": 16502
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.097950421305402e-06,
+      "loss": 0.8589,
+      "step": 16503
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.096844296664368e-06,
+      "loss": 0.9934,
+      "step": 16504
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.095738196183234e-06,
+      "loss": 0.9417,
+      "step": 16505
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.094632119876038e-06,
+      "loss": 0.9167,
+      "step": 16506
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.093526067756821e-06,
+      "loss": 0.9016,
+      "step": 16507
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.092420039839624e-06,
+      "loss": 1.0554,
+      "step": 16508
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.091314036138484e-06,
+      "loss": 1.0143,
+      "step": 16509
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.09020805666745e-06,
+      "loss": 0.9815,
+      "step": 16510
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.089102101440558e-06,
+      "loss": 1.0446,
+      "step": 16511
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.087996170471844e-06,
+      "loss": 1.0804,
+      "step": 16512
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.086890263775353e-06,
+      "loss": 0.9874,
+      "step": 16513
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.085784381365117e-06,
+      "loss": 1.0911,
+      "step": 16514
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.084678523255183e-06,
+      "loss": 0.8586,
+      "step": 16515
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.083572689459583e-06,
+      "loss": 0.8291,
+      "step": 16516
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.082466879992361e-06,
+      "loss": 1.0681,
+      "step": 16517
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.081361094867546e-06,
+      "loss": 1.085,
+      "step": 16518
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.080255334099188e-06,
+      "loss": 1.0262,
+      "step": 16519
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.079149597701314e-06,
+      "loss": 0.9054,
+      "step": 16520
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.078043885687967e-06,
+      "loss": 0.9484,
+      "step": 16521
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.076938198073179e-06,
+      "loss": 0.9941,
+      "step": 16522
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.075832534870985e-06,
+      "loss": 0.8324,
+      "step": 16523
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.07472689609543e-06,
+      "loss": 1.004,
+      "step": 16524
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.073621281760544e-06,
+      "loss": 0.8279,
+      "step": 16525
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.072515691880359e-06,
+      "loss": 0.8761,
+      "step": 16526
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.071410126468917e-06,
+      "loss": 0.953,
+      "step": 16527
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.070304585540245e-06,
+      "loss": 0.9328,
+      "step": 16528
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 8.069199069108386e-06,
+      "loss": 1.0179,
+      "step": 16529
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.068093577187367e-06,
+      "loss": 1.0187,
+      "step": 16530
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.066988109791228e-06,
+      "loss": 1.008,
+      "step": 16531
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.065882666933994e-06,
+      "loss": 1.1635,
+      "step": 16532
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.064777248629706e-06,
+      "loss": 1.0207,
+      "step": 16533
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.063671854892398e-06,
+      "loss": 1.0374,
+      "step": 16534
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.062566485736097e-06,
+      "loss": 1.0125,
+      "step": 16535
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.061461141174836e-06,
+      "loss": 1.1211,
+      "step": 16536
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.060355821222646e-06,
+      "loss": 0.9731,
+      "step": 16537
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.059250525893566e-06,
+      "loss": 0.9375,
+      "step": 16538
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05814525520162e-06,
+      "loss": 1.1075,
+      "step": 16539
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.05704000916084e-06,
+      "loss": 0.9146,
+      "step": 16540
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.055934787785261e-06,
+      "loss": 0.9324,
+      "step": 16541
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.054829591088905e-06,
+      "loss": 0.9515,
+      "step": 16542
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.053724419085811e-06,
+      "loss": 0.964,
+      "step": 16543
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.052619271790005e-06,
+      "loss": 1.1047,
+      "step": 16544
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.051514149215519e-06,
+      "loss": 0.9312,
+      "step": 16545
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.050409051376375e-06,
+      "loss": 1.0868,
+      "step": 16546
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.049303978286607e-06,
+      "loss": 1.0017,
+      "step": 16547
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.048198929960247e-06,
+      "loss": 1.0126,
+      "step": 16548
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04709390641132e-06,
+      "loss": 0.9504,
+      "step": 16549
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.04598890765385e-06,
+      "loss": 0.9678,
+      "step": 16550
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.044883933701867e-06,
+      "loss": 0.9199,
+      "step": 16551
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.043778984569401e-06,
+      "loss": 0.9275,
+      "step": 16552
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.042674060270478e-06,
+      "loss": 0.9512,
+      "step": 16553
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.041569160819122e-06,
+      "loss": 0.8736,
+      "step": 16554
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.040464286229362e-06,
+      "loss": 1.0337,
+      "step": 16555
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.03935943651522e-06,
+      "loss": 0.9843,
+      "step": 16556
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.038254611690726e-06,
+      "loss": 1.0257,
+      "step": 16557
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.037149811769907e-06,
+      "loss": 0.9473,
+      "step": 16558
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.036045036766783e-06,
+      "loss": 0.9809,
+      "step": 16559
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.034940286695378e-06,
+      "loss": 0.9787,
+      "step": 16560
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.033835561569723e-06,
+      "loss": 0.9697,
+      "step": 16561
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.032730861403838e-06,
+      "loss": 1.0065,
+      "step": 16562
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.031626186211749e-06,
+      "loss": 1.1346,
+      "step": 16563
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.030521536007473e-06,
+      "loss": 1.0956,
+      "step": 16564
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.029416910805037e-06,
+      "loss": 0.9709,
+      "step": 16565
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02831231061847e-06,
+      "loss": 1.0056,
+      "step": 16566
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.027207735461784e-06,
+      "loss": 0.8497,
+      "step": 16567
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02610318534901e-06,
+      "loss": 1.0506,
+      "step": 16568
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.024998660294166e-06,
+      "loss": 1.0385,
+      "step": 16569
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.02389416031127e-06,
+      "loss": 0.9576,
+      "step": 16570
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.022789685414348e-06,
+      "loss": 1.0124,
+      "step": 16571
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.021685235617422e-06,
+      "loss": 1.0338,
+      "step": 16572
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.020580810934512e-06,
+      "loss": 1.0384,
+      "step": 16573
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.019476411379631e-06,
+      "loss": 0.9376,
+      "step": 16574
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.018372036966806e-06,
+      "loss": 0.9501,
+      "step": 16575
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.01726768771006e-06,
+      "loss": 0.96,
+      "step": 16576
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.016163363623406e-06,
+      "loss": 0.9655,
+      "step": 16577
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.015059064720863e-06,
+      "loss": 1.0469,
+      "step": 16578
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.013954791016448e-06,
+      "loss": 0.9895,
+      "step": 16579
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.012850542524188e-06,
+      "loss": 1.0601,
+      "step": 16580
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.011746319258094e-06,
+      "loss": 0.8778,
+      "step": 16581
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.010642121232188e-06,
+      "loss": 0.928,
+      "step": 16582
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00953794846048e-06,
+      "loss": 1.042,
+      "step": 16583
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.008433800956996e-06,
+      "loss": 0.9278,
+      "step": 16584
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.007329678735747e-06,
+      "loss": 0.9962,
+      "step": 16585
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.006225581810753e-06,
+      "loss": 0.8678,
+      "step": 16586
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.005121510196028e-06,
+      "loss": 0.9843,
+      "step": 16587
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.004017463905584e-06,
+      "loss": 0.8614,
+      "step": 16588
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.002913442953443e-06,
+      "loss": 0.9728,
+      "step": 16589
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.00180944735362e-06,
+      "loss": 0.9815,
+      "step": 16590
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 8.000705477120127e-06,
+      "loss": 0.9256,
+      "step": 16591
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.999601532266981e-06,
+      "loss": 1.0146,
+      "step": 16592
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.998497612808188e-06,
+      "loss": 1.0397,
+      "step": 16593
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.997393718757773e-06,
+      "loss": 1.0791,
+      "step": 16594
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.996289850129744e-06,
+      "loss": 0.9931,
+      "step": 16595
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.995186006938116e-06,
+      "loss": 1.0218,
+      "step": 16596
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.994082189196897e-06,
+      "loss": 0.9042,
+      "step": 16597
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.992978396920108e-06,
+      "loss": 0.8937,
+      "step": 16598
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.991874630121753e-06,
+      "loss": 1.1052,
+      "step": 16599
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.99077088881585e-06,
+      "loss": 0.938,
+      "step": 16600
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98966717301641e-06,
+      "loss": 0.9876,
+      "step": 16601
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.988563482737438e-06,
+      "loss": 1.0459,
+      "step": 16602
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.98745981799295e-06,
+      "loss": 1.0532,
+      "step": 16603
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.986356178796961e-06,
+      "loss": 0.9419,
+      "step": 16604
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.985252565163472e-06,
+      "loss": 0.924,
+      "step": 16605
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.984148977106501e-06,
+      "loss": 0.948,
+      "step": 16606
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.983045414640049e-06,
+      "loss": 0.8811,
+      "step": 16607
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.981941877778135e-06,
+      "loss": 1.09,
+      "step": 16608
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.980838366534762e-06,
+      "loss": 0.9185,
+      "step": 16609
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.979734880923942e-06,
+      "loss": 1.0513,
+      "step": 16610
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.978631420959676e-06,
+      "loss": 0.9885,
+      "step": 16611
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.977527986655981e-06,
+      "loss": 0.8995,
+      "step": 16612
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.976424578026862e-06,
+      "loss": 0.9141,
+      "step": 16613
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.975321195086327e-06,
+      "loss": 1.0374,
+      "step": 16614
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97421783784838e-06,
+      "loss": 0.9,
+      "step": 16615
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.973114506327026e-06,
+      "loss": 0.8859,
+      "step": 16616
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.97201120053628e-06,
+      "loss": 0.9248,
+      "step": 16617
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.970907920490143e-06,
+      "loss": 0.9982,
+      "step": 16618
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.969804666202618e-06,
+      "loss": 1.0331,
+      "step": 16619
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.968701437687714e-06,
+      "loss": 1.0207,
+      "step": 16620
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.967598234959432e-06,
+      "loss": 0.9018,
+      "step": 16621
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.966495058031786e-06,
+      "loss": 0.9014,
+      "step": 16622
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.965391906918772e-06,
+      "loss": 0.9445,
+      "step": 16623
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.964288781634397e-06,
+      "loss": 1.0355,
+      "step": 16624
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.96318568219266e-06,
+      "loss": 0.9121,
+      "step": 16625
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.962082608607576e-06,
+      "loss": 0.9618,
+      "step": 16626
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.960979560893137e-06,
+      "loss": 0.9898,
+      "step": 16627
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.95987653906335e-06,
+      "loss": 1.0001,
+      "step": 16628
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.958773543132217e-06,
+      "loss": 0.9741,
+      "step": 16629
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.957670573113739e-06,
+      "loss": 1.0163,
+      "step": 16630
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.956567629021922e-06,
+      "loss": 0.924,
+      "step": 16631
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.955464710870765e-06,
+      "loss": 0.9488,
+      "step": 16632
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.954361818674267e-06,
+      "loss": 0.9573,
+      "step": 16633
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.953258952446432e-06,
+      "loss": 1.1159,
+      "step": 16634
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.952156112201257e-06,
+      "loss": 1.0129,
+      "step": 16635
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.951053297952747e-06,
+      "loss": 1.0338,
+      "step": 16636
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.949950509714897e-06,
+      "loss": 1.0313,
+      "step": 16637
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.948847747501712e-06,
+      "loss": 0.9656,
+      "step": 16638
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.947745011327182e-06,
+      "loss": 1.0255,
+      "step": 16639
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.946642301205313e-06,
+      "loss": 0.9435,
+      "step": 16640
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.945539617150108e-06,
+      "loss": 0.9042,
+      "step": 16641
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.944436959175557e-06,
+      "loss": 0.9915,
+      "step": 16642
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.943334327295659e-06,
+      "loss": 1.0649,
+      "step": 16643
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.942231721524411e-06,
+      "loss": 1.0073,
+      "step": 16644
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.941129141875815e-06,
+      "loss": 1.0308,
+      "step": 16645
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.940026588363866e-06,
+      "loss": 0.9246,
+      "step": 16646
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.938924061002556e-06,
+      "loss": 0.9677,
+      "step": 16647
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.937821559805889e-06,
+      "loss": 1.0728,
+      "step": 16648
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93671908478785e-06,
+      "loss": 0.9462,
+      "step": 16649
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.935616635962446e-06,
+      "loss": 0.9835,
+      "step": 16650
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.934514213343666e-06,
+      "loss": 1.0769,
+      "step": 16651
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.933411816945508e-06,
+      "loss": 1.0473,
+      "step": 16652
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93230944678196e-06,
+      "loss": 1.0677,
+      "step": 16653
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.931207102867022e-06,
+      "loss": 0.8801,
+      "step": 16654
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.93010478521469e-06,
+      "loss": 0.9531,
+      "step": 16655
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.929002493838953e-06,
+      "loss": 0.9045,
+      "step": 16656
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.927900228753805e-06,
+      "loss": 0.9799,
+      "step": 16657
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.926797989973237e-06,
+      "loss": 0.889,
+      "step": 16658
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.925695777511247e-06,
+      "loss": 0.9848,
+      "step": 16659
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.924593591381825e-06,
+      "loss": 1.0444,
+      "step": 16660
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92349143159896e-06,
+      "loss": 0.9234,
+      "step": 16661
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.922389298176647e-06,
+      "loss": 1.0377,
+      "step": 16662
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.92128719112887e-06,
+      "loss": 0.9441,
+      "step": 16663
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.920185110469629e-06,
+      "loss": 1.0184,
+      "step": 16664
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.919083056212913e-06,
+      "loss": 0.9855,
+      "step": 16665
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.917981028372713e-06,
+      "loss": 0.9072,
+      "step": 16666
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.916879026963008e-06,
+      "loss": 1.0806,
+      "step": 16667
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.9157770519978e-06,
+      "loss": 1.0139,
+      "step": 16668
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.914675103491074e-06,
+      "loss": 0.8987,
+      "step": 16669
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91357318145682e-06,
+      "loss": 0.9928,
+      "step": 16670
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.912471285909022e-06,
+      "loss": 1.0059,
+      "step": 16671
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.91136941686167e-06,
+      "loss": 0.87,
+      "step": 16672
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.910267574328758e-06,
+      "loss": 0.9734,
+      "step": 16673
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.909165758324267e-06,
+      "loss": 0.9895,
+      "step": 16674
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.908063968862186e-06,
+      "loss": 0.9763,
+      "step": 16675
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.906962205956503e-06,
+      "loss": 0.9533,
+      "step": 16676
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.905860469621197e-06,
+      "loss": 1.0069,
+      "step": 16677
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.904758759870266e-06,
+      "loss": 1.0597,
+      "step": 16678
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.90365707671769e-06,
+      "loss": 0.8898,
+      "step": 16679
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.902555420177456e-06,
+      "loss": 0.9228,
+      "step": 16680
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.901453790263542e-06,
+      "loss": 0.9707,
+      "step": 16681
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.900352186989942e-06,
+      "loss": 0.9264,
+      "step": 16682
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.899250610370638e-06,
+      "loss": 1.0547,
+      "step": 16683
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.898149060419615e-06,
+      "loss": 0.9821,
+      "step": 16684
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.89704753715085e-06,
+      "loss": 1.1177,
+      "step": 16685
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.895946040578332e-06,
+      "loss": 0.9627,
+      "step": 16686
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.894844570716048e-06,
+      "loss": 1.0455,
+      "step": 16687
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.893743127577973e-06,
+      "loss": 1.0351,
+      "step": 16688
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.892641711178094e-06,
+      "loss": 0.9749,
+      "step": 16689
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.891540321530393e-06,
+      "loss": 0.7911,
+      "step": 16690
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.890438958648846e-06,
+      "loss": 1.0198,
+      "step": 16691
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.889337622547441e-06,
+      "loss": 0.9562,
+      "step": 16692
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88823631324016e-06,
+      "loss": 0.8907,
+      "step": 16693
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.88713503074098e-06,
+      "loss": 0.9749,
+      "step": 16694
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.886033775063879e-06,
+      "loss": 0.8845,
+      "step": 16695
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.884932546222842e-06,
+      "loss": 0.9474,
+      "step": 16696
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.883831344231849e-06,
+      "loss": 0.9095,
+      "step": 16697
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.882730169104877e-06,
+      "loss": 0.891,
+      "step": 16698
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.881629020855905e-06,
+      "loss": 0.9721,
+      "step": 16699
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.880527899498909e-06,
+      "loss": 0.8248,
+      "step": 16700
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.879426805047874e-06,
+      "loss": 0.8905,
+      "step": 16701
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.878325737516773e-06,
+      "loss": 0.8864,
+      "step": 16702
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.877224696919587e-06,
+      "loss": 0.8751,
+      "step": 16703
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.876123683270293e-06,
+      "loss": 1.0064,
+      "step": 16704
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.875022696582862e-06,
+      "loss": 1.0276,
+      "step": 16705
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.873921736871274e-06,
+      "loss": 1.0481,
+      "step": 16706
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.872820804149512e-06,
+      "loss": 0.9686,
+      "step": 16707
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.871719898431545e-06,
+      "loss": 0.9943,
+      "step": 16708
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.870619019731345e-06,
+      "loss": 1.0081,
+      "step": 16709
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.869518168062896e-06,
+      "loss": 0.9853,
+      "step": 16710
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.868417343440172e-06,
+      "loss": 0.9476,
+      "step": 16711
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86731654587714e-06,
+      "loss": 0.9533,
+      "step": 16712
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.866215775387782e-06,
+      "loss": 0.8639,
+      "step": 16713
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.865115031986064e-06,
+      "loss": 0.9739,
+      "step": 16714
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.86401431568597e-06,
+      "loss": 0.9467,
+      "step": 16715
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.862913626501466e-06,
+      "loss": 1.0329,
+      "step": 16716
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.861812964446528e-06,
+      "loss": 1.0097,
+      "step": 16717
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.860712329535123e-06,
+      "loss": 0.9333,
+      "step": 16718
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85961172178123e-06,
+      "loss": 1.0142,
+      "step": 16719
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.858511141198816e-06,
+      "loss": 1.008,
+      "step": 16720
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.857410587801859e-06,
+      "loss": 1.0539,
+      "step": 16721
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.856310061604322e-06,
+      "loss": 0.9054,
+      "step": 16722
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.85520956262018e-06,
+      "loss": 0.9852,
+      "step": 16723
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.854109090863402e-06,
+      "loss": 0.9136,
+      "step": 16724
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.853008646347963e-06,
+      "loss": 0.9745,
+      "step": 16725
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.851908229087825e-06,
+      "loss": 1.0326,
+      "step": 16726
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.850807839096964e-06,
+      "loss": 0.9695,
+      "step": 16727
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84970747638934e-06,
+      "loss": 0.8588,
+      "step": 16728
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.848607140978936e-06,
+      "loss": 0.8643,
+      "step": 16729
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.847506832879708e-06,
+      "loss": 0.9596,
+      "step": 16730
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.84640655210563e-06,
+      "loss": 0.9609,
+      "step": 16731
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.845306298670665e-06,
+      "loss": 1.047,
+      "step": 16732
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.844206072588787e-06,
+      "loss": 1.0067,
+      "step": 16733
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.843105873873958e-06,
+      "loss": 0.9323,
+      "step": 16734
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.842005702540147e-06,
+      "loss": 1.0155,
+      "step": 16735
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.840905558601317e-06,
+      "loss": 0.9213,
+      "step": 16736
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.839805442071436e-06,
+      "loss": 0.9885,
+      "step": 16737
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.838705352964472e-06,
+      "loss": 0.9871,
+      "step": 16738
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83760529129439e-06,
+      "loss": 0.9119,
+      "step": 16739
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.83650525707515e-06,
+      "loss": 1.0194,
+      "step": 16740
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.835405250320721e-06,
+      "loss": 1.1011,
+      "step": 16741
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.834305271045062e-06,
+      "loss": 0.8695,
+      "step": 16742
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.833205319262144e-06,
+      "loss": 1.0151,
+      "step": 16743
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.832105394985927e-06,
+      "loss": 0.8694,
+      "step": 16744
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.831005498230375e-06,
+      "loss": 1.0441,
+      "step": 16745
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.829905629009446e-06,
+      "loss": 0.9836,
+      "step": 16746
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.82880578733711e-06,
+      "loss": 1.1464,
+      "step": 16747
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.827705973227326e-06,
+      "loss": 0.961,
+      "step": 16748
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.826606186694056e-06,
+      "loss": 0.9282,
+      "step": 16749
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.825506427751259e-06,
+      "loss": 1.077,
+      "step": 16750
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.824406696412894e-06,
+      "loss": 1.0921,
+      "step": 16751
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.823306992692932e-06,
+      "loss": 0.9831,
+      "step": 16752
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.822207316605325e-06,
+      "loss": 0.9228,
+      "step": 16753
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.821107668164034e-06,
+      "loss": 0.9482,
+      "step": 16754
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.820008047383021e-06,
+      "loss": 0.8468,
+      "step": 16755
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.818908454276239e-06,
+      "loss": 0.8712,
+      "step": 16756
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.817808888857659e-06,
+      "loss": 0.9282,
+      "step": 16757
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.816709351141227e-06,
+      "loss": 0.9088,
+      "step": 16758
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.81560984114091e-06,
+      "loss": 1.0119,
+      "step": 16759
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.814510358870658e-06,
+      "loss": 0.9747,
+      "step": 16760
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.813410904344436e-06,
+      "loss": 0.9968,
+      "step": 16761
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.8123114775762e-06,
+      "loss": 1.0209,
+      "step": 16762
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.811212078579906e-06,
+      "loss": 0.7989,
+      "step": 16763
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.810112707369506e-06,
+      "loss": 1.1038,
+      "step": 16764
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.809013363958958e-06,
+      "loss": 0.9693,
+      "step": 16765
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.807914048362224e-06,
+      "loss": 1.0962,
+      "step": 16766
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.806814760593255e-06,
+      "loss": 0.9136,
+      "step": 16767
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.805715500666003e-06,
+      "loss": 1.013,
+      "step": 16768
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.804616268594429e-06,
+      "loss": 1.0092,
+      "step": 16769
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80351706439248e-06,
+      "loss": 1.0771,
+      "step": 16770
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.802417888074116e-06,
+      "loss": 0.9761,
+      "step": 16771
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.80131873965329e-06,
+      "loss": 1.033,
+      "step": 16772
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.800219619143955e-06,
+      "loss": 1.001,
+      "step": 16773
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.799120526560057e-06,
+      "loss": 0.8857,
+      "step": 16774
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79802146191556e-06,
+      "loss": 1.0129,
+      "step": 16775
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.79692242522441e-06,
+      "loss": 1.0121,
+      "step": 16776
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.795823416500563e-06,
+      "loss": 1.1658,
+      "step": 16777
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.794724435757962e-06,
+      "loss": 1.1234,
+      "step": 16778
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.793625483010563e-06,
+      "loss": 1.0818,
+      "step": 16779
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.792526558272319e-06,
+      "loss": 1.0148,
+      "step": 16780
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.791427661557182e-06,
+      "loss": 0.9541,
+      "step": 16781
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.790328792879096e-06,
+      "loss": 0.9171,
+      "step": 16782
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.789229952252016e-06,
+      "loss": 0.9899,
+      "step": 16783
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.788131139689883e-06,
+      "loss": 0.9571,
+      "step": 16784
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.787032355206656e-06,
+      "loss": 0.9702,
+      "step": 16785
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78593359881628e-06,
+      "loss": 0.9289,
+      "step": 16786
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.784834870532704e-06,
+      "loss": 1.0039,
+      "step": 16787
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.78373617036987e-06,
+      "loss": 1.0156,
+      "step": 16788
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.782637498341735e-06,
+      "loss": 0.874,
+      "step": 16789
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.781538854462243e-06,
+      "loss": 0.9686,
+      "step": 16790
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.780440238745339e-06,
+      "loss": 0.9486,
+      "step": 16791
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.779341651204968e-06,
+      "loss": 0.9006,
+      "step": 16792
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.778243091855077e-06,
+      "loss": 0.9238,
+      "step": 16793
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.777144560709618e-06,
+      "loss": 0.9233,
+      "step": 16794
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77604605778253e-06,
+      "loss": 0.9712,
+      "step": 16795
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.774947583087761e-06,
+      "loss": 0.8987,
+      "step": 16796
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.773849136639256e-06,
+      "loss": 0.963,
+      "step": 16797
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.772750718450953e-06,
+      "loss": 0.9414,
+      "step": 16798
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.771652328536803e-06,
+      "loss": 1.1573,
+      "step": 16799
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.77055396691075e-06,
+      "loss": 0.9491,
+      "step": 16800
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.769455633586736e-06,
+      "loss": 1.2721,
+      "step": 16801
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.768357328578699e-06,
+      "loss": 1.0652,
+      "step": 16802
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.767259051900586e-06,
+      "loss": 1.1183,
+      "step": 16803
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.766160803566343e-06,
+      "loss": 0.9344,
+      "step": 16804
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.765062583589904e-06,
+      "loss": 0.9763,
+      "step": 16805
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.763964391985217e-06,
+      "loss": 1.0121,
+      "step": 16806
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.762866228766216e-06,
+      "loss": 0.8751,
+      "step": 16807
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.761768093946852e-06,
+      "loss": 0.9496,
+      "step": 16808
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.760669987541055e-06,
+      "loss": 0.9448,
+      "step": 16809
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.759571909562774e-06,
+      "loss": 0.919,
+      "step": 16810
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.758473860025943e-06,
+      "loss": 0.8977,
+      "step": 16811
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.7573758389445e-06,
+      "loss": 0.9563,
+      "step": 16812
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.756277846332388e-06,
+      "loss": 0.9816,
+      "step": 16813
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.755179882203547e-06,
+      "loss": 0.8994,
+      "step": 16814
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.754081946571913e-06,
+      "loss": 0.9205,
+      "step": 16815
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 7.75298403945142e-06,
+      "loss": 1.0789,
+      "step": 16816
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.75188616085601e-06,
+      "loss": 0.8638,
+      "step": 16817
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.750788310799622e-06,
+      "loss": 1.0836,
+      "step": 16818
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.749690489296189e-06,
+      "loss": 1.0459,
+      "step": 16819
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.748592696359649e-06,
+      "loss": 1.0294,
+      "step": 16820
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.747494932003933e-06,
+      "loss": 0.9534,
+      "step": 16821
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.746397196242988e-06,
+      "loss": 0.8261,
+      "step": 16822
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.745299489090741e-06,
+      "loss": 1.0328,
+      "step": 16823
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.744201810561129e-06,
+      "loss": 1.0538,
+      "step": 16824
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.743104160668087e-06,
+      "loss": 0.9122,
+      "step": 16825
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.742006539425546e-06,
+      "loss": 0.9536,
+      "step": 16826
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.740908946847445e-06,
+      "loss": 0.9201,
+      "step": 16827
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.739811382947716e-06,
+      "loss": 1.0294,
+      "step": 16828
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.738713847740289e-06,
+      "loss": 0.9039,
+      "step": 16829
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7376163412391e-06,
+      "loss": 1.0293,
+      "step": 16830
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73651886345808e-06,
+      "loss": 0.9435,
+      "step": 16831
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.735421414411166e-06,
+      "loss": 0.9638,
+      "step": 16832
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.734323994112281e-06,
+      "loss": 0.9993,
+      "step": 16833
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.733226602575366e-06,
+      "loss": 1.003,
+      "step": 16834
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.73212923981434e-06,
+      "loss": 0.9188,
+      "step": 16835
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.731031905843145e-06,
+      "loss": 1.1199,
+      "step": 16836
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.729934600675707e-06,
+      "loss": 1.1026,
+      "step": 16837
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.728837324325955e-06,
+      "loss": 0.9869,
+      "step": 16838
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.727740076807821e-06,
+      "loss": 0.9816,
+      "step": 16839
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.726642858135228e-06,
+      "loss": 0.9388,
+      "step": 16840
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.725545668322111e-06,
+      "loss": 1.0599,
+      "step": 16841
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.7244485073824e-06,
+      "loss": 0.9198,
+      "step": 16842
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.723351375330017e-06,
+      "loss": 0.9813,
+      "step": 16843
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.72225427217889e-06,
+      "loss": 0.9237,
+      "step": 16844
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.721157197942954e-06,
+      "loss": 0.9352,
+      "step": 16845
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.720060152636131e-06,
+      "loss": 0.9031,
+      "step": 16846
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.718963136272343e-06,
+      "loss": 1.0339,
+      "step": 16847
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.717866148865525e-06,
+      "loss": 1.0696,
+      "step": 16848
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.716769190429594e-06,
+      "loss": 0.9586,
+      "step": 16849
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.715672260978484e-06,
+      "loss": 0.9412,
+      "step": 16850
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.714575360526114e-06,
+      "loss": 1.0826,
+      "step": 16851
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.713478489086414e-06,
+      "loss": 0.9139,
+      "step": 16852
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.712381646673301e-06,
+      "loss": 1.0143,
+      "step": 16853
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.711284833300705e-06,
+      "loss": 0.9639,
+      "step": 16854
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.710188048982548e-06,
+      "loss": 0.9691,
+      "step": 16855
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.709091293732755e-06,
+      "loss": 1.049,
+      "step": 16856
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.707994567565246e-06,
+      "loss": 0.8543,
+      "step": 16857
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.706897870493943e-06,
+      "loss": 0.9581,
+      "step": 16858
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.705801202532773e-06,
+      "loss": 0.9775,
+      "step": 16859
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.704704563695654e-06,
+      "loss": 1.0049,
+      "step": 16860
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.703607953996508e-06,
+      "loss": 1.0336,
+      "step": 16861
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.702511373449259e-06,
+      "loss": 1.0362,
+      "step": 16862
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.701414822067819e-06,
+      "loss": 1.0491,
+      "step": 16863
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.70031829986612e-06,
+      "loss": 0.972,
+      "step": 16864
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.699221806858073e-06,
+      "loss": 0.962,
+      "step": 16865
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.698125343057603e-06,
+      "loss": 0.9941,
+      "step": 16866
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.697028908478624e-06,
+      "loss": 1.0679,
+      "step": 16867
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69593250313506e-06,
+      "loss": 1.042,
+      "step": 16868
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.69483612704083e-06,
+      "loss": 1.0296,
+      "step": 16869
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.693739780209848e-06,
+      "loss": 0.935,
+      "step": 16870
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.692643462656034e-06,
+      "loss": 1.013,
+      "step": 16871
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.691547174393299e-06,
+      "loss": 0.8981,
+      "step": 16872
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.690450915435573e-06,
+      "loss": 1.0427,
+      "step": 16873
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.689354685796764e-06,
+      "loss": 0.9224,
+      "step": 16874
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.688258485490787e-06,
+      "loss": 0.971,
+      "step": 16875
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.687162314531564e-06,
+      "loss": 0.9892,
+      "step": 16876
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.686066172933002e-06,
+      "loss": 0.9455,
+      "step": 16877
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.684970060709023e-06,
+      "loss": 0.8751,
+      "step": 16878
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.683873977873543e-06,
+      "loss": 0.9426,
+      "step": 16879
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.682777924440472e-06,
+      "loss": 1.15,
+      "step": 16880
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.681681900423722e-06,
+      "loss": 0.942,
+      "step": 16881
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.680585905837211e-06,
+      "loss": 0.922,
+      "step": 16882
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.679489940694854e-06,
+      "loss": 0.9543,
+      "step": 16883
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.678394005010562e-06,
+      "loss": 0.9349,
+      "step": 16884
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.677298098798245e-06,
+      "loss": 1.0964,
+      "step": 16885
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.676202222071812e-06,
+      "loss": 0.9579,
+      "step": 16886
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.675106374845185e-06,
+      "loss": 1.0015,
+      "step": 16887
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.674010557132269e-06,
+      "loss": 0.9496,
+      "step": 16888
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.672914768946974e-06,
+      "loss": 0.9318,
+      "step": 16889
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.671819010303218e-06,
+      "loss": 0.8845,
+      "step": 16890
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.670723281214898e-06,
+      "loss": 0.9758,
+      "step": 16891
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.669627581695935e-06,
+      "loss": 0.9589,
+      "step": 16892
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.668531911760238e-06,
+      "loss": 1.0132,
+      "step": 16893
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.667436271421714e-06,
+      "loss": 1.0257,
+      "step": 16894
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.666340660694265e-06,
+      "loss": 1.0234,
+      "step": 16895
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.665245079591809e-06,
+      "loss": 1.0167,
+      "step": 16896
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.664149528128252e-06,
+      "loss": 0.8623,
+      "step": 16897
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6630540063175e-06,
+      "loss": 1.0114,
+      "step": 16898
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.66195851417346e-06,
+      "loss": 1.0154,
+      "step": 16899
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.660863051710037e-06,
+      "loss": 0.8745,
+      "step": 16900
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.659767618941143e-06,
+      "loss": 0.8509,
+      "step": 16901
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65867221588068e-06,
+      "loss": 0.997,
+      "step": 16902
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.657576842542555e-06,
+      "loss": 0.9539,
+      "step": 16903
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.656481498940675e-06,
+      "loss": 1.0237,
+      "step": 16904
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.65538618508894e-06,
+      "loss": 0.9406,
+      "step": 16905
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.654290901001258e-06,
+      "loss": 0.8965,
+      "step": 16906
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.653195646691535e-06,
+      "loss": 1.0199,
+      "step": 16907
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.652100422173677e-06,
+      "loss": 1.0816,
+      "step": 16908
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.651005227461574e-06,
+      "loss": 0.8927,
+      "step": 16909
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.649910062569143e-06,
+      "loss": 1.0084,
+      "step": 16910
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.648814927510285e-06,
+      "loss": 0.9747,
+      "step": 16911
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.647719822298899e-06,
+      "loss": 0.9621,
+      "step": 16912
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.646624746948886e-06,
+      "loss": 0.9469,
+      "step": 16913
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.645529701474146e-06,
+      "loss": 0.9998,
+      "step": 16914
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.644434685888589e-06,
+      "loss": 1.1178,
+      "step": 16915
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.643339700206107e-06,
+      "loss": 0.9703,
+      "step": 16916
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.642244744440606e-06,
+      "loss": 0.9836,
+      "step": 16917
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.641149818605984e-06,
+      "loss": 1.0119,
+      "step": 16918
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.640054922716136e-06,
+      "loss": 0.9627,
+      "step": 16919
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63896005678497e-06,
+      "loss": 0.9933,
+      "step": 16920
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.637865220826381e-06,
+      "loss": 1.1195,
+      "step": 16921
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.636770414854268e-06,
+      "loss": 1.015,
+      "step": 16922
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.635675638882524e-06,
+      "loss": 0.9869,
+      "step": 16923
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.634580892925053e-06,
+      "loss": 1.0748,
+      "step": 16924
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.633486176995753e-06,
+      "loss": 0.9484,
+      "step": 16925
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.632391491108519e-06,
+      "loss": 0.8891,
+      "step": 16926
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.631296835277246e-06,
+      "loss": 1.0426,
+      "step": 16927
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.63020220951583e-06,
+      "loss": 0.8777,
+      "step": 16928
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.629107613838173e-06,
+      "loss": 0.8831,
+      "step": 16929
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6280130482581625e-06,
+      "loss": 1.0244,
+      "step": 16930
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6269185127897e-06,
+      "loss": 1.0568,
+      "step": 16931
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.625824007446677e-06,
+      "loss": 0.8543,
+      "step": 16932
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.624729532242985e-06,
+      "loss": 1.012,
+      "step": 16933
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.623635087192522e-06,
+      "loss": 0.9722,
+      "step": 16934
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.622540672309185e-06,
+      "loss": 0.9124,
+      "step": 16935
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6214462876068604e-06,
+      "loss": 0.989,
+      "step": 16936
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.620351933099443e-06,
+      "loss": 1.0219,
+      "step": 16937
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.619257608800828e-06,
+      "loss": 1.0666,
+      "step": 16938
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.618163314724904e-06,
+      "loss": 0.9608,
+      "step": 16939
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.617069050885566e-06,
+      "loss": 0.8679,
+      "step": 16940
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.615974817296703e-06,
+      "loss": 0.8957,
+      "step": 16941
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.614880613972203e-06,
+      "loss": 0.943,
+      "step": 16942
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.613786440925964e-06,
+      "loss": 1.072,
+      "step": 16943
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.612692298171871e-06,
+      "loss": 0.9758,
+      "step": 16944
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6115981857238165e-06,
+      "loss": 1.028,
+      "step": 16945
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.610504103595687e-06,
+      "loss": 0.9903,
+      "step": 16946
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.609410051801369e-06,
+      "loss": 1.0153,
+      "step": 16947
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.608316030354757e-06,
+      "loss": 0.9492,
+      "step": 16948
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.607222039269739e-06,
+      "loss": 1.0001,
+      "step": 16949
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.6061280785602e-06,
+      "loss": 0.9564,
+      "step": 16950
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.605034148240024e-06,
+      "loss": 1.0354,
+      "step": 16951
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.603940248323109e-06,
+      "loss": 0.976,
+      "step": 16952
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.602846378823333e-06,
+      "loss": 1.109,
+      "step": 16953
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.601752539754583e-06,
+      "loss": 1.0421,
+      "step": 16954
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.600658731130749e-06,
+      "loss": 0.9994,
+      "step": 16955
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.599564952965709e-06,
+      "loss": 0.88,
+      "step": 16956
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.598471205273357e-06,
+      "loss": 0.999,
+      "step": 16957
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.597377488067571e-06,
+      "loss": 0.8863,
+      "step": 16958
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.596283801362241e-06,
+      "loss": 0.8516,
+      "step": 16959
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.595190145171245e-06,
+      "loss": 0.9793,
+      "step": 16960
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.59409651950847e-06,
+      "loss": 1.0629,
+      "step": 16961
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.593002924387799e-06,
+      "loss": 0.9056,
+      "step": 16962
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.591909359823116e-06,
+      "loss": 0.8898,
+      "step": 16963
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5908158258283e-06,
+      "loss": 1.0792,
+      "step": 16964
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.589722322417233e-06,
+      "loss": 1.0183,
+      "step": 16965
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5886288496038045e-06,
+      "loss": 1.1607,
+      "step": 16966
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.587535407401887e-06,
+      "loss": 0.9495,
+      "step": 16967
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.586441995825364e-06,
+      "loss": 1.0236,
+      "step": 16968
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.585348614888119e-06,
+      "loss": 0.9609,
+      "step": 16969
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.584255264604023e-06,
+      "loss": 0.9279,
+      "step": 16970
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5831619449869676e-06,
+      "loss": 1.0279,
+      "step": 16971
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.582068656050824e-06,
+      "loss": 0.9393,
+      "step": 16972
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.580975397809476e-06,
+      "loss": 0.8612,
+      "step": 16973
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.579882170276796e-06,
+      "loss": 0.8947,
+      "step": 16974
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.578788973466667e-06,
+      "loss": 0.9959,
+      "step": 16975
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.577695807392967e-06,
+      "loss": 1.1696,
+      "step": 16976
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.576602672069574e-06,
+      "loss": 0.9865,
+      "step": 16977
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.57550956751036e-06,
+      "loss": 0.9981,
+      "step": 16978
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.574416493729202e-06,
+      "loss": 1.1662,
+      "step": 16979
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.573323450739984e-06,
+      "loss": 0.9077,
+      "step": 16980
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.572230438556576e-06,
+      "loss": 1.014,
+      "step": 16981
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.571137457192852e-06,
+      "loss": 1.1204,
+      "step": 16982
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.570044506662691e-06,
+      "loss": 0.9706,
+      "step": 16983
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.568951586979962e-06,
+      "loss": 1.1024,
+      "step": 16984
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.567858698158547e-06,
+      "loss": 0.9994,
+      "step": 16985
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.566765840212314e-06,
+      "loss": 0.8813,
+      "step": 16986
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.565673013155139e-06,
+      "loss": 0.8851,
+      "step": 16987
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.564580217000892e-06,
+      "loss": 1.0288,
+      "step": 16988
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.563487451763447e-06,
+      "loss": 1.0341,
+      "step": 16989
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.56239471745668e-06,
+      "loss": 0.9404,
+      "step": 16990
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.561302014094462e-06,
+      "loss": 1.022,
+      "step": 16991
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.560209341690659e-06,
+      "loss": 0.9826,
+      "step": 16992
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.559116700259144e-06,
+      "loss": 0.9987,
+      "step": 16993
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.558024089813793e-06,
+      "loss": 1.0372,
+      "step": 16994
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.556931510368473e-06,
+      "loss": 0.9658,
+      "step": 16995
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.555838961937051e-06,
+      "loss": 0.9429,
+      "step": 16996
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.554746444533401e-06,
+      "loss": 0.9002,
+      "step": 16997
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.553653958171386e-06,
+      "loss": 1.0122,
+      "step": 16998
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.552561502864882e-06,
+      "loss": 0.9713,
+      "step": 16999
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.551469078627756e-06,
+      "loss": 1.0096,
+      "step": 17000
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.550376685473875e-06,
+      "loss": 0.9488,
+      "step": 17001
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5492843234171e-06,
+      "loss": 1.0763,
+      "step": 17002
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.548191992471307e-06,
+      "loss": 1.0009,
+      "step": 17003
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.547099692650361e-06,
+      "loss": 0.9556,
+      "step": 17004
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.546007423968127e-06,
+      "loss": 0.9349,
+      "step": 17005
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5449151864384685e-06,
+      "loss": 1.0125,
+      "step": 17006
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5438229800752525e-06,
+      "loss": 0.9083,
+      "step": 17007
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.542730804892349e-06,
+      "loss": 0.9323,
+      "step": 17008
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.541638660903617e-06,
+      "loss": 0.9376,
+      "step": 17009
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.540546548122926e-06,
+      "loss": 1.0171,
+      "step": 17010
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.539454466564136e-06,
+      "loss": 0.8636,
+      "step": 17011
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.538362416241108e-06,
+      "loss": 0.8336,
+      "step": 17012
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.537270397167708e-06,
+      "loss": 0.8558,
+      "step": 17013
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.536178409357805e-06,
+      "loss": 0.956,
+      "step": 17014
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5350864528252545e-06,
+      "loss": 0.9266,
+      "step": 17015
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.533994527583915e-06,
+      "loss": 0.9923,
+      "step": 17016
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.532902633647657e-06,
+      "loss": 1.0341,
+      "step": 17017
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.531810771030338e-06,
+      "loss": 0.9266,
+      "step": 17018
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.530718939745819e-06,
+      "loss": 0.8985,
+      "step": 17019
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.529627139807958e-06,
+      "loss": 1.0586,
+      "step": 17020
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.528535371230615e-06,
+      "loss": 1.0052,
+      "step": 17021
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5274436340276554e-06,
+      "loss": 1.0418,
+      "step": 17022
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.526351928212934e-06,
+      "loss": 0.8707,
+      "step": 17023
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.525260253800311e-06,
+      "loss": 0.8961,
+      "step": 17024
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.524168610803644e-06,
+      "loss": 1.1274,
+      "step": 17025
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.523076999236787e-06,
+      "loss": 0.9739,
+      "step": 17026
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.521985419113604e-06,
+      "loss": 0.8546,
+      "step": 17027
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.520893870447953e-06,
+      "loss": 0.9756,
+      "step": 17028
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.519802353253687e-06,
+      "loss": 1.0028,
+      "step": 17029
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5187108675446585e-06,
+      "loss": 1.0276,
+      "step": 17030
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51761941333473e-06,
+      "loss": 1.1045,
+      "step": 17031
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.516527990637758e-06,
+      "loss": 0.9093,
+      "step": 17032
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.515436599467594e-06,
+      "loss": 1.0647,
+      "step": 17033
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.514345239838095e-06,
+      "loss": 0.9268,
+      "step": 17034
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.51325391176311e-06,
+      "loss": 1.0477,
+      "step": 17035
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.512162615256502e-06,
+      "loss": 0.9235,
+      "step": 17036
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.511071350332118e-06,
+      "loss": 0.9822,
+      "step": 17037
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.509980117003814e-06,
+      "loss": 0.9388,
+      "step": 17038
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.508888915285444e-06,
+      "loss": 0.959,
+      "step": 17039
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.507797745190854e-06,
+      "loss": 0.9071,
+      "step": 17040
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.5067066067339025e-06,
+      "loss": 0.9302,
+      "step": 17041
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.505615499928441e-06,
+      "loss": 1.0064,
+      "step": 17042
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.504524424788318e-06,
+      "loss": 1.0398,
+      "step": 17043
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.503433381327381e-06,
+      "loss": 1.0698,
+      "step": 17044
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.502342369559487e-06,
+      "loss": 1.0286,
+      "step": 17045
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.501251389498488e-06,
+      "loss": 0.8744,
+      "step": 17046
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.500160441158224e-06,
+      "loss": 0.9981,
+      "step": 17047
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4990695245525525e-06,
+      "loss": 1.1121,
+      "step": 17048
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.497978639695316e-06,
+      "loss": 0.8868,
+      "step": 17049
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.496887786600369e-06,
+      "loss": 1.0879,
+      "step": 17050
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.495796965281554e-06,
+      "loss": 0.8756,
+      "step": 17051
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.494706175752725e-06,
+      "loss": 1.0284,
+      "step": 17052
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.493615418027725e-06,
+      "loss": 0.8951,
+      "step": 17053
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4925246921203965e-06,
+      "loss": 0.9504,
+      "step": 17054
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.491433998044593e-06,
+      "loss": 1.0214,
+      "step": 17055
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.49034333581416e-06,
+      "loss": 0.9872,
+      "step": 17056
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48925270544294e-06,
+      "loss": 0.9072,
+      "step": 17057
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.488162106944777e-06,
+      "loss": 0.9417,
+      "step": 17058
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.48707154033352e-06,
+      "loss": 0.9428,
+      "step": 17059
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4859810056230135e-06,
+      "loss": 1.0583,
+      "step": 17060
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.484890502827096e-06,
+      "loss": 1.0811,
+      "step": 17061
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.483800031959618e-06,
+      "loss": 0.9831,
+      "step": 17062
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.482709593034414e-06,
+      "loss": 0.9758,
+      "step": 17063
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.481619186065336e-06,
+      "loss": 0.9868,
+      "step": 17064
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.480528811066221e-06,
+      "loss": 1.0166,
+      "step": 17065
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.479438468050913e-06,
+      "loss": 0.9907,
+      "step": 17066
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.47834815703325e-06,
+      "loss": 0.9311,
+      "step": 17067
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.477257878027077e-06,
+      "loss": 0.9321,
+      "step": 17068
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.476167631046234e-06,
+      "loss": 0.9345,
+      "step": 17069
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.475077416104563e-06,
+      "loss": 1.0361,
+      "step": 17070
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.473987233215898e-06,
+      "loss": 0.9064,
+      "step": 17071
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.472897082394081e-06,
+      "loss": 0.9209,
+      "step": 17072
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.471806963652956e-06,
+      "loss": 0.9694,
+      "step": 17073
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4707168770063574e-06,
+      "loss": 1.0411,
+      "step": 17074
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4696268224681235e-06,
+      "loss": 0.9731,
+      "step": 17075
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.468536800052093e-06,
+      "loss": 0.9244,
+      "step": 17076
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4674468097721e-06,
+      "loss": 0.9482,
+      "step": 17077
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.466356851641987e-06,
+      "loss": 0.9637,
+      "step": 17078
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.465266925675588e-06,
+      "loss": 1.0989,
+      "step": 17079
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.464177031886741e-06,
+      "loss": 0.9589,
+      "step": 17080
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.463087170289278e-06,
+      "loss": 0.9278,
+      "step": 17081
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.461997340897033e-06,
+      "loss": 0.9859,
+      "step": 17082
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.460907543723852e-06,
+      "loss": 0.9011,
+      "step": 17083
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4598177787835606e-06,
+      "loss": 0.8976,
+      "step": 17084
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.458728046089991e-06,
+      "loss": 0.9671,
+      "step": 17085
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.457638345656981e-06,
+      "loss": 0.9515,
+      "step": 17086
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.456548677498366e-06,
+      "loss": 1.0187,
+      "step": 17087
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.455459041627977e-06,
+      "loss": 0.8762,
+      "step": 17088
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.454369438059645e-06,
+      "loss": 1.036,
+      "step": 17089
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.453279866807204e-06,
+      "loss": 0.9775,
+      "step": 17090
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.452190327884481e-06,
+      "loss": 0.9276,
+      "step": 17091
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.451100821305316e-06,
+      "loss": 1.1008,
+      "step": 17092
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.450011347083532e-06,
+      "loss": 1.0854,
+      "step": 17093
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.448921905232966e-06,
+      "loss": 0.9867,
+      "step": 17094
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.447832495767442e-06,
+      "loss": 1.0286,
+      "step": 17095
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.44674311870079e-06,
+      "loss": 0.9914,
+      "step": 17096
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.445653774046846e-06,
+      "loss": 0.8941,
+      "step": 17097
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.4445644618194345e-06,
+      "loss": 1.0515,
+      "step": 17098
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.443475182032383e-06,
+      "loss": 1.0519,
+      "step": 17099
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.442385934699516e-06,
+      "loss": 0.8603,
+      "step": 17100
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.441296719834672e-06,
+      "loss": 0.9666,
+      "step": 17101
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.44020753745167e-06,
+      "loss": 0.9131,
+      "step": 17102
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.439118387564338e-06,
+      "loss": 0.993,
+      "step": 17103
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 7.438029270186504e-06,
+      "loss": 0.9902,
+      "step": 17104
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.436940185331989e-06,
+      "loss": 0.9581,
+      "step": 17105
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.435851133014625e-06,
+      "loss": 0.998,
+      "step": 17106
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.434762113248236e-06,
+      "loss": 0.8773,
+      "step": 17107
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.433673126046646e-06,
+      "loss": 1.0249,
+      "step": 17108
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.432584171423676e-06,
+      "loss": 0.9775,
+      "step": 17109
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.43149524939315e-06,
+      "loss": 1.0024,
+      "step": 17110
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.430406359968901e-06,
+      "loss": 1.0574,
+      "step": 17111
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.429317503164741e-06,
+      "loss": 1.0355,
+      "step": 17112
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.428228678994498e-06,
+      "loss": 1.1663,
+      "step": 17113
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.427139887471989e-06,
+      "loss": 0.9405,
+      "step": 17114
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.426051128611045e-06,
+      "loss": 1.0291,
+      "step": 17115
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4249624024254814e-06,
+      "loss": 0.8411,
+      "step": 17116
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.423873708929119e-06,
+      "loss": 1.006,
+      "step": 17117
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.42278504813578e-06,
+      "loss": 0.9692,
+      "step": 17118
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.421696420059281e-06,
+      "loss": 0.9036,
+      "step": 17119
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.420607824713446e-06,
+      "loss": 0.9836,
+      "step": 17120
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.419519262112096e-06,
+      "loss": 0.9644,
+      "step": 17121
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4184307322690465e-06,
+      "loss": 1.0208,
+      "step": 17122
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.417342235198115e-06,
+      "loss": 1.043,
+      "step": 17123
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.416253770913118e-06,
+      "loss": 0.9584,
+      "step": 17124
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.415165339427881e-06,
+      "loss": 1.004,
+      "step": 17125
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.414076940756216e-06,
+      "loss": 0.9158,
+      "step": 17126
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.412988574911939e-06,
+      "loss": 0.9922,
+      "step": 17127
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.411900241908866e-06,
+      "loss": 0.9435,
+      "step": 17128
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.410811941760818e-06,
+      "loss": 0.9479,
+      "step": 17129
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.409723674481606e-06,
+      "loss": 1.0496,
+      "step": 17130
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.408635440085048e-06,
+      "loss": 0.9425,
+      "step": 17131
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.407547238584958e-06,
+      "loss": 0.9803,
+      "step": 17132
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.406459069995144e-06,
+      "loss": 1.03,
+      "step": 17133
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.40537093432943e-06,
+      "loss": 1.066,
+      "step": 17134
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.404282831601626e-06,
+      "loss": 0.9044,
+      "step": 17135
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.403194761825544e-06,
+      "loss": 0.9566,
+      "step": 17136
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4021067250149946e-06,
+      "loss": 0.9636,
+      "step": 17137
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.4010187211837905e-06,
+      "loss": 0.9938,
+      "step": 17138
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3999307503457496e-06,
+      "loss": 1.0015,
+      "step": 17139
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.398842812514677e-06,
+      "loss": 1.0048,
+      "step": 17140
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.397754907704387e-06,
+      "loss": 0.9863,
+      "step": 17141
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.396667035928686e-06,
+      "loss": 0.9442,
+      "step": 17142
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.39557919720139e-06,
+      "loss": 1.0385,
+      "step": 17143
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.394491391536305e-06,
+      "loss": 0.9991,
+      "step": 17144
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.393403618947242e-06,
+      "loss": 1.0263,
+      "step": 17145
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3923158794480085e-06,
+      "loss": 0.9135,
+      "step": 17146
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.391228173052409e-06,
+      "loss": 0.9783,
+      "step": 17147
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3901404997742595e-06,
+      "loss": 0.8247,
+      "step": 17148
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3890528596273654e-06,
+      "loss": 1.018,
+      "step": 17149
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.387965252625533e-06,
+      "loss": 1.0428,
+      "step": 17150
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3868776787825645e-06,
+      "loss": 0.9905,
+      "step": 17151
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3857901381122735e-06,
+      "loss": 0.9171,
+      "step": 17152
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.384702630628464e-06,
+      "loss": 0.8658,
+      "step": 17153
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.383615156344938e-06,
+      "loss": 1.0347,
+      "step": 17154
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.382527715275506e-06,
+      "loss": 0.8825,
+      "step": 17155
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.381440307433965e-06,
+      "loss": 0.9869,
+      "step": 17156
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.380352932834128e-06,
+      "loss": 0.8738,
+      "step": 17157
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3792655914897934e-06,
+      "loss": 0.9736,
+      "step": 17158
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3781782834147675e-06,
+      "loss": 1.0055,
+      "step": 17159
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3770910086228535e-06,
+      "loss": 0.9491,
+      "step": 17160
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.376003767127846e-06,
+      "loss": 0.9869,
+      "step": 17161
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.374916558943557e-06,
+      "loss": 0.8974,
+      "step": 17162
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.373829384083786e-06,
+      "loss": 1.0995,
+      "step": 17163
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.372742242562332e-06,
+      "loss": 0.9755,
+      "step": 17164
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3716551343929955e-06,
+      "loss": 1.0439,
+      "step": 17165
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.370568059589579e-06,
+      "loss": 0.9149,
+      "step": 17166
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3694810181658845e-06,
+      "loss": 0.8074,
+      "step": 17167
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.368394010135706e-06,
+      "loss": 0.9765,
+      "step": 17168
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.367307035512849e-06,
+      "loss": 0.9811,
+      "step": 17169
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.366220094311103e-06,
+      "loss": 0.9707,
+      "step": 17170
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.365133186544278e-06,
+      "loss": 0.996,
+      "step": 17171
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.364046312226163e-06,
+      "loss": 0.9759,
+      "step": 17172
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.362959471370562e-06,
+      "loss": 0.9715,
+      "step": 17173
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.361872663991268e-06,
+      "loss": 0.9918,
+      "step": 17174
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.360785890102074e-06,
+      "loss": 0.883,
+      "step": 17175
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.359699149716783e-06,
+      "loss": 0.8589,
+      "step": 17176
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.358612442849191e-06,
+      "loss": 0.9675,
+      "step": 17177
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3575257695130886e-06,
+      "loss": 0.9405,
+      "step": 17178
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3564391297222704e-06,
+      "loss": 0.9923,
+      "step": 17179
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.355352523490538e-06,
+      "loss": 0.9824,
+      "step": 17180
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.354265950831681e-06,
+      "loss": 0.9122,
+      "step": 17181
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3531794117594905e-06,
+      "loss": 1.026,
+      "step": 17182
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.352092906287765e-06,
+      "loss": 0.976,
+      "step": 17183
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.351006434430289e-06,
+      "loss": 1.0083,
+      "step": 17184
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.349919996200865e-06,
+      "loss": 1.0291,
+      "step": 17185
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.348833591613278e-06,
+      "loss": 0.9789,
+      "step": 17186
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.347747220681325e-06,
+      "loss": 0.982,
+      "step": 17187
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.346660883418792e-06,
+      "loss": 0.9649,
+      "step": 17188
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.345574579839469e-06,
+      "loss": 1.1241,
+      "step": 17189
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.344488309957152e-06,
+      "loss": 0.9805,
+      "step": 17190
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.343402073785628e-06,
+      "loss": 0.9604,
+      "step": 17191
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.342315871338686e-06,
+      "loss": 0.9468,
+      "step": 17192
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.341229702630112e-06,
+      "loss": 0.9565,
+      "step": 17193
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.340143567673701e-06,
+      "loss": 0.97,
+      "step": 17194
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3390574664832385e-06,
+      "loss": 0.7873,
+      "step": 17195
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.337971399072508e-06,
+      "loss": 1.0989,
+      "step": 17196
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.336885365455304e-06,
+      "loss": 1.0607,
+      "step": 17197
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.335799365645404e-06,
+      "loss": 1.0677,
+      "step": 17198
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.334713399656604e-06,
+      "loss": 0.9582,
+      "step": 17199
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.333627467502685e-06,
+      "loss": 1.1254,
+      "step": 17200
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.332541569197433e-06,
+      "loss": 1.0203,
+      "step": 17201
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.331455704754633e-06,
+      "loss": 0.9051,
+      "step": 17202
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.330369874188068e-06,
+      "loss": 1.0801,
+      "step": 17203
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.329284077511529e-06,
+      "loss": 1.0315,
+      "step": 17204
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.328198314738794e-06,
+      "loss": 0.9449,
+      "step": 17205
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.327112585883646e-06,
+      "loss": 0.838,
+      "step": 17206
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.326026890959867e-06,
+      "loss": 0.9454,
+      "step": 17207
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3249412299812465e-06,
+      "loss": 1.1084,
+      "step": 17208
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.323855602961561e-06,
+      "loss": 0.9293,
+      "step": 17209
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.322770009914593e-06,
+      "loss": 0.9639,
+      "step": 17210
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.321684450854125e-06,
+      "loss": 0.9554,
+      "step": 17211
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.320598925793932e-06,
+      "loss": 0.9557,
+      "step": 17212
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3195134347478016e-06,
+      "loss": 0.9519,
+      "step": 17213
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.318427977729513e-06,
+      "loss": 0.9249,
+      "step": 17214
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.317342554752844e-06,
+      "loss": 0.9107,
+      "step": 17215
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.316257165831573e-06,
+      "loss": 0.9871,
+      "step": 17216
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.315171810979476e-06,
+      "loss": 0.9562,
+      "step": 17217
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3140864902103385e-06,
+      "loss": 1.0051,
+      "step": 17218
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.313001203537934e-06,
+      "loss": 0.9507,
+      "step": 17219
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3119159509760385e-06,
+      "loss": 0.9854,
+      "step": 17220
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.310830732538428e-06,
+      "loss": 1.2146,
+      "step": 17221
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.309745548238885e-06,
+      "loss": 0.9488,
+      "step": 17222
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.308660398091183e-06,
+      "loss": 1.0801,
+      "step": 17223
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.307575282109094e-06,
+      "loss": 0.8307,
+      "step": 17224
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3064902003063975e-06,
+      "loss": 0.8806,
+      "step": 17225
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.305405152696862e-06,
+      "loss": 0.9473,
+      "step": 17226
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.3043201392942686e-06,
+      "loss": 0.9044,
+      "step": 17227
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.303235160112392e-06,
+      "loss": 0.9364,
+      "step": 17228
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.302150215165e-06,
+      "loss": 0.9223,
+      "step": 17229
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.301065304465869e-06,
+      "loss": 1.0739,
+      "step": 17230
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.299980428028765e-06,
+      "loss": 0.8852,
+      "step": 17231
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.298895585867473e-06,
+      "loss": 1.0978,
+      "step": 17232
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.297810777995755e-06,
+      "loss": 0.9356,
+      "step": 17233
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.296726004427385e-06,
+      "loss": 0.9305,
+      "step": 17234
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.29564126517613e-06,
+      "loss": 0.9721,
+      "step": 17235
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.294556560255768e-06,
+      "loss": 1.0464,
+      "step": 17236
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.293471889680063e-06,
+      "loss": 1.007,
+      "step": 17237
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.292387253462788e-06,
+      "loss": 1.0302,
+      "step": 17238
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.291302651617711e-06,
+      "loss": 0.9036,
+      "step": 17239
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.290218084158596e-06,
+      "loss": 0.9733,
+      "step": 17240
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.289133551099217e-06,
+      "loss": 0.9995,
+      "step": 17241
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.288049052453341e-06,
+      "loss": 0.8496,
+      "step": 17242
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.286964588234736e-06,
+      "loss": 1.1555,
+      "step": 17243
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.285880158457165e-06,
+      "loss": 1.0522,
+      "step": 17244
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.284795763134394e-06,
+      "loss": 1.0079,
+      "step": 17245
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.283711402280196e-06,
+      "loss": 0.894,
+      "step": 17246
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.282627075908333e-06,
+      "loss": 0.9974,
+      "step": 17247
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.281542784032566e-06,
+      "loss": 1.1242,
+      "step": 17248
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.280458526666662e-06,
+      "loss": 0.8127,
+      "step": 17249
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27937430382439e-06,
+      "loss": 1.0068,
+      "step": 17250
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.278290115519508e-06,
+      "loss": 0.9658,
+      "step": 17251
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.277205961765785e-06,
+      "loss": 0.9151,
+      "step": 17252
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.27612184257698e-06,
+      "loss": 1.0476,
+      "step": 17253
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.275037757966852e-06,
+      "loss": 0.962,
+      "step": 17254
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.273953707949168e-06,
+      "loss": 0.9866,
+      "step": 17255
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.272869692537694e-06,
+      "loss": 1.0346,
+      "step": 17256
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2717857117461834e-06,
+      "loss": 1.0366,
+      "step": 17257
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.270701765588398e-06,
+      "loss": 0.9383,
+      "step": 17258
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.269617854078097e-06,
+      "loss": 1.0258,
+      "step": 17259
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2685339772290486e-06,
+      "loss": 0.9158,
+      "step": 17260
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.267450135055004e-06,
+      "loss": 1.1115,
+      "step": 17261
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.266366327569727e-06,
+      "loss": 0.954,
+      "step": 17262
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.26528255478697e-06,
+      "loss": 0.9941,
+      "step": 17263
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.264198816720498e-06,
+      "loss": 1.0126,
+      "step": 17264
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.263115113384065e-06,
+      "loss": 0.9367,
+      "step": 17265
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2620314447914306e-06,
+      "loss": 0.9077,
+      "step": 17266
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.260947810956349e-06,
+      "loss": 0.9758,
+      "step": 17267
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.259864211892576e-06,
+      "loss": 1.0749,
+      "step": 17268
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.258780647613868e-06,
+      "loss": 0.9593,
+      "step": 17269
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.257697118133986e-06,
+      "loss": 0.9878,
+      "step": 17270
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.256613623466677e-06,
+      "loss": 0.9514,
+      "step": 17271
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.255530163625701e-06,
+      "loss": 0.974,
+      "step": 17272
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.254446738624807e-06,
+      "loss": 1.0509,
+      "step": 17273
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2533633484777566e-06,
+      "loss": 0.9245,
+      "step": 17274
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.252279993198296e-06,
+      "loss": 1.0968,
+      "step": 17275
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.251196672800183e-06,
+      "loss": 0.9212,
+      "step": 17276
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.250113387297162e-06,
+      "loss": 0.8815,
+      "step": 17277
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.249030136702995e-06,
+      "loss": 0.8929,
+      "step": 17278
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2479469210314266e-06,
+      "loss": 0.9488,
+      "step": 17279
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.246863740296213e-06,
+      "loss": 0.8387,
+      "step": 17280
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2457805945111e-06,
+      "loss": 0.8591,
+      "step": 17281
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.244697483689838e-06,
+      "loss": 0.8855,
+      "step": 17282
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.243614407846179e-06,
+      "loss": 0.9428,
+      "step": 17283
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.242531366993875e-06,
+      "loss": 0.8731,
+      "step": 17284
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.241448361146668e-06,
+      "loss": 0.9939,
+      "step": 17285
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.240365390318307e-06,
+      "loss": 0.9847,
+      "step": 17286
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.239282454522549e-06,
+      "loss": 0.921,
+      "step": 17287
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.238199553773133e-06,
+      "loss": 0.8823,
+      "step": 17288
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.237116688083808e-06,
+      "loss": 0.9773,
+      "step": 17289
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.236033857468323e-06,
+      "loss": 0.998,
+      "step": 17290
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.234951061940417e-06,
+      "loss": 1.0659,
+      "step": 17291
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.233868301513847e-06,
+      "loss": 1.1093,
+      "step": 17292
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.232785576202348e-06,
+      "loss": 1.0517,
+      "step": 17293
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.231702886019672e-06,
+      "loss": 0.9613,
+      "step": 17294
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.230620230979557e-06,
+      "loss": 0.8084,
+      "step": 17295
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.22953761109575e-06,
+      "loss": 1.0139,
+      "step": 17296
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.228455026381995e-06,
+      "loss": 1.0291,
+      "step": 17297
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.227372476852037e-06,
+      "loss": 0.9975,
+      "step": 17298
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2262899625196144e-06,
+      "loss": 1.0289,
+      "step": 17299
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.225207483398469e-06,
+      "loss": 0.9637,
+      "step": 17300
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.224125039502349e-06,
+      "loss": 1.0432,
+      "step": 17301
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2230426308449916e-06,
+      "loss": 0.8319,
+      "step": 17302
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.221960257440135e-06,
+      "loss": 1.0207,
+      "step": 17303
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.220877919301523e-06,
+      "loss": 0.9446,
+      "step": 17304
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.219795616442892e-06,
+      "loss": 0.9632,
+      "step": 17305
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.218713348877987e-06,
+      "loss": 1.0502,
+      "step": 17306
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.217631116620541e-06,
+      "loss": 1.0075,
+      "step": 17307
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.216548919684299e-06,
+      "loss": 1.0887,
+      "step": 17308
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.215466758082994e-06,
+      "loss": 0.9561,
+      "step": 17309
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.214384631830361e-06,
+      "loss": 0.8932,
+      "step": 17310
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.213302540940147e-06,
+      "loss": 1.0838,
+      "step": 17311
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.212220485426083e-06,
+      "loss": 1.0203,
+      "step": 17312
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.211138465301903e-06,
+      "loss": 0.8163,
+      "step": 17313
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2100564805813444e-06,
+      "loss": 1.0779,
+      "step": 17314
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.208974531278145e-06,
+      "loss": 1.0462,
+      "step": 17315
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.207892617406041e-06,
+      "loss": 0.9737,
+      "step": 17316
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.206810738978763e-06,
+      "loss": 0.9967,
+      "step": 17317
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.205728896010046e-06,
+      "loss": 1.0571,
+      "step": 17318
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.2046470885136215e-06,
+      "loss": 0.9576,
+      "step": 17319
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.203565316503225e-06,
+      "loss": 1.0829,
+      "step": 17320
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.202483579992594e-06,
+      "loss": 0.9086,
+      "step": 17321
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.201401878995454e-06,
+      "loss": 0.9906,
+      "step": 17322
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.200320213525538e-06,
+      "loss": 1.0007,
+      "step": 17323
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.199238583596576e-06,
+      "loss": 1.0704,
+      "step": 17324
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.198156989222306e-06,
+      "loss": 1.0229,
+      "step": 17325
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.197075430416454e-06,
+      "loss": 0.9163,
+      "step": 17326
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.195993907192745e-06,
+      "loss": 1.0147,
+      "step": 17327
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.194912419564913e-06,
+      "loss": 1.0233,
+      "step": 17328
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.193830967546691e-06,
+      "loss": 0.9741,
+      "step": 17329
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.192749551151806e-06,
+      "loss": 0.9858,
+      "step": 17330
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.191668170393979e-06,
+      "loss": 1.0158,
+      "step": 17331
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.190586825286947e-06,
+      "loss": 0.999,
+      "step": 17332
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1895055158444285e-06,
+      "loss": 0.9332,
+      "step": 17333
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.188424242080157e-06,
+      "loss": 0.9753,
+      "step": 17334
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.187343004007861e-06,
+      "loss": 0.9355,
+      "step": 17335
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.18626180164126e-06,
+      "loss": 0.9056,
+      "step": 17336
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.185180634994081e-06,
+      "loss": 0.9867,
+      "step": 17337
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1840995040800474e-06,
+      "loss": 0.9616,
+      "step": 17338
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.183018408912891e-06,
+      "loss": 0.8879,
+      "step": 17339
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1819373495063305e-06,
+      "loss": 0.9576,
+      "step": 17340
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.180856325874088e-06,
+      "loss": 1.0159,
+      "step": 17341
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.179775338029889e-06,
+      "loss": 0.9889,
+      "step": 17342
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.178694385987459e-06,
+      "loss": 0.798,
+      "step": 17343
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.177613469760516e-06,
+      "loss": 1.0213,
+      "step": 17344
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.176532589362784e-06,
+      "loss": 0.9719,
+      "step": 17345
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.175451744807986e-06,
+      "loss": 0.8468,
+      "step": 17346
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1743709361098355e-06,
+      "loss": 1.0571,
+      "step": 17347
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.173290163282061e-06,
+      "loss": 0.951,
+      "step": 17348
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.172209426338381e-06,
+      "loss": 0.92,
+      "step": 17349
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.171128725292516e-06,
+      "loss": 1.0498,
+      "step": 17350
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.170048060158179e-06,
+      "loss": 0.9898,
+      "step": 17351
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1689674309490895e-06,
+      "loss": 1.1932,
+      "step": 17352
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.167886837678976e-06,
+      "loss": 0.8914,
+      "step": 17353
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.166806280361547e-06,
+      "loss": 0.9587,
+      "step": 17354
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.165725759010521e-06,
+      "loss": 0.9085,
+      "step": 17355
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.164645273639613e-06,
+      "loss": 1.0766,
+      "step": 17356
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.163564824262548e-06,
+      "loss": 0.9448,
+      "step": 17357
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1624844108930315e-06,
+      "loss": 0.9993,
+      "step": 17358
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.161404033544787e-06,
+      "loss": 0.8493,
+      "step": 17359
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.160323692231527e-06,
+      "loss": 0.9185,
+      "step": 17360
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.15924338696696e-06,
+      "loss": 0.9715,
+      "step": 17361
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.158163117764807e-06,
+      "loss": 0.8656,
+      "step": 17362
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.157082884638782e-06,
+      "loss": 0.9622,
+      "step": 17363
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.156002687602597e-06,
+      "loss": 0.936,
+      "step": 17364
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.15492252666996e-06,
+      "loss": 0.876,
+      "step": 17365
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.153842401854587e-06,
+      "loss": 0.9116,
+      "step": 17366
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.152762313170194e-06,
+      "loss": 0.9797,
+      "step": 17367
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.151682260630485e-06,
+      "loss": 0.9387,
+      "step": 17368
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.150602244249177e-06,
+      "loss": 0.967,
+      "step": 17369
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.149522264039972e-06,
+      "loss": 1.0259,
+      "step": 17370
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.148442320016591e-06,
+      "loss": 0.9486,
+      "step": 17371
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1473624121927356e-06,
+      "loss": 0.9793,
+      "step": 17372
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.146282540582119e-06,
+      "loss": 0.9727,
+      "step": 17373
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1452027051984486e-06,
+      "loss": 1.0912,
+      "step": 17374
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.144122906055429e-06,
+      "loss": 0.8996,
+      "step": 17375
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.14304314316677e-06,
+      "loss": 1.1345,
+      "step": 17376
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.141963416546184e-06,
+      "loss": 1.0842,
+      "step": 17377
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.1408837262073725e-06,
+      "loss": 1.0893,
+      "step": 17378
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.139804072164041e-06,
+      "loss": 1.014,
+      "step": 17379
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.138724454429895e-06,
+      "loss": 1.0352,
+      "step": 17380
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.137644873018645e-06,
+      "loss": 0.9607,
+      "step": 17381
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.136565327943992e-06,
+      "loss": 1.0052,
+      "step": 17382
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.135485819219643e-06,
+      "loss": 1.084,
+      "step": 17383
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.134406346859297e-06,
+      "loss": 1.0027,
+      "step": 17384
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.133326910876663e-06,
+      "loss": 0.961,
+      "step": 17385
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.132247511285442e-06,
+      "loss": 0.8716,
+      "step": 17386
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.131168148099338e-06,
+      "loss": 1.0115,
+      "step": 17387
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.130088821332051e-06,
+      "loss": 0.9692,
+      "step": 17388
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.129009530997279e-06,
+      "loss": 0.8977,
+      "step": 17389
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.12793027710873e-06,
+      "loss": 1.1678,
+      "step": 17390
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 7.126851059680105e-06,
+      "loss": 0.9285,
+      "step": 17391
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.125771878725098e-06,
+      "loss": 1.0072,
+      "step": 17392
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.124692734257415e-06,
+      "loss": 0.9199,
+      "step": 17393
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.123613626290748e-06,
+      "loss": 0.985,
+      "step": 17394
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.122534554838805e-06,
+      "loss": 0.9217,
+      "step": 17395
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.121455519915279e-06,
+      "loss": 0.955,
+      "step": 17396
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.120376521533869e-06,
+      "loss": 0.9635,
+      "step": 17397
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.11929755970827e-06,
+      "loss": 1.0073,
+      "step": 17398
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1182186344521855e-06,
+      "loss": 0.9652,
+      "step": 17399
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.117139745779306e-06,
+      "loss": 1.0077,
+      "step": 17400
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.116060893703333e-06,
+      "loss": 1.0255,
+      "step": 17401
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.114982078237955e-06,
+      "loss": 0.9509,
+      "step": 17402
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.113903299396871e-06,
+      "loss": 0.9715,
+      "step": 17403
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.112824557193778e-06,
+      "loss": 1.1534,
+      "step": 17404
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.111745851642369e-06,
+      "loss": 0.9109,
+      "step": 17405
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1106671827563345e-06,
+      "loss": 1.0234,
+      "step": 17406
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.109588550549371e-06,
+      "loss": 0.9555,
+      "step": 17407
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1085099550351675e-06,
+      "loss": 0.9446,
+      "step": 17408
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.107431396227424e-06,
+      "loss": 1.0149,
+      "step": 17409
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.106352874139826e-06,
+      "loss": 0.9395,
+      "step": 17410
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10527438878607e-06,
+      "loss": 0.9736,
+      "step": 17411
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.104195940179837e-06,
+      "loss": 0.9903,
+      "step": 17412
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.1031175283348294e-06,
+      "loss": 0.9726,
+      "step": 17413
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.10203915326473e-06,
+      "loss": 1.0388,
+      "step": 17414
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.100960814983234e-06,
+      "loss": 1.04,
+      "step": 17415
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0998825135040235e-06,
+      "loss": 0.9207,
+      "step": 17416
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.098804248840789e-06,
+      "loss": 0.8864,
+      "step": 17417
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.097726021007226e-06,
+      "loss": 1.0647,
+      "step": 17418
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.096647830017016e-06,
+      "loss": 1.0948,
+      "step": 17419
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.095569675883845e-06,
+      "loss": 0.957,
+      "step": 17420
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.094491558621403e-06,
+      "loss": 1.0262,
+      "step": 17421
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.093413478243372e-06,
+      "loss": 0.9459,
+      "step": 17422
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.092335434763445e-06,
+      "loss": 0.9855,
+      "step": 17423
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0912574281953e-06,
+      "loss": 1.0588,
+      "step": 17424
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.090179458552628e-06,
+      "loss": 0.9132,
+      "step": 17425
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.089101525849107e-06,
+      "loss": 1.0328,
+      "step": 17426
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.088023630098427e-06,
+      "loss": 0.9334,
+      "step": 17427
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.086945771314269e-06,
+      "loss": 0.9491,
+      "step": 17428
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.085867949510318e-06,
+      "loss": 1.0858,
+      "step": 17429
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.084790164700253e-06,
+      "loss": 1.141,
+      "step": 17430
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.083712416897755e-06,
+      "loss": 0.9191,
+      "step": 17431
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.082634706116512e-06,
+      "loss": 0.9067,
+      "step": 17432
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0815570323702035e-06,
+      "loss": 1.0506,
+      "step": 17433
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.080479395672506e-06,
+      "loss": 0.9742,
+      "step": 17434
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.079401796037099e-06,
+      "loss": 0.941,
+      "step": 17435
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.078324233477669e-06,
+      "loss": 0.9702,
+      "step": 17436
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.077246708007893e-06,
+      "loss": 0.9215,
+      "step": 17437
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0761692196414466e-06,
+      "loss": 1.0106,
+      "step": 17438
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07509176839201e-06,
+      "loss": 0.9959,
+      "step": 17439
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.07401435427326e-06,
+      "loss": 1.1164,
+      "step": 17440
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.072936977298874e-06,
+      "loss": 1.0006,
+      "step": 17441
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.071859637482534e-06,
+      "loss": 0.8548,
+      "step": 17442
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.070782334837911e-06,
+      "loss": 0.9556,
+      "step": 17443
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.069705069378683e-06,
+      "loss": 0.9557,
+      "step": 17444
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.06862784111852e-06,
+      "loss": 1.0164,
+      "step": 17445
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.067550650071107e-06,
+      "loss": 1.0032,
+      "step": 17446
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.066473496250113e-06,
+      "loss": 1.0734,
+      "step": 17447
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.065396379669212e-06,
+      "loss": 0.9516,
+      "step": 17448
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0643193003420755e-06,
+      "loss": 1.0532,
+      "step": 17449
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.063242258282382e-06,
+      "loss": 0.9017,
+      "step": 17450
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.062165253503801e-06,
+      "loss": 0.914,
+      "step": 17451
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.061088286020007e-06,
+      "loss": 1.0522,
+      "step": 17452
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0600113558446705e-06,
+      "loss": 0.9339,
+      "step": 17453
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.058934462991457e-06,
+      "loss": 0.9853,
+      "step": 17454
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.057857607474046e-06,
+      "loss": 1.0312,
+      "step": 17455
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.056780789306105e-06,
+      "loss": 0.9653,
+      "step": 17456
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.055704008501303e-06,
+      "loss": 0.9442,
+      "step": 17457
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0546272650733085e-06,
+      "loss": 0.9501,
+      "step": 17458
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.053550559035788e-06,
+      "loss": 1.0123,
+      "step": 17459
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.052473890402418e-06,
+      "loss": 0.9679,
+      "step": 17460
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.051397259186862e-06,
+      "loss": 0.9816,
+      "step": 17461
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.050320665402785e-06,
+      "loss": 0.9123,
+      "step": 17462
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.049244109063853e-06,
+      "loss": 1.0341,
+      "step": 17463
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.048167590183741e-06,
+      "loss": 1.0849,
+      "step": 17464
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.047091108776108e-06,
+      "loss": 0.9145,
+      "step": 17465
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.046014664854622e-06,
+      "loss": 0.9077,
+      "step": 17466
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.044938258432949e-06,
+      "loss": 1.1062,
+      "step": 17467
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.043861889524746e-06,
+      "loss": 1.0463,
+      "step": 17468
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.042785558143686e-06,
+      "loss": 1.0133,
+      "step": 17469
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.041709264303431e-06,
+      "loss": 1.016,
+      "step": 17470
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.040633008017643e-06,
+      "loss": 0.9416,
+      "step": 17471
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.039556789299984e-06,
+      "loss": 0.8634,
+      "step": 17472
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.038480608164113e-06,
+      "loss": 0.9133,
+      "step": 17473
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.037404464623701e-06,
+      "loss": 0.9296,
+      "step": 17474
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0363283586924e-06,
+      "loss": 0.7725,
+      "step": 17475
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0352522903838785e-06,
+      "loss": 0.9503,
+      "step": 17476
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.034176259711787e-06,
+      "loss": 0.8953,
+      "step": 17477
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.033100266689796e-06,
+      "loss": 1.0588,
+      "step": 17478
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.032024311331559e-06,
+      "loss": 0.9066,
+      "step": 17479
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.030948393650737e-06,
+      "loss": 1.0252,
+      "step": 17480
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.029872513660987e-06,
+      "loss": 0.9497,
+      "step": 17481
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.028796671375964e-06,
+      "loss": 1.0379,
+      "step": 17482
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.027720866809331e-06,
+      "loss": 1.1108,
+      "step": 17483
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0266450999747435e-06,
+      "loss": 0.9779,
+      "step": 17484
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.025569370885858e-06,
+      "loss": 0.8827,
+      "step": 17485
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.024493679556328e-06,
+      "loss": 0.9484,
+      "step": 17486
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.023418025999809e-06,
+      "loss": 0.9751,
+      "step": 17487
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.022342410229961e-06,
+      "loss": 0.9521,
+      "step": 17488
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.021266832260431e-06,
+      "loss": 1.0135,
+      "step": 17489
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.020191292104883e-06,
+      "loss": 1.0014,
+      "step": 17490
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.01911578977696e-06,
+      "loss": 0.9126,
+      "step": 17491
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.018040325290323e-06,
+      "loss": 1.0585,
+      "step": 17492
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.01696489865862e-06,
+      "loss": 0.8778,
+      "step": 17493
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.015889509895509e-06,
+      "loss": 0.926,
+      "step": 17494
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.014814159014635e-06,
+      "loss": 0.9797,
+      "step": 17495
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0137388460296495e-06,
+      "loss": 0.9096,
+      "step": 17496
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.012663570954207e-06,
+      "loss": 0.8535,
+      "step": 17497
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.011588333801957e-06,
+      "loss": 0.9897,
+      "step": 17498
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.010513134586549e-06,
+      "loss": 0.982,
+      "step": 17499
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0094379733216326e-06,
+      "loss": 1.0963,
+      "step": 17500
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.008362850020851e-06,
+      "loss": 1.0754,
+      "step": 17501
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0072877646978636e-06,
+      "loss": 0.9805,
+      "step": 17502
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.006212717366307e-06,
+      "loss": 0.9442,
+      "step": 17503
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.005137708039838e-06,
+      "loss": 0.9446,
+      "step": 17504
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.004062736732094e-06,
+      "loss": 0.9134,
+      "step": 17505
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.00298780345673e-06,
+      "loss": 1.0369,
+      "step": 17506
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.0019129082273865e-06,
+      "loss": 0.9404,
+      "step": 17507
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 7.000838051057713e-06,
+      "loss": 0.9473,
+      "step": 17508
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99976323196135e-06,
+      "loss": 0.9513,
+      "step": 17509
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.998688450951944e-06,
+      "loss": 0.9078,
+      "step": 17510
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9976137080431384e-06,
+      "loss": 1.0001,
+      "step": 17511
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.996539003248581e-06,
+      "loss": 0.9926,
+      "step": 17512
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.995464336581908e-06,
+      "loss": 0.8875,
+      "step": 17513
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.994389708056767e-06,
+      "loss": 1.0396,
+      "step": 17514
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.993315117686794e-06,
+      "loss": 1.0601,
+      "step": 17515
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.99224056548564e-06,
+      "loss": 1.0389,
+      "step": 17516
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.991166051466937e-06,
+      "loss": 1.1044,
+      "step": 17517
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.990091575644331e-06,
+      "loss": 0.9374,
+      "step": 17518
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9890171380314576e-06,
+      "loss": 1.0116,
+      "step": 17519
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.987942738641963e-06,
+      "loss": 1.1176,
+      "step": 17520
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.98686837748948e-06,
+      "loss": 1.0242,
+      "step": 17521
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.985794054587654e-06,
+      "loss": 0.9714,
+      "step": 17522
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.984719769950114e-06,
+      "loss": 0.9585,
+      "step": 17523
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.983645523590501e-06,
+      "loss": 0.8566,
+      "step": 17524
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9825713155224585e-06,
+      "loss": 0.9116,
+      "step": 17525
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.981497145759618e-06,
+      "loss": 1.0063,
+      "step": 17526
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.980423014315615e-06,
+      "loss": 0.9484,
+      "step": 17527
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.979348921204088e-06,
+      "loss": 0.9439,
+      "step": 17528
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9782748664386655e-06,
+      "loss": 0.974,
+      "step": 17529
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.977200850032992e-06,
+      "loss": 0.984,
+      "step": 17530
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.976126872000697e-06,
+      "loss": 0.8806,
+      "step": 17531
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.975052932355414e-06,
+      "loss": 0.8883,
+      "step": 17532
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.973979031110774e-06,
+      "loss": 1.058,
+      "step": 17533
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.972905168280415e-06,
+      "loss": 0.8162,
+      "step": 17534
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.971831343877966e-06,
+      "loss": 0.9272,
+      "step": 17535
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.970757557917063e-06,
+      "loss": 1.0138,
+      "step": 17536
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.969683810411329e-06,
+      "loss": 0.9268,
+      "step": 17537
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9686101013744e-06,
+      "loss": 0.9066,
+      "step": 17538
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.967536430819909e-06,
+      "loss": 0.9607,
+      "step": 17539
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.966462798761484e-06,
+      "loss": 1.0009,
+      "step": 17540
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.965389205212752e-06,
+      "loss": 0.9929,
+      "step": 17541
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9643156501873436e-06,
+      "loss": 0.9474,
+      "step": 17542
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9632421336988834e-06,
+      "loss": 1.0051,
+      "step": 17543
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.962168655761007e-06,
+      "loss": 0.9944,
+      "step": 17544
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.961095216387334e-06,
+      "loss": 1.0612,
+      "step": 17545
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9600218155915e-06,
+      "loss": 0.9838,
+      "step": 17546
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.958948453387122e-06,
+      "loss": 0.8854,
+      "step": 17547
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.957875129787831e-06,
+      "loss": 0.8431,
+      "step": 17548
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.956801844807255e-06,
+      "loss": 1.0191,
+      "step": 17549
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.955728598459015e-06,
+      "loss": 0.9605,
+      "step": 17550
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.954655390756735e-06,
+      "loss": 1.039,
+      "step": 17551
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.953582221714038e-06,
+      "loss": 1.0356,
+      "step": 17552
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.952509091344554e-06,
+      "loss": 1.029,
+      "step": 17553
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.951435999661902e-06,
+      "loss": 1.038,
+      "step": 17554
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.950362946679703e-06,
+      "loss": 0.8928,
+      "step": 17555
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.94928993241158e-06,
+      "loss": 0.9747,
+      "step": 17556
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.948216956871153e-06,
+      "loss": 0.9902,
+      "step": 17557
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9471440200720494e-06,
+      "loss": 0.9654,
+      "step": 17558
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.946071122027882e-06,
+      "loss": 0.8802,
+      "step": 17559
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.944998262752278e-06,
+      "loss": 0.9484,
+      "step": 17560
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.943925442258848e-06,
+      "loss": 0.8852,
+      "step": 17561
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.942852660561218e-06,
+      "loss": 0.9509,
+      "step": 17562
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9417799176730076e-06,
+      "loss": 0.9479,
+      "step": 17563
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.940707213607831e-06,
+      "loss": 0.9888,
+      "step": 17564
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9396345483793056e-06,
+      "loss": 0.9758,
+      "step": 17565
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.938561922001049e-06,
+      "loss": 1.1084,
+      "step": 17566
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.937489334486682e-06,
+      "loss": 1.0129,
+      "step": 17567
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9364167858498176e-06,
+      "loss": 0.9791,
+      "step": 17568
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9353442761040694e-06,
+      "loss": 0.8936,
+      "step": 17569
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.934271805263051e-06,
+      "loss": 1.0021,
+      "step": 17570
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.933199373340387e-06,
+      "loss": 0.9496,
+      "step": 17571
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.932126980349681e-06,
+      "loss": 0.8968,
+      "step": 17572
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.931054626304553e-06,
+      "loss": 0.8963,
+      "step": 17573
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.929982311218614e-06,
+      "loss": 0.9216,
+      "step": 17574
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.928910035105472e-06,
+      "loss": 0.9076,
+      "step": 17575
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.927837797978747e-06,
+      "loss": 1.0145,
+      "step": 17576
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.926765599852048e-06,
+      "loss": 0.9993,
+      "step": 17577
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.925693440738985e-06,
+      "loss": 1.0701,
+      "step": 17578
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.924621320653169e-06,
+      "loss": 0.8649,
+      "step": 17579
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.923549239608207e-06,
+      "loss": 0.9271,
+      "step": 17580
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.922477197617716e-06,
+      "loss": 1.0434,
+      "step": 17581
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.921405194695301e-06,
+      "loss": 1.0262,
+      "step": 17582
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.920333230854571e-06,
+      "loss": 0.9585,
+      "step": 17583
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.919261306109131e-06,
+      "loss": 0.9789,
+      "step": 17584
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.918189420472596e-06,
+      "loss": 0.7993,
+      "step": 17585
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.917117573958567e-06,
+      "loss": 0.9386,
+      "step": 17586
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.916045766580654e-06,
+      "loss": 1.0773,
+      "step": 17587
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.914973998352463e-06,
+      "loss": 0.9895,
+      "step": 17588
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.913902269287595e-06,
+      "loss": 0.9298,
+      "step": 17589
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.912830579399661e-06,
+      "loss": 0.9922,
+      "step": 17590
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.911758928702265e-06,
+      "loss": 1.0345,
+      "step": 17591
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.910687317209011e-06,
+      "loss": 0.9766,
+      "step": 17592
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.9096157449334975e-06,
+      "loss": 0.9863,
+      "step": 17593
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.908544211889332e-06,
+      "loss": 1.1327,
+      "step": 17594
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.90747271809012e-06,
+      "loss": 0.9192,
+      "step": 17595
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.906401263549461e-06,
+      "loss": 0.9508,
+      "step": 17596
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.905329848280956e-06,
+      "loss": 1.0593,
+      "step": 17597
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.904258472298204e-06,
+      "loss": 0.9462,
+      "step": 17598
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.903187135614813e-06,
+      "loss": 1.0483,
+      "step": 17599
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.902115838244376e-06,
+      "loss": 1.0102,
+      "step": 17600
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.901044580200498e-06,
+      "loss": 1.0243,
+      "step": 17601
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.899973361496776e-06,
+      "loss": 1.0637,
+      "step": 17602
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.898902182146804e-06,
+      "loss": 0.9393,
+      "step": 17603
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.897831042164187e-06,
+      "loss": 0.9625,
+      "step": 17604
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.896759941562523e-06,
+      "loss": 0.9717,
+      "step": 17605
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.895688880355404e-06,
+      "loss": 1.0513,
+      "step": 17606
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.894617858556434e-06,
+      "loss": 0.9973,
+      "step": 17607
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.893546876179198e-06,
+      "loss": 0.8648,
+      "step": 17608
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.892475933237303e-06,
+      "loss": 0.9005,
+      "step": 17609
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8914050297443405e-06,
+      "loss": 0.8986,
+      "step": 17610
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.890334165713903e-06,
+      "loss": 0.9348,
+      "step": 17611
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8892633411595846e-06,
+      "loss": 0.9702,
+      "step": 17612
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.888192556094984e-06,
+      "loss": 0.9427,
+      "step": 17613
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88712181053369e-06,
+      "loss": 0.8906,
+      "step": 17614
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.886051104489298e-06,
+      "loss": 0.9052,
+      "step": 17615
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.884980437975399e-06,
+      "loss": 1.075,
+      "step": 17616
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.88390981100558e-06,
+      "loss": 1.0037,
+      "step": 17617
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.882839223593438e-06,
+      "loss": 0.8563,
+      "step": 17618
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.881768675752567e-06,
+      "loss": 0.929,
+      "step": 17619
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.880698167496549e-06,
+      "loss": 1.0279,
+      "step": 17620
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.87962769883898e-06,
+      "loss": 0.9559,
+      "step": 17621
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8785572697934424e-06,
+      "loss": 1.0256,
+      "step": 17622
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.877486880373532e-06,
+      "loss": 0.9876,
+      "step": 17623
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.876416530592833e-06,
+      "loss": 1.0068,
+      "step": 17624
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.875346220464938e-06,
+      "loss": 0.8054,
+      "step": 17625
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.874275950003424e-06,
+      "loss": 0.9686,
+      "step": 17626
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.873205719221889e-06,
+      "loss": 0.9869,
+      "step": 17627
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.872135528133914e-06,
+      "loss": 1.0506,
+      "step": 17628
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.871065376753086e-06,
+      "loss": 0.9598,
+      "step": 17629
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.869995265092988e-06,
+      "loss": 1.0353,
+      "step": 17630
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.868925193167205e-06,
+      "loss": 0.9658,
+      "step": 17631
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8678551609893225e-06,
+      "loss": 1.0035,
+      "step": 17632
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.866785168572927e-06,
+      "loss": 1.081,
+      "step": 17633
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8657152159315965e-06,
+      "loss": 0.9853,
+      "step": 17634
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.864645303078916e-06,
+      "loss": 0.9762,
+      "step": 17635
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.863575430028465e-06,
+      "loss": 1.0785,
+      "step": 17636
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.862505596793831e-06,
+      "loss": 1.0521,
+      "step": 17637
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8614358033885895e-06,
+      "loss": 0.8571,
+      "step": 17638
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.860366049826327e-06,
+      "loss": 0.9083,
+      "step": 17639
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.859296336120616e-06,
+      "loss": 0.9808,
+      "step": 17640
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.858226662285041e-06,
+      "loss": 0.9275,
+      "step": 17641
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.857157028333181e-06,
+      "loss": 0.9365,
+      "step": 17642
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8560874342786165e-06,
+      "loss": 0.8674,
+      "step": 17643
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85501788013492e-06,
+      "loss": 0.91,
+      "step": 17644
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.85394836591567e-06,
+      "loss": 0.9795,
+      "step": 17645
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.852878891634451e-06,
+      "loss": 1.0652,
+      "step": 17646
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.851809457304833e-06,
+      "loss": 0.9055,
+      "step": 17647
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.850740062940393e-06,
+      "loss": 0.8619,
+      "step": 17648
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.849670708554707e-06,
+      "loss": 1.0122,
+      "step": 17649
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.848601394161349e-06,
+      "loss": 0.9623,
+      "step": 17650
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.847532119773896e-06,
+      "loss": 1.0106,
+      "step": 17651
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8464628854059214e-06,
+      "loss": 0.9813,
+      "step": 17652
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.845393691070998e-06,
+      "loss": 1.0206,
+      "step": 17653
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.844324536782696e-06,
+      "loss": 0.9245,
+      "step": 17654
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.843255422554594e-06,
+      "loss": 0.9689,
+      "step": 17655
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8421863484002615e-06,
+      "loss": 0.9463,
+      "step": 17656
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.84111731433327e-06,
+      "loss": 1.0619,
+      "step": 17657
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8400483203671885e-06,
+      "loss": 0.9549,
+      "step": 17658
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8389793665155855e-06,
+      "loss": 1.05,
+      "step": 17659
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.837910452792041e-06,
+      "loss": 0.9577,
+      "step": 17660
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.836841579210116e-06,
+      "loss": 1.0829,
+      "step": 17661
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83577274578338e-06,
+      "loss": 0.958,
+      "step": 17662
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.834703952525405e-06,
+      "loss": 0.9392,
+      "step": 17663
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.833635199449754e-06,
+      "loss": 0.8956,
+      "step": 17664
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.83256648657e-06,
+      "loss": 1.0884,
+      "step": 17665
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.831497813899706e-06,
+      "loss": 0.8109,
+      "step": 17666
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.830429181452442e-06,
+      "loss": 1.0005,
+      "step": 17667
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.829360589241767e-06,
+      "loss": 0.9218,
+      "step": 17668
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.828292037281253e-06,
+      "loss": 0.9978,
+      "step": 17669
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.827223525584466e-06,
+      "loss": 1.1112,
+      "step": 17670
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.826155054164967e-06,
+      "loss": 0.9016,
+      "step": 17671
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.825086623036317e-06,
+      "loss": 0.9484,
+      "step": 17672
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.824018232212082e-06,
+      "loss": 0.9491,
+      "step": 17673
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8229498817058295e-06,
+      "loss": 0.9485,
+      "step": 17674
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.821881571531117e-06,
+      "loss": 1.0673,
+      "step": 17675
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.820813301701504e-06,
+      "loss": 0.8808,
+      "step": 17676
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.819745072230559e-06,
+      "loss": 0.9777,
+      "step": 17677
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.818676883131834e-06,
+      "loss": 0.9584,
+      "step": 17678
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 6.8176087344188945e-06,
+      "loss": 0.8974,
+      "step": 17679
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.816540626105304e-06,
+      "loss": 1.0169,
+      "step": 17680
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.815472558204617e-06,
+      "loss": 0.8052,
+      "step": 17681
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.814404530730388e-06,
+      "loss": 0.853,
+      "step": 17682
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.81333654369618e-06,
+      "loss": 0.9613,
+      "step": 17683
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.812268597115557e-06,
+      "loss": 0.9665,
+      "step": 17684
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.811200691002068e-06,
+      "loss": 1.022,
+      "step": 17685
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.810132825369269e-06,
+      "loss": 1.0596,
+      "step": 17686
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.809065000230717e-06,
+      "loss": 0.8761,
+      "step": 17687
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.807997215599974e-06,
+      "loss": 0.9543,
+      "step": 17688
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.806929471490591e-06,
+      "loss": 0.9898,
+      "step": 17689
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80586176791612e-06,
+      "loss": 0.9224,
+      "step": 17690
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.804794104890119e-06,
+      "loss": 0.9573,
+      "step": 17691
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.803726482426136e-06,
+      "loss": 0.9585,
+      "step": 17692
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.80265890053773e-06,
+      "loss": 0.9569,
+      "step": 17693
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.801591359238453e-06,
+      "loss": 1.1681,
+      "step": 17694
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.8005238585418574e-06,
+      "loss": 0.9963,
+      "step": 17695
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.799456398461488e-06,
+      "loss": 0.912,
+      "step": 17696
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.798388979010904e-06,
+      "loss": 1.0371,
+      "step": 17697
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7973216002036535e-06,
+      "loss": 0.9072,
+      "step": 17698
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.796254262053287e-06,
+      "loss": 0.9783,
+      "step": 17699
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.79518696457335e-06,
+      "loss": 1.0602,
+      "step": 17700
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.794119707777391e-06,
+      "loss": 1.0305,
+      "step": 17701
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.793052491678969e-06,
+      "loss": 0.9858,
+      "step": 17702
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.791985316291621e-06,
+      "loss": 0.8764,
+      "step": 17703
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.790918181628899e-06,
+      "loss": 0.9341,
+      "step": 17704
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.789851087704351e-06,
+      "loss": 1.1534,
+      "step": 17705
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7887840345315145e-06,
+      "loss": 0.9339,
+      "step": 17706
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7877170221239476e-06,
+      "loss": 1.0231,
+      "step": 17707
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.78665005049519e-06,
+      "loss": 0.9932,
+      "step": 17708
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.785583119658788e-06,
+      "loss": 0.9774,
+      "step": 17709
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.784516229628279e-06,
+      "loss": 0.9502,
+      "step": 17710
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.783449380417217e-06,
+      "loss": 0.8357,
+      "step": 17711
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.782382572039141e-06,
+      "loss": 0.9684,
+      "step": 17712
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.781315804507593e-06,
+      "loss": 0.8408,
+      "step": 17713
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.780249077836117e-06,
+      "loss": 0.9836,
+      "step": 17714
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77918239203825e-06,
+      "loss": 0.9968,
+      "step": 17715
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.77811574712754e-06,
+      "loss": 1.0976,
+      "step": 17716
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.777049143117522e-06,
+      "loss": 0.9128,
+      "step": 17717
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.775982580021742e-06,
+      "loss": 0.9315,
+      "step": 17718
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.774916057853732e-06,
+      "loss": 0.9286,
+      "step": 17719
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.773849576627039e-06,
+      "loss": 1.0298,
+      "step": 17720
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.772783136355197e-06,
+      "loss": 1.0295,
+      "step": 17721
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.771716737051745e-06,
+      "loss": 0.9028,
+      "step": 17722
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.770650378730222e-06,
+      "loss": 1.0246,
+      "step": 17723
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.769584061404159e-06,
+      "loss": 1.0157,
+      "step": 17724
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.768517785087099e-06,
+      "loss": 0.9248,
+      "step": 17725
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.767451549792578e-06,
+      "loss": 0.9156,
+      "step": 17726
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.766385355534129e-06,
+      "loss": 1.0708,
+      "step": 17727
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.765319202325289e-06,
+      "loss": 0.8589,
+      "step": 17728
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7642530901795876e-06,
+      "loss": 0.955,
+      "step": 17729
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.763187019110566e-06,
+      "loss": 0.9944,
+      "step": 17730
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.762120989131751e-06,
+      "loss": 0.9248,
+      "step": 17731
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.761055000256681e-06,
+      "loss": 0.959,
+      "step": 17732
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7599890524988806e-06,
+      "loss": 1.0006,
+      "step": 17733
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.758923145871892e-06,
+      "loss": 0.8906,
+      "step": 17734
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.75785728038924e-06,
+      "loss": 0.978,
+      "step": 17735
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7567914560644564e-06,
+      "loss": 1.0207,
+      "step": 17736
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.755725672911071e-06,
+      "loss": 0.994,
+      "step": 17737
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.754659930942613e-06,
+      "loss": 0.963,
+      "step": 17738
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.753594230172613e-06,
+      "loss": 1.107,
+      "step": 17739
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.752528570614602e-06,
+      "loss": 1.0121,
+      "step": 17740
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.751462952282104e-06,
+      "loss": 0.9376,
+      "step": 17741
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.750397375188651e-06,
+      "loss": 1.0664,
+      "step": 17742
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.749331839347761e-06,
+      "loss": 0.8867,
+      "step": 17743
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.748266344772974e-06,
+      "loss": 1.0038,
+      "step": 17744
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.747200891477806e-06,
+      "loss": 1.0647,
+      "step": 17745
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.746135479475787e-06,
+      "loss": 1.1229,
+      "step": 17746
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.745070108780437e-06,
+      "loss": 0.9983,
+      "step": 17747
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.744004779405289e-06,
+      "loss": 0.8851,
+      "step": 17748
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.742939491363859e-06,
+      "loss": 0.978,
+      "step": 17749
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.741874244669677e-06,
+      "loss": 0.9973,
+      "step": 17750
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.74080903933626e-06,
+      "loss": 0.9722,
+      "step": 17751
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7397438753771305e-06,
+      "loss": 1.0136,
+      "step": 17752
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.738678752805818e-06,
+      "loss": 1.0259,
+      "step": 17753
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.737613671635839e-06,
+      "loss": 0.9728,
+      "step": 17754
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.736548631880711e-06,
+      "loss": 0.995,
+      "step": 17755
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.735483633553958e-06,
+      "loss": 0.9008,
+      "step": 17756
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.734418676669099e-06,
+      "loss": 1.0747,
+      "step": 17757
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7333537612396545e-06,
+      "loss": 1.0818,
+      "step": 17758
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.732288887279141e-06,
+      "loss": 1.0685,
+      "step": 17759
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.73122405480108e-06,
+      "loss": 1.1157,
+      "step": 17760
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.730159263818983e-06,
+      "loss": 0.9808,
+      "step": 17761
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.729094514346376e-06,
+      "loss": 0.9741,
+      "step": 17762
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.728029806396768e-06,
+      "loss": 1.0423,
+      "step": 17763
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.726965139983681e-06,
+      "loss": 0.9121,
+      "step": 17764
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.725900515120625e-06,
+      "loss": 0.9893,
+      "step": 17765
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.724835931821114e-06,
+      "loss": 0.9432,
+      "step": 17766
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.723771390098671e-06,
+      "loss": 0.8304,
+      "step": 17767
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.722706889966806e-06,
+      "loss": 1.0241,
+      "step": 17768
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.721642431439029e-06,
+      "loss": 0.8556,
+      "step": 17769
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.720578014528857e-06,
+      "loss": 0.9081,
+      "step": 17770
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7195136392497975e-06,
+      "loss": 0.9323,
+      "step": 17771
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.718449305615369e-06,
+      "loss": 0.9619,
+      "step": 17772
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.717385013639077e-06,
+      "loss": 1.0074,
+      "step": 17773
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.716320763334439e-06,
+      "loss": 0.9552,
+      "step": 17774
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.715256554714955e-06,
+      "loss": 0.9094,
+      "step": 17775
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.714192387794143e-06,
+      "loss": 0.9162,
+      "step": 17776
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.713128262585514e-06,
+      "loss": 0.9075,
+      "step": 17777
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.712064179102571e-06,
+      "loss": 0.8895,
+      "step": 17778
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7110001373588215e-06,
+      "loss": 0.9804,
+      "step": 17779
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.709936137367775e-06,
+      "loss": 0.9569,
+      "step": 17780
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.708872179142944e-06,
+      "loss": 1.0413,
+      "step": 17781
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.707808262697828e-06,
+      "loss": 0.9749,
+      "step": 17782
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.706744388045936e-06,
+      "loss": 1.083,
+      "step": 17783
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.705680555200773e-06,
+      "loss": 1.0408,
+      "step": 17784
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.704616764175839e-06,
+      "loss": 0.9574,
+      "step": 17785
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.7035530149846475e-06,
+      "loss": 0.9018,
+      "step": 17786
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.702489307640701e-06,
+      "loss": 0.9593,
+      "step": 17787
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.701425642157498e-06,
+      "loss": 0.9423,
+      "step": 17788
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.70036201854854e-06,
+      "loss": 1.0186,
+      "step": 17789
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.699298436827335e-06,
+      "loss": 0.932,
+      "step": 17790
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.698234897007384e-06,
+      "loss": 0.9238,
+      "step": 17791
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.697171399102187e-06,
+      "loss": 1.0126,
+      "step": 17792
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.696107943125243e-06,
+      "loss": 0.9945,
+      "step": 17793
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.695044529090053e-06,
+      "loss": 0.9759,
+      "step": 17794
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.69398115701012e-06,
+      "loss": 1.1198,
+      "step": 17795
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.692917826898942e-06,
+      "loss": 0.931,
+      "step": 17796
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6918545387700125e-06,
+      "loss": 0.8805,
+      "step": 17797
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6907912926368355e-06,
+      "loss": 1.0748,
+      "step": 17798
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.689728088512904e-06,
+      "loss": 0.9922,
+      "step": 17799
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.688664926411717e-06,
+      "loss": 1.0381,
+      "step": 17800
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.687601806346775e-06,
+      "loss": 0.8874,
+      "step": 17801
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.686538728331571e-06,
+      "loss": 0.9446,
+      "step": 17802
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.685475692379594e-06,
+      "loss": 0.9037,
+      "step": 17803
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.684412698504348e-06,
+      "loss": 1.1842,
+      "step": 17804
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.683349746719324e-06,
+      "loss": 0.9833,
+      "step": 17805
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6822868370380165e-06,
+      "loss": 0.9403,
+      "step": 17806
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.681223969473916e-06,
+      "loss": 1.0517,
+      "step": 17807
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.680161144040516e-06,
+      "loss": 0.9154,
+      "step": 17808
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.679098360751312e-06,
+      "loss": 1.0868,
+      "step": 17809
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6780356196197935e-06,
+      "loss": 0.8957,
+      "step": 17810
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.676972920659452e-06,
+      "loss": 0.8819,
+      "step": 17811
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.67591026388378e-06,
+      "loss": 1.0428,
+      "step": 17812
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6748476493062596e-06,
+      "loss": 0.9918,
+      "step": 17813
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.673785076940389e-06,
+      "loss": 1.0385,
+      "step": 17814
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.672722546799655e-06,
+      "loss": 0.9632,
+      "step": 17815
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6716600588975455e-06,
+      "loss": 1.0357,
+      "step": 17816
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.670597613247544e-06,
+      "loss": 0.8937,
+      "step": 17817
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.669535209863145e-06,
+      "loss": 0.9524,
+      "step": 17818
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.668472848757834e-06,
+      "loss": 0.9836,
+      "step": 17819
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.667410529945095e-06,
+      "loss": 0.847,
+      "step": 17820
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.666348253438414e-06,
+      "loss": 0.9854,
+      "step": 17821
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.665286019251274e-06,
+      "loss": 1.0607,
+      "step": 17822
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.664223827397166e-06,
+      "loss": 0.8984,
+      "step": 17823
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.663161677889569e-06,
+      "loss": 0.9375,
+      "step": 17824
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.662099570741969e-06,
+      "loss": 0.9207,
+      "step": 17825
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.661037505967849e-06,
+      "loss": 1.0592,
+      "step": 17826
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6599754835806875e-06,
+      "loss": 0.9782,
+      "step": 17827
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.658913503593971e-06,
+      "loss": 0.8793,
+      "step": 17828
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.657851566021182e-06,
+      "loss": 1.0298,
+      "step": 17829
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6567896708757986e-06,
+      "loss": 0.9825,
+      "step": 17830
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.655727818171298e-06,
+      "loss": 1.1241,
+      "step": 17831
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.654666007921167e-06,
+      "loss": 0.8782,
+      "step": 17832
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.653604240138882e-06,
+      "loss": 1.0489,
+      "step": 17833
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.652542514837921e-06,
+      "loss": 0.9673,
+      "step": 17834
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.651480832031763e-06,
+      "loss": 0.9161,
+      "step": 17835
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6504191917338835e-06,
+      "loss": 0.97,
+      "step": 17836
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.649357593957764e-06,
+      "loss": 1.0338,
+      "step": 17837
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.648296038716877e-06,
+      "loss": 0.8873,
+      "step": 17838
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.647234526024702e-06,
+      "loss": 0.9087,
+      "step": 17839
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6461730558947135e-06,
+      "loss": 0.9403,
+      "step": 17840
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.645111628340381e-06,
+      "loss": 1.0384,
+      "step": 17841
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.644050243375186e-06,
+      "loss": 0.935,
+      "step": 17842
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.642988901012602e-06,
+      "loss": 1.056,
+      "step": 17843
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.641927601266099e-06,
+      "loss": 0.9738,
+      "step": 17844
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.640866344149151e-06,
+      "loss": 0.9195,
+      "step": 17845
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.639805129675231e-06,
+      "loss": 0.9218,
+      "step": 17846
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.638743957857812e-06,
+      "loss": 1.0691,
+      "step": 17847
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.637682828710362e-06,
+      "loss": 0.9771,
+      "step": 17848
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.636621742246354e-06,
+      "loss": 1.0611,
+      "step": 17849
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6355606984792555e-06,
+      "loss": 0.947,
+      "step": 17850
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.63449969742254e-06,
+      "loss": 1.1713,
+      "step": 17851
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.633438739089673e-06,
+      "loss": 1.0109,
+      "step": 17852
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.632377823494127e-06,
+      "loss": 0.8134,
+      "step": 17853
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.631316950649363e-06,
+      "loss": 0.9379,
+      "step": 17854
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.630256120568856e-06,
+      "loss": 0.9085,
+      "step": 17855
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.629195333266068e-06,
+      "loss": 0.9551,
+      "step": 17856
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6281345887544715e-06,
+      "loss": 0.9092,
+      "step": 17857
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.627073887047522e-06,
+      "loss": 0.981,
+      "step": 17858
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.626013228158692e-06,
+      "loss": 1.0162,
+      "step": 17859
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.624952612101446e-06,
+      "loss": 0.9736,
+      "step": 17860
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.623892038889248e-06,
+      "loss": 0.8663,
+      "step": 17861
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6228315085355585e-06,
+      "loss": 0.8876,
+      "step": 17862
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.621771021053845e-06,
+      "loss": 1.0213,
+      "step": 17863
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.620710576457563e-06,
+      "loss": 0.9452,
+      "step": 17864
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6196501747601835e-06,
+      "loss": 1.0127,
+      "step": 17865
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.618589815975161e-06,
+      "loss": 0.9787,
+      "step": 17866
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.617529500115961e-06,
+      "loss": 1.0008,
+      "step": 17867
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6164692271960385e-06,
+      "loss": 0.936,
+      "step": 17868
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6154089972288596e-06,
+      "loss": 0.8866,
+      "step": 17869
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.614348810227878e-06,
+      "loss": 0.9688,
+      "step": 17870
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.6132886662065575e-06,
+      "loss": 0.9393,
+      "step": 17871
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.612228565178351e-06,
+      "loss": 1.0526,
+      "step": 17872
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.611168507156718e-06,
+      "loss": 0.9128,
+      "step": 17873
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.610108492155119e-06,
+      "loss": 0.9845,
+      "step": 17874
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.609048520187009e-06,
+      "loss": 0.9714,
+      "step": 17875
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.60798859126584e-06,
+      "loss": 0.9442,
+      "step": 17876
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.606928705405072e-06,
+      "loss": 0.9066,
+      "step": 17877
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.605868862618154e-06,
+      "loss": 0.9925,
+      "step": 17878
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.604809062918549e-06,
+      "loss": 1.062,
+      "step": 17879
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.603749306319705e-06,
+      "loss": 0.8921,
+      "step": 17880
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.602689592835076e-06,
+      "loss": 0.9404,
+      "step": 17881
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.601629922478114e-06,
+      "loss": 0.9332,
+      "step": 17882
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.600570295262272e-06,
+      "loss": 1.0135,
+      "step": 17883
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.599510711201005e-06,
+      "loss": 0.8458,
+      "step": 17884
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.59845117030776e-06,
+      "loss": 0.9295,
+      "step": 17885
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.597391672595988e-06,
+      "loss": 0.9465,
+      "step": 17886
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.596332218079137e-06,
+      "loss": 0.9592,
+      "step": 17887
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.595272806770662e-06,
+      "loss": 1.0821,
+      "step": 17888
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.59421343868401e-06,
+      "loss": 0.9334,
+      "step": 17889
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.593154113832624e-06,
+      "loss": 1.0462,
+      "step": 17890
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.592094832229958e-06,
+      "loss": 0.9545,
+      "step": 17891
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.591035593889454e-06,
+      "loss": 0.8529,
+      "step": 17892
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.589976398824564e-06,
+      "loss": 1.0839,
+      "step": 17893
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.588917247048732e-06,
+      "loss": 0.8938,
+      "step": 17894
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5878581385754035e-06,
+      "loss": 0.9356,
+      "step": 17895
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58679907341802e-06,
+      "loss": 0.9531,
+      "step": 17896
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5857400515900305e-06,
+      "loss": 0.9955,
+      "step": 17897
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5846810731048806e-06,
+      "loss": 0.9161,
+      "step": 17898
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58362213797601e-06,
+      "loss": 1.1057,
+      "step": 17899
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.58256324621686e-06,
+      "loss": 0.9671,
+      "step": 17900
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.581504397840874e-06,
+      "loss": 1.0165,
+      "step": 17901
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5804455928614985e-06,
+      "loss": 0.9703,
+      "step": 17902
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.579386831292171e-06,
+      "loss": 0.8766,
+      "step": 17903
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.578328113146329e-06,
+      "loss": 0.9497,
+      "step": 17904
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.577269438437419e-06,
+      "loss": 1.0914,
+      "step": 17905
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.576210807178872e-06,
+      "loss": 1.0168,
+      "step": 17906
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.575152219384135e-06,
+      "loss": 0.9547,
+      "step": 17907
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.574093675066645e-06,
+      "loss": 0.9201,
+      "step": 17908
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.573035174239839e-06,
+      "loss": 1.0101,
+      "step": 17909
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5719767169171475e-06,
+      "loss": 1.1162,
+      "step": 17910
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.570918303112017e-06,
+      "loss": 1.0238,
+      "step": 17911
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.569859932837883e-06,
+      "loss": 0.8887,
+      "step": 17912
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.568801606108177e-06,
+      "loss": 1.0152,
+      "step": 17913
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.567743322936333e-06,
+      "loss": 0.978,
+      "step": 17914
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.566685083335786e-06,
+      "loss": 0.9198,
+      "step": 17915
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.565626887319976e-06,
+      "loss": 0.9236,
+      "step": 17916
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.56456873490233e-06,
+      "loss": 0.7561,
+      "step": 17917
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.563510626096286e-06,
+      "loss": 0.8762,
+      "step": 17918
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.562452560915273e-06,
+      "loss": 0.9881,
+      "step": 17919
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.561394539372718e-06,
+      "loss": 0.9574,
+      "step": 17920
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5603365614820605e-06,
+      "loss": 0.986,
+      "step": 17921
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.55927862725673e-06,
+      "loss": 0.8729,
+      "step": 17922
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.558220736710155e-06,
+      "loss": 0.9672,
+      "step": 17923
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.557162889855759e-06,
+      "loss": 0.9997,
+      "step": 17924
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.556105086706981e-06,
+      "loss": 0.9317,
+      "step": 17925
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.555047327277247e-06,
+      "loss": 0.9756,
+      "step": 17926
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.55398961157998e-06,
+      "loss": 0.9361,
+      "step": 17927
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5529319396286106e-06,
+      "loss": 0.9302,
+      "step": 17928
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.551874311436564e-06,
+      "loss": 1.0761,
+      "step": 17929
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.55081672701727e-06,
+      "loss": 1.0047,
+      "step": 17930
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.549759186384153e-06,
+      "loss": 0.835,
+      "step": 17931
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.548701689550636e-06,
+      "loss": 0.9969,
+      "step": 17932
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.547644236530147e-06,
+      "loss": 0.9435,
+      "step": 17933
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5465868273361026e-06,
+      "loss": 0.9791,
+      "step": 17934
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.545529461981932e-06,
+      "loss": 1.0748,
+      "step": 17935
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.544472140481062e-06,
+      "loss": 0.9779,
+      "step": 17936
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.543414862846908e-06,
+      "loss": 0.9943,
+      "step": 17937
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.542357629092892e-06,
+      "loss": 0.8274,
+      "step": 17938
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5413004392324385e-06,
+      "loss": 1.0148,
+      "step": 17939
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.540243293278969e-06,
+      "loss": 0.8895,
+      "step": 17940
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5391861912459e-06,
+      "loss": 1.0547,
+      "step": 17941
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.538129133146653e-06,
+      "loss": 0.9331,
+      "step": 17942
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.537072118994645e-06,
+      "loss": 0.96,
+      "step": 17943
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.536015148803298e-06,
+      "loss": 0.9936,
+      "step": 17944
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.534958222586028e-06,
+      "loss": 0.9285,
+      "step": 17945
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.533901340356252e-06,
+      "loss": 0.9447,
+      "step": 17946
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.532844502127389e-06,
+      "loss": 0.9672,
+      "step": 17947
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.531787707912848e-06,
+      "loss": 0.9468,
+      "step": 17948
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.530730957726052e-06,
+      "loss": 0.933,
+      "step": 17949
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.529674251580416e-06,
+      "loss": 0.9738,
+      "step": 17950
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.528617589489352e-06,
+      "loss": 0.9689,
+      "step": 17951
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.527560971466269e-06,
+      "loss": 0.9263,
+      "step": 17952
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52650439752459e-06,
+      "loss": 1.0441,
+      "step": 17953
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.525447867677726e-06,
+      "loss": 1.0147,
+      "step": 17954
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.524391381939082e-06,
+      "loss": 0.9484,
+      "step": 17955
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.523334940322079e-06,
+      "loss": 0.9661,
+      "step": 17956
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.522278542840118e-06,
+      "loss": 1.0599,
+      "step": 17957
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.52122218950662e-06,
+      "loss": 1.0965,
+      "step": 17958
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.5201658803349886e-06,
+      "loss": 0.9032,
+      "step": 17959
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.519109615338637e-06,
+      "loss": 1.0078,
+      "step": 17960
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.518053394530972e-06,
+      "loss": 1.0229,
+      "step": 17961
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.516997217925398e-06,
+      "loss": 0.9132,
+      "step": 17962
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.515941085535329e-06,
+      "loss": 0.9323,
+      "step": 17963
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.514884997374172e-06,
+      "loss": 0.9881,
+      "step": 17964
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.51382895345533e-06,
+      "loss": 1.0344,
+      "step": 17965
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 6.512772953792209e-06,
+      "loss": 0.9391,
+      "step": 17966
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5117169983982206e-06,
+      "loss": 0.9484,
+      "step": 17967
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.510661087286765e-06,
+      "loss": 0.9284,
+      "step": 17968
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.509605220471245e-06,
+      "loss": 0.9218,
+      "step": 17969
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.5085493979650695e-06,
+      "loss": 1.1179,
+      "step": 17970
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.507493619781635e-06,
+      "loss": 0.9643,
+      "step": 17971
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.506437885934351e-06,
+      "loss": 0.9463,
+      "step": 17972
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.505382196436617e-06,
+      "loss": 1.0294,
+      "step": 17973
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.504326551301835e-06,
+      "loss": 0.9509,
+      "step": 17974
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.503270950543405e-06,
+      "loss": 0.9327,
+      "step": 17975
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.502215394174727e-06,
+      "loss": 1.0172,
+      "step": 17976
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.501159882209201e-06,
+      "loss": 0.9398,
+      "step": 17977
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.500104414660231e-06,
+      "loss": 0.9499,
+      "step": 17978
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.499048991541208e-06,
+      "loss": 0.9472,
+      "step": 17979
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.497993612865534e-06,
+      "loss": 1.023,
+      "step": 17980
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.496938278646612e-06,
+      "loss": 0.8804,
+      "step": 17981
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4958829888978325e-06,
+      "loss": 0.882,
+      "step": 17982
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.494827743632593e-06,
+      "loss": 0.9785,
+      "step": 17983
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.49377254286429e-06,
+      "loss": 0.9265,
+      "step": 17984
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.492717386606315e-06,
+      "loss": 1.0353,
+      "step": 17985
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.491662274872072e-06,
+      "loss": 1.0033,
+      "step": 17986
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.490607207674949e-06,
+      "loss": 1.1063,
+      "step": 17987
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.489552185028342e-06,
+      "loss": 0.9807,
+      "step": 17988
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.488497206945641e-06,
+      "loss": 1.1353,
+      "step": 17989
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.487442273440239e-06,
+      "loss": 0.9912,
+      "step": 17990
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.486387384525532e-06,
+      "loss": 1.0136,
+      "step": 17991
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.48533254021491e-06,
+      "loss": 0.972,
+      "step": 17992
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4842777405217605e-06,
+      "loss": 0.8916,
+      "step": 17993
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4832229854594754e-06,
+      "loss": 0.9608,
+      "step": 17994
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4821682750414485e-06,
+      "loss": 0.8816,
+      "step": 17995
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.481113609281067e-06,
+      "loss": 0.9221,
+      "step": 17996
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.480058988191713e-06,
+      "loss": 0.9255,
+      "step": 17997
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4790044117867845e-06,
+      "loss": 0.8423,
+      "step": 17998
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.477949880079659e-06,
+      "loss": 1.0464,
+      "step": 17999
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.476895393083734e-06,
+      "loss": 1.0509,
+      "step": 18000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.475840950812389e-06,
+      "loss": 1.0933,
+      "step": 18001
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.474786553279011e-06,
+      "loss": 0.9091,
+      "step": 18002
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4737322004969825e-06,
+      "loss": 0.8699,
+      "step": 18003
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.472677892479693e-06,
+      "loss": 0.9255,
+      "step": 18004
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.471623629240528e-06,
+      "loss": 0.846,
+      "step": 18005
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.470569410792866e-06,
+      "loss": 1.0632,
+      "step": 18006
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.469515237150091e-06,
+      "loss": 0.9917,
+      "step": 18007
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.468461108325583e-06,
+      "loss": 1.0691,
+      "step": 18008
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.467407024332732e-06,
+      "loss": 0.9444,
+      "step": 18009
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.466352985184913e-06,
+      "loss": 1.0721,
+      "step": 18010
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.465298990895506e-06,
+      "loss": 1.0466,
+      "step": 18011
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.464245041477894e-06,
+      "loss": 0.9163,
+      "step": 18012
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4631911369454525e-06,
+      "loss": 0.9476,
+      "step": 18013
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.462137277311566e-06,
+      "loss": 0.9772,
+      "step": 18014
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4610834625896105e-06,
+      "loss": 0.9421,
+      "step": 18015
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.460029692792964e-06,
+      "loss": 0.9868,
+      "step": 18016
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.458975967934999e-06,
+      "loss": 0.9749,
+      "step": 18017
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.457922288029099e-06,
+      "loss": 0.9034,
+      "step": 18018
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.456868653088639e-06,
+      "loss": 0.9956,
+      "step": 18019
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.455815063126993e-06,
+      "loss": 1.0055,
+      "step": 18020
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.454761518157534e-06,
+      "loss": 1.0243,
+      "step": 18021
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.453708018193637e-06,
+      "loss": 1.0069,
+      "step": 18022
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.452654563248679e-06,
+      "loss": 1.0458,
+      "step": 18023
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.451601153336034e-06,
+      "loss": 1.0173,
+      "step": 18024
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.450547788469068e-06,
+      "loss": 0.9075,
+      "step": 18025
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.449494468661161e-06,
+      "loss": 0.895,
+      "step": 18026
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.448441193925676e-06,
+      "loss": 0.9424,
+      "step": 18027
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.447387964275991e-06,
+      "loss": 1.0018,
+      "step": 18028
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.446334779725475e-06,
+      "loss": 0.9739,
+      "step": 18029
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.445281640287498e-06,
+      "loss": 0.9712,
+      "step": 18030
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4442285459754226e-06,
+      "loss": 0.9024,
+      "step": 18031
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.443175496802625e-06,
+      "loss": 1.032,
+      "step": 18032
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.442122492782474e-06,
+      "loss": 0.9803,
+      "step": 18033
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.441069533928335e-06,
+      "loss": 0.9731,
+      "step": 18034
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.440016620253572e-06,
+      "loss": 1.0723,
+      "step": 18035
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4389637517715505e-06,
+      "loss": 0.9972,
+      "step": 18036
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.437910928495645e-06,
+      "loss": 0.9225,
+      "step": 18037
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.436858150439211e-06,
+      "loss": 0.9661,
+      "step": 18038
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.435805417615622e-06,
+      "loss": 1.1562,
+      "step": 18039
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.434752730038236e-06,
+      "loss": 0.9484,
+      "step": 18040
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.433700087720413e-06,
+      "loss": 0.9467,
+      "step": 18041
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.432647490675525e-06,
+      "loss": 1.2194,
+      "step": 18042
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.431594938916932e-06,
+      "loss": 0.991,
+      "step": 18043
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.430542432457993e-06,
+      "loss": 1.0599,
+      "step": 18044
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.429489971312065e-06,
+      "loss": 0.9833,
+      "step": 18045
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.428437555492517e-06,
+      "loss": 0.8975,
+      "step": 18046
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.427385185012709e-06,
+      "loss": 0.9134,
+      "step": 18047
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.426332859885993e-06,
+      "loss": 0.9325,
+      "step": 18048
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.425280580125736e-06,
+      "loss": 0.9563,
+      "step": 18049
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4242283457452866e-06,
+      "loss": 0.8285,
+      "step": 18050
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.423176156758015e-06,
+      "loss": 0.9975,
+      "step": 18051
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.422124013177269e-06,
+      "loss": 0.8787,
+      "step": 18052
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.421071915016409e-06,
+      "loss": 0.9512,
+      "step": 18053
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4200198622887914e-06,
+      "loss": 1.0307,
+      "step": 18054
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.418967855007765e-06,
+      "loss": 0.9469,
+      "step": 18055
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.417915893186693e-06,
+      "loss": 0.8876,
+      "step": 18056
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.416863976838929e-06,
+      "loss": 1.1172,
+      "step": 18057
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.415812105977823e-06,
+      "loss": 1.0018,
+      "step": 18058
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.414760280616726e-06,
+      "loss": 0.9965,
+      "step": 18059
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.413708500768997e-06,
+      "loss": 0.9267,
+      "step": 18060
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.412656766447987e-06,
+      "loss": 0.9073,
+      "step": 18061
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.411605077667042e-06,
+      "loss": 0.964,
+      "step": 18062
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.410553434439519e-06,
+      "loss": 0.9357,
+      "step": 18063
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.409501836778762e-06,
+      "loss": 0.7953,
+      "step": 18064
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4084502846981265e-06,
+      "loss": 0.9653,
+      "step": 18065
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.407398778210959e-06,
+      "loss": 0.9137,
+      "step": 18066
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.40634731733061e-06,
+      "loss": 0.9504,
+      "step": 18067
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.405295902070425e-06,
+      "loss": 1.0013,
+      "step": 18068
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4042445324437485e-06,
+      "loss": 1.0631,
+      "step": 18069
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.403193208463933e-06,
+      "loss": 0.9263,
+      "step": 18070
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4021419301443254e-06,
+      "loss": 0.9672,
+      "step": 18071
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.401090697498265e-06,
+      "loss": 1.034,
+      "step": 18072
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.4000395105391e-06,
+      "loss": 0.834,
+      "step": 18073
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.398988369280176e-06,
+      "loss": 1.0122,
+      "step": 18074
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.397937273734838e-06,
+      "loss": 1.023,
+      "step": 18075
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3968862239164255e-06,
+      "loss": 0.9433,
+      "step": 18076
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.395835219838284e-06,
+      "loss": 1.0872,
+      "step": 18077
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3947842615137515e-06,
+      "loss": 1.0281,
+      "step": 18078
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3937333489561746e-06,
+      "loss": 0.9559,
+      "step": 18079
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.392682482178892e-06,
+      "loss": 1.0636,
+      "step": 18080
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3916316611952465e-06,
+      "loss": 0.9276,
+      "step": 18081
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.390580886018576e-06,
+      "loss": 0.9279,
+      "step": 18082
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.389530156662215e-06,
+      "loss": 0.9774,
+      "step": 18083
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.388479473139508e-06,
+      "loss": 1.0051,
+      "step": 18084
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.387428835463795e-06,
+      "loss": 0.9373,
+      "step": 18085
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3863782436484074e-06,
+      "loss": 1.0576,
+      "step": 18086
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.385327697706683e-06,
+      "loss": 0.8953,
+      "step": 18087
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.384277197651965e-06,
+      "loss": 0.9287,
+      "step": 18088
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.383226743497582e-06,
+      "loss": 1.0107,
+      "step": 18089
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.382176335256871e-06,
+      "loss": 0.9979,
+      "step": 18090
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.38112597294317e-06,
+      "loss": 0.9978,
+      "step": 18091
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.380075656569803e-06,
+      "loss": 0.9664,
+      "step": 18092
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.379025386150116e-06,
+      "loss": 0.8306,
+      "step": 18093
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.377975161697434e-06,
+      "loss": 1.026,
+      "step": 18094
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.376924983225093e-06,
+      "loss": 1.0103,
+      "step": 18095
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.375874850746419e-06,
+      "loss": 1.0466,
+      "step": 18096
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.374824764274747e-06,
+      "loss": 0.9618,
+      "step": 18097
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.373774723823412e-06,
+      "loss": 0.9245,
+      "step": 18098
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.372724729405739e-06,
+      "loss": 0.9729,
+      "step": 18099
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3716747810350555e-06,
+      "loss": 1.0285,
+      "step": 18100
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.37062487872469e-06,
+      "loss": 0.9691,
+      "step": 18101
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3695750224879784e-06,
+      "loss": 0.9333,
+      "step": 18102
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.368525212338241e-06,
+      "loss": 1.03,
+      "step": 18103
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.367475448288807e-06,
+      "loss": 0.8798,
+      "step": 18104
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.366425730353003e-06,
+      "loss": 0.9912,
+      "step": 18105
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36537605854415e-06,
+      "loss": 0.9697,
+      "step": 18106
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3643264328755826e-06,
+      "loss": 1.0461,
+      "step": 18107
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.363276853360617e-06,
+      "loss": 0.9226,
+      "step": 18108
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.362227320012584e-06,
+      "loss": 0.9186,
+      "step": 18109
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3611778328448e-06,
+      "loss": 0.8719,
+      "step": 18110
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.36012839187059e-06,
+      "loss": 0.8866,
+      "step": 18111
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.359078997103283e-06,
+      "loss": 0.9849,
+      "step": 18112
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.358029648556194e-06,
+      "loss": 0.9268,
+      "step": 18113
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.356980346242645e-06,
+      "loss": 0.9322,
+      "step": 18114
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3559310901759545e-06,
+      "loss": 1.0265,
+      "step": 18115
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.354881880369448e-06,
+      "loss": 0.8512,
+      "step": 18116
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.353832716836443e-06,
+      "loss": 0.9502,
+      "step": 18117
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.352783599590256e-06,
+      "loss": 0.9139,
+      "step": 18118
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.351734528644207e-06,
+      "loss": 1.016,
+      "step": 18119
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.35068550401161e-06,
+      "loss": 0.9232,
+      "step": 18120
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.349636525705787e-06,
+      "loss": 1.0898,
+      "step": 18121
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.348587593740053e-06,
+      "loss": 0.9778,
+      "step": 18122
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.347538708127724e-06,
+      "loss": 0.9506,
+      "step": 18123
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.346489868882113e-06,
+      "loss": 1.0144,
+      "step": 18124
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.345441076016533e-06,
+      "loss": 0.9796,
+      "step": 18125
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.344392329544305e-06,
+      "loss": 1.1057,
+      "step": 18126
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.34334362947874e-06,
+      "loss": 0.9907,
+      "step": 18127
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.342294975833146e-06,
+      "loss": 1.0102,
+      "step": 18128
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.341246368620838e-06,
+      "loss": 0.9062,
+      "step": 18129
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3401978078551315e-06,
+      "loss": 0.8924,
+      "step": 18130
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.339149293549336e-06,
+      "loss": 0.9039,
+      "step": 18131
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.338100825716758e-06,
+      "loss": 1.0348,
+      "step": 18132
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.337052404370711e-06,
+      "loss": 1.0436,
+      "step": 18133
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.336004029524501e-06,
+      "loss": 0.9565,
+      "step": 18134
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.334955701191438e-06,
+      "loss": 0.9487,
+      "step": 18135
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.333907419384837e-06,
+      "loss": 0.9338,
+      "step": 18136
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.332859184117999e-06,
+      "loss": 0.9576,
+      "step": 18137
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.331810995404226e-06,
+      "loss": 0.9257,
+      "step": 18138
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.330762853256834e-06,
+      "loss": 0.9111,
+      "step": 18139
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.329714757689127e-06,
+      "loss": 0.9485,
+      "step": 18140
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.328666708714408e-06,
+      "loss": 1.0669,
+      "step": 18141
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.327618706345979e-06,
+      "loss": 0.948,
+      "step": 18142
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.326570750597146e-06,
+      "loss": 0.9457,
+      "step": 18143
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.325522841481217e-06,
+      "loss": 0.8679,
+      "step": 18144
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.32447497901149e-06,
+      "loss": 1.0508,
+      "step": 18145
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.323427163201269e-06,
+      "loss": 0.9367,
+      "step": 18146
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3223793940638555e-06,
+      "loss": 0.9397,
+      "step": 18147
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.321331671612547e-06,
+      "loss": 0.978,
+      "step": 18148
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.320283995860648e-06,
+      "loss": 0.9252,
+      "step": 18149
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.31923636682146e-06,
+      "loss": 0.9542,
+      "step": 18150
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.318188784508279e-06,
+      "loss": 0.9865,
+      "step": 18151
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.317141248934402e-06,
+      "loss": 0.9417,
+      "step": 18152
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.316093760113129e-06,
+      "loss": 0.9215,
+      "step": 18153
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.315046318057763e-06,
+      "loss": 1.0227,
+      "step": 18154
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.313998922781592e-06,
+      "loss": 0.9901,
+      "step": 18155
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.312951574297919e-06,
+      "loss": 0.9651,
+      "step": 18156
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.311904272620033e-06,
+      "loss": 0.9245,
+      "step": 18157
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.310857017761236e-06,
+      "loss": 0.922,
+      "step": 18158
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.309809809734819e-06,
+      "loss": 1.0722,
+      "step": 18159
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.308762648554079e-06,
+      "loss": 0.9538,
+      "step": 18160
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.307715534232306e-06,
+      "loss": 0.9537,
+      "step": 18161
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.3066684667827904e-06,
+      "loss": 1.0622,
+      "step": 18162
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.305621446218828e-06,
+      "loss": 1.1135,
+      "step": 18163
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.304574472553715e-06,
+      "loss": 0.9826,
+      "step": 18164
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.303527545800735e-06,
+      "loss": 1.0554,
+      "step": 18165
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.302480665973177e-06,
+      "loss": 0.9882,
+      "step": 18166
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.301433833084337e-06,
+      "loss": 0.9707,
+      "step": 18167
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.300387047147504e-06,
+      "loss": 0.9371,
+      "step": 18168
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.299340308175962e-06,
+      "loss": 0.9206,
+      "step": 18169
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.298293616183002e-06,
+      "loss": 1.0046,
+      "step": 18170
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.297246971181909e-06,
+      "loss": 1.0183,
+      "step": 18171
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.296200373185974e-06,
+      "loss": 1.0202,
+      "step": 18172
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.295153822208479e-06,
+      "loss": 0.9733,
+      "step": 18173
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.294107318262713e-06,
+      "loss": 0.8894,
+      "step": 18174
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.293060861361959e-06,
+      "loss": 0.9726,
+      "step": 18175
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.292014451519499e-06,
+      "loss": 0.9476,
+      "step": 18176
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.29096808874862e-06,
+      "loss": 0.8732,
+      "step": 18177
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.289921773062608e-06,
+      "loss": 1.0875,
+      "step": 18178
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.288875504474738e-06,
+      "loss": 0.9714,
+      "step": 18179
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.287829282998299e-06,
+      "loss": 0.9312,
+      "step": 18180
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.286783108646568e-06,
+      "loss": 0.73,
+      "step": 18181
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2857369814328305e-06,
+      "loss": 0.938,
+      "step": 18182
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.28469090137036e-06,
+      "loss": 1.072,
+      "step": 18183
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.283644868472445e-06,
+      "loss": 1.0041,
+      "step": 18184
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.282598882752353e-06,
+      "loss": 0.9614,
+      "step": 18185
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.281552944223373e-06,
+      "loss": 0.921,
+      "step": 18186
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2805070528987765e-06,
+      "loss": 0.9608,
+      "step": 18187
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.279461208791846e-06,
+      "loss": 1.069,
+      "step": 18188
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.278415411915854e-06,
+      "loss": 1.0556,
+      "step": 18189
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.277369662284074e-06,
+      "loss": 1.0206,
+      "step": 18190
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.276323959909787e-06,
+      "loss": 1.0295,
+      "step": 18191
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.275278304806269e-06,
+      "loss": 0.943,
+      "step": 18192
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.274232696986786e-06,
+      "loss": 0.9708,
+      "step": 18193
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2731871364646165e-06,
+      "loss": 0.9127,
+      "step": 18194
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.272141623253039e-06,
+      "loss": 0.9452,
+      "step": 18195
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.271096157365317e-06,
+      "loss": 0.9751,
+      "step": 18196
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2700507388147256e-06,
+      "loss": 0.836,
+      "step": 18197
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26900536761454e-06,
+      "loss": 0.9649,
+      "step": 18198
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.267960043778021e-06,
+      "loss": 1.0445,
+      "step": 18199
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26691476731845e-06,
+      "loss": 1.0207,
+      "step": 18200
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.265869538249088e-06,
+      "loss": 0.9042,
+      "step": 18201
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.26482435658321e-06,
+      "loss": 0.9382,
+      "step": 18202
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2637792223340786e-06,
+      "loss": 1.0384,
+      "step": 18203
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.262734135514965e-06,
+      "loss": 1.0515,
+      "step": 18204
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2616890961391344e-06,
+      "loss": 0.9438,
+      "step": 18205
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.260644104219855e-06,
+      "loss": 0.9883,
+      "step": 18206
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.259599159770391e-06,
+      "loss": 0.9543,
+      "step": 18207
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.258554262804006e-06,
+      "loss": 0.9505,
+      "step": 18208
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.257509413333971e-06,
+      "loss": 0.799,
+      "step": 18209
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.256464611373546e-06,
+      "loss": 0.9775,
+      "step": 18210
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.255419856935993e-06,
+      "loss": 0.9345,
+      "step": 18211
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.254375150034577e-06,
+      "loss": 1.1221,
+      "step": 18212
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.253330490682555e-06,
+      "loss": 0.954,
+      "step": 18213
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.252285878893196e-06,
+      "loss": 0.9788,
+      "step": 18214
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.251241314679758e-06,
+      "loss": 1.0412,
+      "step": 18215
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.250196798055502e-06,
+      "loss": 0.9619,
+      "step": 18216
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.249152329033685e-06,
+      "loss": 0.9957,
+      "step": 18217
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.248107907627566e-06,
+      "loss": 0.9293,
+      "step": 18218
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2470635338504105e-06,
+      "loss": 1.0225,
+      "step": 18219
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.24601920771547e-06,
+      "loss": 1.0055,
+      "step": 18220
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.244974929236003e-06,
+      "loss": 0.9682,
+      "step": 18221
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2439306984252646e-06,
+      "loss": 0.8796,
+      "step": 18222
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.242886515296517e-06,
+      "loss": 0.9177,
+      "step": 18223
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.241842379863012e-06,
+      "loss": 0.9082,
+      "step": 18224
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.240798292138002e-06,
+      "loss": 0.9657,
+      "step": 18225
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.239754252134746e-06,
+      "loss": 0.9771,
+      "step": 18226
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.23871025986649e-06,
+      "loss": 1.0191,
+      "step": 18227
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.237666315346495e-06,
+      "loss": 0.9606,
+      "step": 18228
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.236622418588016e-06,
+      "loss": 0.8487,
+      "step": 18229
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.235578569604297e-06,
+      "loss": 0.9694,
+      "step": 18230
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.234534768408591e-06,
+      "loss": 0.9735,
+      "step": 18231
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.233491015014148e-06,
+      "loss": 0.9073,
+      "step": 18232
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.232447309434224e-06,
+      "loss": 0.912,
+      "step": 18233
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.231403651682064e-06,
+      "loss": 0.9736,
+      "step": 18234
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.230360041770915e-06,
+      "loss": 1.1116,
+      "step": 18235
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.229316479714027e-06,
+      "loss": 0.9439,
+      "step": 18236
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.228272965524653e-06,
+      "loss": 0.9629,
+      "step": 18237
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.227229499216034e-06,
+      "loss": 0.973,
+      "step": 18238
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.226186080801415e-06,
+      "loss": 0.8888,
+      "step": 18239
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2251427102940475e-06,
+      "loss": 1.0238,
+      "step": 18240
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.22409938770717e-06,
+      "loss": 0.9097,
+      "step": 18241
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.223056113054033e-06,
+      "loss": 0.8705,
+      "step": 18242
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.222012886347881e-06,
+      "loss": 0.9706,
+      "step": 18243
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.220969707601954e-06,
+      "loss": 1.0463,
+      "step": 18244
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.219926576829494e-06,
+      "loss": 0.9157,
+      "step": 18245
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.218883494043741e-06,
+      "loss": 1.0142,
+      "step": 18246
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2178404592579464e-06,
+      "loss": 0.8788,
+      "step": 18247
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.216797472485345e-06,
+      "loss": 0.9386,
+      "step": 18248
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.215754533739176e-06,
+      "loss": 0.9377,
+      "step": 18249
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2147116430326785e-06,
+      "loss": 0.9949,
+      "step": 18250
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.213668800379095e-06,
+      "loss": 0.9728,
+      "step": 18251
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.212626005791663e-06,
+      "loss": 0.956,
+      "step": 18252
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.2115832592836225e-06,
+      "loss": 1.0082,
+      "step": 18253
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 6.210540560868206e-06,
+      "loss": 0.9886,
+      "step": 18254
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20949791055865e-06,
+      "loss": 0.9832,
+      "step": 18255
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.208455308368195e-06,
+      "loss": 1.0045,
+      "step": 18256
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.207412754310076e-06,
+      "loss": 1.0776,
+      "step": 18257
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.206370248397526e-06,
+      "loss": 0.9693,
+      "step": 18258
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2053277906437805e-06,
+      "loss": 0.9,
+      "step": 18259
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.2042853810620675e-06,
+      "loss": 0.9431,
+      "step": 18260
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.20324301966563e-06,
+      "loss": 1.1243,
+      "step": 18261
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.202200706467695e-06,
+      "loss": 0.9979,
+      "step": 18262
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.201158441481492e-06,
+      "loss": 1.0533,
+      "step": 18263
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.200116224720254e-06,
+      "loss": 0.9701,
+      "step": 18264
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.199074056197215e-06,
+      "loss": 0.9129,
+      "step": 18265
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.198031935925601e-06,
+      "loss": 1.0415,
+      "step": 18266
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.196989863918644e-06,
+      "loss": 1.0785,
+      "step": 18267
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.195947840189572e-06,
+      "loss": 1.0351,
+      "step": 18268
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.194905864751608e-06,
+      "loss": 0.8598,
+      "step": 18269
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.193863937617988e-06,
+      "loss": 0.9791,
+      "step": 18270
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.192822058801935e-06,
+      "loss": 1.0982,
+      "step": 18271
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.191780228316675e-06,
+      "loss": 0.9454,
+      "step": 18272
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1907384461754305e-06,
+      "loss": 1.0557,
+      "step": 18273
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.189696712391432e-06,
+      "loss": 0.9334,
+      "step": 18274
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.188655026977904e-06,
+      "loss": 0.9793,
+      "step": 18275
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.187613389948066e-06,
+      "loss": 0.8808,
+      "step": 18276
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.186571801315147e-06,
+      "loss": 1.0165,
+      "step": 18277
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.185530261092359e-06,
+      "loss": 0.9124,
+      "step": 18278
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.184488769292937e-06,
+      "loss": 0.9629,
+      "step": 18279
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.183447325930094e-06,
+      "loss": 0.9739,
+      "step": 18280
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.182405931017055e-06,
+      "loss": 0.8021,
+      "step": 18281
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.181364584567039e-06,
+      "loss": 0.947,
+      "step": 18282
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.180323286593259e-06,
+      "loss": 0.8901,
+      "step": 18283
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.179282037108945e-06,
+      "loss": 0.8975,
+      "step": 18284
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17824083612731e-06,
+      "loss": 0.8563,
+      "step": 18285
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.17719968366157e-06,
+      "loss": 1.0165,
+      "step": 18286
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1761585797249425e-06,
+      "loss": 1.0202,
+      "step": 18287
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.175117524330647e-06,
+      "loss": 0.9379,
+      "step": 18288
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.174076517491899e-06,
+      "loss": 0.9562,
+      "step": 18289
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.173035559221911e-06,
+      "loss": 0.9982,
+      "step": 18290
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1719946495339e-06,
+      "loss": 1.0255,
+      "step": 18291
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.170953788441074e-06,
+      "loss": 1.0808,
+      "step": 18292
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.169912975956658e-06,
+      "loss": 1.0558,
+      "step": 18293
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.168872212093854e-06,
+      "loss": 1.0076,
+      "step": 18294
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.167831496865881e-06,
+      "loss": 0.9905,
+      "step": 18295
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.166790830285947e-06,
+      "loss": 0.9955,
+      "step": 18296
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.16575021236726e-06,
+      "loss": 1.0307,
+      "step": 18297
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.164709643123035e-06,
+      "loss": 0.9309,
+      "step": 18298
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.163669122566483e-06,
+      "loss": 0.8768,
+      "step": 18299
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.162628650710809e-06,
+      "loss": 0.9135,
+      "step": 18300
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1615882275692216e-06,
+      "loss": 0.9843,
+      "step": 18301
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.160547853154932e-06,
+      "loss": 1.126,
+      "step": 18302
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.159507527481148e-06,
+      "loss": 0.933,
+      "step": 18303
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.158467250561069e-06,
+      "loss": 0.9474,
+      "step": 18304
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.157427022407909e-06,
+      "loss": 1.0335,
+      "step": 18305
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.156386843034865e-06,
+      "loss": 0.9756,
+      "step": 18306
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.155346712455152e-06,
+      "loss": 0.9806,
+      "step": 18307
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.154306630681965e-06,
+      "loss": 1.0207,
+      "step": 18308
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1532665977285134e-06,
+      "loss": 0.9707,
+      "step": 18309
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.152226613607997e-06,
+      "loss": 1.0668,
+      "step": 18310
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1511866783336165e-06,
+      "loss": 0.8805,
+      "step": 18311
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.150146791918577e-06,
+      "loss": 1.0487,
+      "step": 18312
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.149106954376081e-06,
+      "loss": 1.0712,
+      "step": 18313
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.148067165719324e-06,
+      "loss": 1.006,
+      "step": 18314
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.147027425961505e-06,
+      "loss": 0.8787,
+      "step": 18315
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.14598773511583e-06,
+      "loss": 0.9691,
+      "step": 18316
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.144948093195495e-06,
+      "loss": 1.0875,
+      "step": 18317
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.143908500213693e-06,
+      "loss": 0.9439,
+      "step": 18318
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.142868956183629e-06,
+      "loss": 0.9328,
+      "step": 18319
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.141829461118489e-06,
+      "loss": 0.9382,
+      "step": 18320
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.140790015031481e-06,
+      "loss": 0.8986,
+      "step": 18321
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.139750617935791e-06,
+      "loss": 0.994,
+      "step": 18322
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.138711269844622e-06,
+      "loss": 1.0661,
+      "step": 18323
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.137671970771159e-06,
+      "loss": 0.9647,
+      "step": 18324
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.136632720728602e-06,
+      "loss": 0.9321,
+      "step": 18325
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.135593519730143e-06,
+      "loss": 0.9834,
+      "step": 18326
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.134554367788974e-06,
+      "loss": 0.9624,
+      "step": 18327
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.133515264918285e-06,
+      "loss": 1.0669,
+      "step": 18328
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.132476211131267e-06,
+      "loss": 1.0322,
+      "step": 18329
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1314372064411135e-06,
+      "loss": 0.9438,
+      "step": 18330
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.130398250861014e-06,
+      "loss": 1.0257,
+      "step": 18331
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.129359344404153e-06,
+      "loss": 1.0417,
+      "step": 18332
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1283204870837255e-06,
+      "loss": 1.0146,
+      "step": 18333
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1272816789129105e-06,
+      "loss": 0.9055,
+      "step": 18334
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.126242919904907e-06,
+      "loss": 1.0116,
+      "step": 18335
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.125204210072892e-06,
+      "loss": 0.973,
+      "step": 18336
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.124165549430059e-06,
+      "loss": 1.0572,
+      "step": 18337
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.123126937989585e-06,
+      "loss": 0.8374,
+      "step": 18338
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.122088375764659e-06,
+      "loss": 0.9634,
+      "step": 18339
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.121049862768469e-06,
+      "loss": 0.9184,
+      "step": 18340
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1200113990141965e-06,
+      "loss": 1.0288,
+      "step": 18341
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.118972984515019e-06,
+      "loss": 0.9903,
+      "step": 18342
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.117934619284123e-06,
+      "loss": 0.8305,
+      "step": 18343
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.116896303334693e-06,
+      "loss": 0.9215,
+      "step": 18344
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1158580366799066e-06,
+      "loss": 0.914,
+      "step": 18345
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.114819819332944e-06,
+      "loss": 0.9175,
+      "step": 18346
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.113781651306987e-06,
+      "loss": 0.8498,
+      "step": 18347
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.112743532615211e-06,
+      "loss": 0.931,
+      "step": 18348
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.111705463270796e-06,
+      "loss": 0.9544,
+      "step": 18349
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.110667443286926e-06,
+      "loss": 0.8831,
+      "step": 18350
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.109629472676773e-06,
+      "loss": 0.9349,
+      "step": 18351
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.108591551453511e-06,
+      "loss": 0.9984,
+      "step": 18352
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.107553679630316e-06,
+      "loss": 0.8757,
+      "step": 18353
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.106515857220372e-06,
+      "loss": 1.0017,
+      "step": 18354
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.105478084236849e-06,
+      "loss": 1.0367,
+      "step": 18355
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.104440360692917e-06,
+      "loss": 0.852,
+      "step": 18356
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.103402686601751e-06,
+      "loss": 1.0648,
+      "step": 18357
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.102365061976532e-06,
+      "loss": 0.8694,
+      "step": 18358
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.101327486830422e-06,
+      "loss": 1.0728,
+      "step": 18359
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.1002899611766e-06,
+      "loss": 0.9956,
+      "step": 18360
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.099252485028233e-06,
+      "loss": 0.9355,
+      "step": 18361
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.098215058398488e-06,
+      "loss": 0.9696,
+      "step": 18362
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.097177681300541e-06,
+      "loss": 0.8953,
+      "step": 18363
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.096140353747562e-06,
+      "loss": 0.9663,
+      "step": 18364
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.095103075752717e-06,
+      "loss": 1.073,
+      "step": 18365
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0940658473291715e-06,
+      "loss": 0.9495,
+      "step": 18366
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.093028668490091e-06,
+      "loss": 0.9162,
+      "step": 18367
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.091991539248652e-06,
+      "loss": 1.0252,
+      "step": 18368
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.090954459618013e-06,
+      "loss": 0.9656,
+      "step": 18369
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.08991742961134e-06,
+      "loss": 0.9435,
+      "step": 18370
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.088880449241795e-06,
+      "loss": 0.9414,
+      "step": 18371
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.087843518522552e-06,
+      "loss": 0.8321,
+      "step": 18372
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.086806637466764e-06,
+      "loss": 0.9169,
+      "step": 18373
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0857698060876e-06,
+      "loss": 0.8348,
+      "step": 18374
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0847330243982206e-06,
+      "loss": 1.0274,
+      "step": 18375
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.083696292411783e-06,
+      "loss": 1.0004,
+      "step": 18376
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.082659610141454e-06,
+      "loss": 1.029,
+      "step": 18377
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.081622977600394e-06,
+      "loss": 0.9825,
+      "step": 18378
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.080586394801761e-06,
+      "loss": 0.9631,
+      "step": 18379
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0795498617587115e-06,
+      "loss": 1.1565,
+      "step": 18380
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.078513378484403e-06,
+      "loss": 0.849,
+      "step": 18381
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.077476944992003e-06,
+      "loss": 1.0118,
+      "step": 18382
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.076440561294658e-06,
+      "loss": 0.9892,
+      "step": 18383
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.075404227405531e-06,
+      "loss": 0.8828,
+      "step": 18384
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.074367943337772e-06,
+      "loss": 0.9276,
+      "step": 18385
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.073331709104544e-06,
+      "loss": 0.8706,
+      "step": 18386
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.072295524718995e-06,
+      "loss": 1.0315,
+      "step": 18387
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.071259390194284e-06,
+      "loss": 0.9214,
+      "step": 18388
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0702233055435614e-06,
+      "loss": 0.895,
+      "step": 18389
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.069187270779976e-06,
+      "loss": 1.0485,
+      "step": 18390
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.068151285916687e-06,
+      "loss": 1.0106,
+      "step": 18391
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.067115350966845e-06,
+      "loss": 0.9766,
+      "step": 18392
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.066079465943597e-06,
+      "loss": 0.9727,
+      "step": 18393
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0650436308600955e-06,
+      "loss": 0.8718,
+      "step": 18394
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.064007845729487e-06,
+      "loss": 0.8212,
+      "step": 18395
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.062972110564926e-06,
+      "loss": 0.9848,
+      "step": 18396
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.061936425379558e-06,
+      "loss": 0.9326,
+      "step": 18397
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.060900790186532e-06,
+      "loss": 0.9871,
+      "step": 18398
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.05986520499899e-06,
+      "loss": 0.7601,
+      "step": 18399
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.058829669830084e-06,
+      "loss": 1.0228,
+      "step": 18400
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.057794184692957e-06,
+      "loss": 1.0416,
+      "step": 18401
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.056758749600758e-06,
+      "loss": 1.0871,
+      "step": 18402
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.055723364566626e-06,
+      "loss": 0.9748,
+      "step": 18403
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.054688029603704e-06,
+      "loss": 1.0664,
+      "step": 18404
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0536527447251405e-06,
+      "loss": 1.0632,
+      "step": 18405
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0526175099440785e-06,
+      "loss": 0.9615,
+      "step": 18406
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.051582325273655e-06,
+      "loss": 0.9928,
+      "step": 18407
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.050547190727016e-06,
+      "loss": 0.9801,
+      "step": 18408
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.049512106317297e-06,
+      "loss": 1.0306,
+      "step": 18409
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.048477072057644e-06,
+      "loss": 0.9684,
+      "step": 18410
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.047442087961191e-06,
+      "loss": 1.0106,
+      "step": 18411
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.046407154041082e-06,
+      "loss": 0.9,
+      "step": 18412
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.045372270310448e-06,
+      "loss": 1.037,
+      "step": 18413
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.044337436782434e-06,
+      "loss": 1.1518,
+      "step": 18414
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0433026534701715e-06,
+      "loss": 0.9586,
+      "step": 18415
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.042267920386802e-06,
+      "loss": 0.9327,
+      "step": 18416
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.041233237545454e-06,
+      "loss": 1.0216,
+      "step": 18417
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.040198604959269e-06,
+      "loss": 1.0076,
+      "step": 18418
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0391640226413774e-06,
+      "loss": 0.9396,
+      "step": 18419
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.038129490604915e-06,
+      "loss": 0.8878,
+      "step": 18420
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.037095008863012e-06,
+      "loss": 0.9708,
+      "step": 18421
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.036060577428802e-06,
+      "loss": 0.9954,
+      "step": 18422
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.035026196315421e-06,
+      "loss": 0.9364,
+      "step": 18423
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.033991865535996e-06,
+      "loss": 0.8542,
+      "step": 18424
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.032957585103657e-06,
+      "loss": 0.9016,
+      "step": 18425
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.031923355031536e-06,
+      "loss": 0.8867,
+      "step": 18426
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.030889175332757e-06,
+      "loss": 0.9852,
+      "step": 18427
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.029855046020456e-06,
+      "loss": 1.0214,
+      "step": 18428
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.028820967107758e-06,
+      "loss": 0.9628,
+      "step": 18429
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.027786938607789e-06,
+      "loss": 0.9598,
+      "step": 18430
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.026752960533676e-06,
+      "loss": 1.0618,
+      "step": 18431
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.025719032898543e-06,
+      "loss": 1.0299,
+      "step": 18432
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0246851557155215e-06,
+      "loss": 0.9817,
+      "step": 18433
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.023651328997733e-06,
+      "loss": 0.9754,
+      "step": 18434
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.022617552758298e-06,
+      "loss": 1.023,
+      "step": 18435
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.021583827010341e-06,
+      "loss": 1.0052,
+      "step": 18436
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.020550151766991e-06,
+      "loss": 1.0263,
+      "step": 18437
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.019516527041366e-06,
+      "loss": 0.9743,
+      "step": 18438
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.018482952846585e-06,
+      "loss": 0.8757,
+      "step": 18439
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0174494291957745e-06,
+      "loss": 0.913,
+      "step": 18440
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.016415956102047e-06,
+      "loss": 1.0961,
+      "step": 18441
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.015382533578529e-06,
+      "loss": 0.9535,
+      "step": 18442
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.014349161638335e-06,
+      "loss": 0.8358,
+      "step": 18443
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.013315840294589e-06,
+      "loss": 0.9358,
+      "step": 18444
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.012282569560403e-06,
+      "loss": 0.9746,
+      "step": 18445
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.011249349448892e-06,
+      "loss": 0.9324,
+      "step": 18446
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0102161799731816e-06,
+      "loss": 0.9336,
+      "step": 18447
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.009183061146382e-06,
+      "loss": 1.015,
+      "step": 18448
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.008149992981606e-06,
+      "loss": 0.9232,
+      "step": 18449
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.00711697549197e-06,
+      "loss": 0.7627,
+      "step": 18450
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0060840086905915e-06,
+      "loss": 0.928,
+      "step": 18451
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.0050510925905805e-06,
+      "loss": 0.9529,
+      "step": 18452
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.004018227205046e-06,
+      "loss": 0.9366,
+      "step": 18453
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.002985412547108e-06,
+      "loss": 1.0629,
+      "step": 18454
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.001952648629868e-06,
+      "loss": 0.9138,
+      "step": 18455
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 6.000919935466443e-06,
+      "loss": 1.0569,
+      "step": 18456
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.999887273069944e-06,
+      "loss": 1.0123,
+      "step": 18457
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.998854661453478e-06,
+      "loss": 0.9352,
+      "step": 18458
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9978221006301515e-06,
+      "loss": 0.8926,
+      "step": 18459
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.996789590613073e-06,
+      "loss": 1.1084,
+      "step": 18460
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.995757131415354e-06,
+      "loss": 1.0393,
+      "step": 18461
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.994724723050101e-06,
+      "loss": 1.0001,
+      "step": 18462
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.993692365530412e-06,
+      "loss": 0.9065,
+      "step": 18463
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9926600588694e-06,
+      "loss": 1.0162,
+      "step": 18464
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.991627803080169e-06,
+      "loss": 0.8549,
+      "step": 18465
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.990595598175823e-06,
+      "loss": 0.9492,
+      "step": 18466
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.989563444169463e-06,
+      "loss": 0.8656,
+      "step": 18467
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9885313410741954e-06,
+      "loss": 1.0646,
+      "step": 18468
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.987499288903115e-06,
+      "loss": 1.0742,
+      "step": 18469
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.986467287669331e-06,
+      "loss": 0.9389,
+      "step": 18470
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.985435337385944e-06,
+      "loss": 0.9888,
+      "step": 18471
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.984403438066053e-06,
+      "loss": 0.9153,
+      "step": 18472
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.983371589722753e-06,
+      "loss": 0.9237,
+      "step": 18473
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.982339792369146e-06,
+      "loss": 1.0105,
+      "step": 18474
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9813080460183344e-06,
+      "loss": 0.9448,
+      "step": 18475
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.980276350683413e-06,
+      "loss": 0.9067,
+      "step": 18476
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9792447063774765e-06,
+      "loss": 0.9765,
+      "step": 18477
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.978213113113621e-06,
+      "loss": 1.0555,
+      "step": 18478
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.977181570904949e-06,
+      "loss": 1.0208,
+      "step": 18479
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.976150079764547e-06,
+      "loss": 0.9128,
+      "step": 18480
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.975118639705516e-06,
+      "loss": 0.8364,
+      "step": 18481
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.974087250740947e-06,
+      "loss": 0.9173,
+      "step": 18482
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9730559128839295e-06,
+      "loss": 1.1207,
+      "step": 18483
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9720246261475615e-06,
+      "loss": 0.8803,
+      "step": 18484
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9709933905449345e-06,
+      "loss": 0.9757,
+      "step": 18485
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.969962206089138e-06,
+      "loss": 0.9099,
+      "step": 18486
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.968931072793261e-06,
+      "loss": 0.9774,
+      "step": 18487
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.967899990670394e-06,
+      "loss": 0.9442,
+      "step": 18488
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.966868959733629e-06,
+      "loss": 0.9944,
+      "step": 18489
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.965837979996053e-06,
+      "loss": 0.9291,
+      "step": 18490
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.964807051470755e-06,
+      "loss": 0.9423,
+      "step": 18491
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.963776174170817e-06,
+      "loss": 0.9318,
+      "step": 18492
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.962745348109334e-06,
+      "loss": 1.0914,
+      "step": 18493
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.961714573299384e-06,
+      "loss": 0.9804,
+      "step": 18494
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.96068384975406e-06,
+      "loss": 0.9555,
+      "step": 18495
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.959653177486442e-06,
+      "loss": 0.8861,
+      "step": 18496
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95862255650961e-06,
+      "loss": 0.8708,
+      "step": 18497
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.957591986836655e-06,
+      "loss": 1.1551,
+      "step": 18498
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9565614684806584e-06,
+      "loss": 0.8747,
+      "step": 18499
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.955531001454699e-06,
+      "loss": 1.1226,
+      "step": 18500
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.954500585771861e-06,
+      "loss": 0.9695,
+      "step": 18501
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.95347022144522e-06,
+      "loss": 0.9727,
+      "step": 18502
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.952439908487865e-06,
+      "loss": 0.9157,
+      "step": 18503
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.951409646912868e-06,
+      "loss": 0.9229,
+      "step": 18504
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.950379436733312e-06,
+      "loss": 0.9538,
+      "step": 18505
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.949349277962271e-06,
+      "loss": 1.0535,
+      "step": 18506
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9483191706128285e-06,
+      "loss": 0.8803,
+      "step": 18507
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.947289114698055e-06,
+      "loss": 0.9816,
+      "step": 18508
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.946259110231032e-06,
+      "loss": 0.991,
+      "step": 18509
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.945229157224832e-06,
+      "loss": 0.9812,
+      "step": 18510
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.944199255692528e-06,
+      "loss": 0.9349,
+      "step": 18511
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9431694056471975e-06,
+      "loss": 0.9856,
+      "step": 18512
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.942139607101918e-06,
+      "loss": 0.8539,
+      "step": 18513
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9411098600697535e-06,
+      "loss": 1.0178,
+      "step": 18514
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.940080164563783e-06,
+      "loss": 0.9329,
+      "step": 18515
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9390505205970715e-06,
+      "loss": 0.9576,
+      "step": 18516
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9380209281826985e-06,
+      "loss": 0.9158,
+      "step": 18517
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.936991387333728e-06,
+      "loss": 0.9424,
+      "step": 18518
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.935961898063234e-06,
+      "loss": 0.9821,
+      "step": 18519
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.934932460384279e-06,
+      "loss": 0.9142,
+      "step": 18520
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.933903074309939e-06,
+      "loss": 0.9187,
+      "step": 18521
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.932873739853276e-06,
+      "loss": 0.8922,
+      "step": 18522
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.931844457027363e-06,
+      "loss": 1.0038,
+      "step": 18523
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9308152258452615e-06,
+      "loss": 1.0056,
+      "step": 18524
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.929786046320033e-06,
+      "loss": 1.0323,
+      "step": 18525
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9287569184647506e-06,
+      "loss": 0.8698,
+      "step": 18526
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.927727842292479e-06,
+      "loss": 0.8099,
+      "step": 18527
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.926698817816278e-06,
+      "loss": 1.0108,
+      "step": 18528
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.925669845049211e-06,
+      "loss": 0.9539,
+      "step": 18529
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.924640924004339e-06,
+      "loss": 0.9757,
+      "step": 18530
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92361205469473e-06,
+      "loss": 0.9053,
+      "step": 18531
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.922583237133439e-06,
+      "loss": 0.955,
+      "step": 18532
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.92155447133353e-06,
+      "loss": 0.8692,
+      "step": 18533
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.920525757308058e-06,
+      "loss": 0.915,
+      "step": 18534
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.91949709507009e-06,
+      "loss": 0.9591,
+      "step": 18535
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9184684846326775e-06,
+      "loss": 0.9976,
+      "step": 18536
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.917439926008883e-06,
+      "loss": 1.0086,
+      "step": 18537
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.91641141921176e-06,
+      "loss": 0.9995,
+      "step": 18538
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9153829642543654e-06,
+      "loss": 0.9445,
+      "step": 18539
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.9143545611497574e-06,
+      "loss": 0.9194,
+      "step": 18540
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.913326209910992e-06,
+      "loss": 1.127,
+      "step": 18541
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.91229791055112e-06,
+      "loss": 0.9444,
+      "step": 18542
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.911269663083199e-06,
+      "loss": 0.9306,
+      "step": 18543
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.910241467520277e-06,
+      "loss": 1.0578,
+      "step": 18544
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.909213323875413e-06,
+      "loss": 0.8548,
+      "step": 18545
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.908185232161655e-06,
+      "loss": 0.9702,
+      "step": 18546
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9071571923920565e-06,
+      "loss": 1.041,
+      "step": 18547
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.906129204579663e-06,
+      "loss": 0.9667,
+      "step": 18548
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.905101268737533e-06,
+      "loss": 0.9256,
+      "step": 18549
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.904073384878708e-06,
+      "loss": 1.0346,
+      "step": 18550
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.903045553016242e-06,
+      "loss": 0.9402,
+      "step": 18551
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9020177731631805e-06,
+      "loss": 0.8541,
+      "step": 18552
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.9009900453325664e-06,
+      "loss": 0.9689,
+      "step": 18553
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.899962369537456e-06,
+      "loss": 1.0509,
+      "step": 18554
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8989347457908904e-06,
+      "loss": 1.0253,
+      "step": 18555
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.897907174105912e-06,
+      "loss": 0.982,
+      "step": 18556
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.896879654495567e-06,
+      "loss": 0.9742,
+      "step": 18557
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.895852186972904e-06,
+      "loss": 0.96,
+      "step": 18558
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8948247715509645e-06,
+      "loss": 1.058,
+      "step": 18559
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.893797408242785e-06,
+      "loss": 0.8126,
+      "step": 18560
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.892770097061416e-06,
+      "loss": 1.0049,
+      "step": 18561
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8917428380198905e-06,
+      "loss": 1.0468,
+      "step": 18562
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.890715631131256e-06,
+      "loss": 1.0629,
+      "step": 18563
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.889688476408552e-06,
+      "loss": 0.9027,
+      "step": 18564
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.888661373864816e-06,
+      "loss": 0.9521,
+      "step": 18565
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.887634323513085e-06,
+      "loss": 1.0153,
+      "step": 18566
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.886607325366396e-06,
+      "loss": 0.8436,
+      "step": 18567
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.885580379437794e-06,
+      "loss": 0.8639,
+      "step": 18568
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8845534857403094e-06,
+      "loss": 0.9333,
+      "step": 18569
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.883526644286981e-06,
+      "loss": 1.0444,
+      "step": 18570
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8824998550908375e-06,
+      "loss": 0.9576,
+      "step": 18571
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.881473118164925e-06,
+      "loss": 0.9009,
+      "step": 18572
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.880446433522272e-06,
+      "loss": 0.9401,
+      "step": 18573
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.879419801175909e-06,
+      "loss": 0.9278,
+      "step": 18574
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.878393221138874e-06,
+      "loss": 0.9399,
+      "step": 18575
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.877366693424192e-06,
+      "loss": 0.9899,
+      "step": 18576
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8763402180449e-06,
+      "loss": 1.1397,
+      "step": 18577
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.875313795014031e-06,
+      "loss": 0.9121,
+      "step": 18578
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.874287424344609e-06,
+      "loss": 0.888,
+      "step": 18579
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.873261106049667e-06,
+      "loss": 1.0094,
+      "step": 18580
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.872234840142229e-06,
+      "loss": 1.1017,
+      "step": 18581
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.871208626635332e-06,
+      "loss": 0.963,
+      "step": 18582
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.870182465541998e-06,
+      "loss": 0.8673,
+      "step": 18583
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.869156356875252e-06,
+      "loss": 0.9743,
+      "step": 18584
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.868130300648121e-06,
+      "loss": 0.8397,
+      "step": 18585
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.867104296873634e-06,
+      "loss": 1.1211,
+      "step": 18586
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.866078345564812e-06,
+      "loss": 0.9989,
+      "step": 18587
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.865052446734682e-06,
+      "loss": 0.8933,
+      "step": 18588
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.864026600396267e-06,
+      "loss": 0.9087,
+      "step": 18589
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.863000806562584e-06,
+      "loss": 0.9594,
+      "step": 18590
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8619750652466615e-06,
+      "loss": 0.897,
+      "step": 18591
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8609493764615225e-06,
+      "loss": 0.8889,
+      "step": 18592
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8599237402201825e-06,
+      "loss": 0.983,
+      "step": 18593
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8588981565356615e-06,
+      "loss": 1.0256,
+      "step": 18594
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.857872625420981e-06,
+      "loss": 0.9763,
+      "step": 18595
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.856847146889162e-06,
+      "loss": 0.8526,
+      "step": 18596
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.855821720953221e-06,
+      "loss": 1.1382,
+      "step": 18597
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.854796347626172e-06,
+      "loss": 0.9601,
+      "step": 18598
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8537710269210335e-06,
+      "loss": 1.0823,
+      "step": 18599
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.852745758850826e-06,
+      "loss": 0.9479,
+      "step": 18600
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.851720543428559e-06,
+      "loss": 1.0349,
+      "step": 18601
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.850695380667253e-06,
+      "loss": 0.8821,
+      "step": 18602
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.849670270579917e-06,
+      "loss": 1.0892,
+      "step": 18603
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.848645213179565e-06,
+      "loss": 1.1514,
+      "step": 18604
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.84762020847921e-06,
+      "loss": 1.0342,
+      "step": 18605
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8465952564918684e-06,
+      "loss": 0.839,
+      "step": 18606
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8455703572305475e-06,
+      "loss": 1.0173,
+      "step": 18607
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8445455107082574e-06,
+      "loss": 0.8765,
+      "step": 18608
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.843520716938008e-06,
+      "loss": 0.9759,
+      "step": 18609
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8424959759328135e-06,
+      "loss": 0.9225,
+      "step": 18610
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.841471287705678e-06,
+      "loss": 0.9617,
+      "step": 18611
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.840446652269613e-06,
+      "loss": 1.0477,
+      "step": 18612
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.839422069637619e-06,
+      "loss": 0.9791,
+      "step": 18613
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.838397539822711e-06,
+      "loss": 0.8332,
+      "step": 18614
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.837373062837892e-06,
+      "loss": 1.1021,
+      "step": 18615
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.836348638696168e-06,
+      "loss": 0.9544,
+      "step": 18616
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.835324267410542e-06,
+      "loss": 1.0454,
+      "step": 18617
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.834299948994016e-06,
+      "loss": 0.9473,
+      "step": 18618
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.833275683459598e-06,
+      "loss": 0.9262,
+      "step": 18619
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.832251470820286e-06,
+      "loss": 0.885,
+      "step": 18620
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.831227311089088e-06,
+      "loss": 0.9174,
+      "step": 18621
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.830203204279003e-06,
+      "loss": 1.0295,
+      "step": 18622
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.829179150403027e-06,
+      "loss": 1.0091,
+      "step": 18623
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8281551494741676e-06,
+      "loss": 0.9892,
+      "step": 18624
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.827131201505421e-06,
+      "loss": 0.9559,
+      "step": 18625
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.826107306509783e-06,
+      "loss": 0.9522,
+      "step": 18626
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.825083464500252e-06,
+      "loss": 1.0117,
+      "step": 18627
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.824059675489826e-06,
+      "loss": 0.8225,
+      "step": 18628
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.823035939491508e-06,
+      "loss": 1.0586,
+      "step": 18629
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.822012256518288e-06,
+      "loss": 1.1279,
+      "step": 18630
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.820988626583162e-06,
+      "loss": 0.8855,
+      "step": 18631
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.81996504969912e-06,
+      "loss": 0.8595,
+      "step": 18632
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.818941525879165e-06,
+      "loss": 0.9548,
+      "step": 18633
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.817918055136282e-06,
+      "loss": 0.9607,
+      "step": 18634
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.816894637483472e-06,
+      "loss": 0.9885,
+      "step": 18635
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.815871272933722e-06,
+      "loss": 0.9365,
+      "step": 18636
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.814847961500021e-06,
+      "loss": 1.0002,
+      "step": 18637
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.813824703195364e-06,
+      "loss": 0.9212,
+      "step": 18638
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.812801498032741e-06,
+      "loss": 0.9745,
+      "step": 18639
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.811778346025134e-06,
+      "loss": 0.9148,
+      "step": 18640
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.810755247185538e-06,
+      "loss": 1.0179,
+      "step": 18641
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.809732201526943e-06,
+      "loss": 0.9315,
+      "step": 18642
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.808709209062333e-06,
+      "loss": 0.8351,
+      "step": 18643
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.807686269804695e-06,
+      "loss": 0.9466,
+      "step": 18644
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.806663383767015e-06,
+      "loss": 1.0315,
+      "step": 18645
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8056405509622725e-06,
+      "loss": 0.8599,
+      "step": 18646
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.804617771403463e-06,
+      "loss": 1.0202,
+      "step": 18647
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.803595045103559e-06,
+      "loss": 0.8955,
+      "step": 18648
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.802572372075553e-06,
+      "loss": 0.9891,
+      "step": 18649
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.8015497523324235e-06,
+      "loss": 1.0082,
+      "step": 18650
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.800527185887149e-06,
+      "loss": 0.8642,
+      "step": 18651
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7995046727527185e-06,
+      "loss": 0.956,
+      "step": 18652
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.798482212942107e-06,
+      "loss": 1.0302,
+      "step": 18653
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.797459806468291e-06,
+      "loss": 0.9948,
+      "step": 18654
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.796437453344255e-06,
+      "loss": 0.9797,
+      "step": 18655
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7954151535829796e-06,
+      "loss": 1.0233,
+      "step": 18656
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.794392907197439e-06,
+      "loss": 0.864,
+      "step": 18657
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79337071420061e-06,
+      "loss": 0.9506,
+      "step": 18658
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79234857460547e-06,
+      "loss": 0.9866,
+      "step": 18659
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.79132648842499e-06,
+      "loss": 1.1299,
+      "step": 18660
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.790304455672153e-06,
+      "loss": 0.9535,
+      "step": 18661
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.789282476359926e-06,
+      "loss": 1.0108,
+      "step": 18662
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.78826055050129e-06,
+      "loss": 1.0362,
+      "step": 18663
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.787238678109214e-06,
+      "loss": 0.9853,
+      "step": 18664
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.786216859196666e-06,
+      "loss": 0.9509,
+      "step": 18665
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.785195093776627e-06,
+      "loss": 0.9122,
+      "step": 18666
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.784173381862061e-06,
+      "loss": 0.9572,
+      "step": 18667
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.783151723465939e-06,
+      "loss": 1.0195,
+      "step": 18668
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.782130118601231e-06,
+      "loss": 0.9316,
+      "step": 18669
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.781108567280911e-06,
+      "loss": 0.8556,
+      "step": 18670
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.780087069517943e-06,
+      "loss": 1.0507,
+      "step": 18671
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.779065625325294e-06,
+      "loss": 1.0392,
+      "step": 18672
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7780442347159315e-06,
+      "loss": 1.0591,
+      "step": 18673
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.777022897702818e-06,
+      "loss": 0.9765,
+      "step": 18674
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.776001614298928e-06,
+      "loss": 0.9191,
+      "step": 18675
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.774980384517215e-06,
+      "loss": 0.994,
+      "step": 18676
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.773959208370654e-06,
+      "loss": 0.899,
+      "step": 18677
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.772938085872204e-06,
+      "loss": 1.063,
+      "step": 18678
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.771917017034824e-06,
+      "loss": 0.9973,
+      "step": 18679
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.770896001871483e-06,
+      "loss": 0.9704,
+      "step": 18680
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.769875040395139e-06,
+      "loss": 1.0728,
+      "step": 18681
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.768854132618748e-06,
+      "loss": 1.0422,
+      "step": 18682
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.767833278555277e-06,
+      "loss": 0.9487,
+      "step": 18683
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.766812478217685e-06,
+      "loss": 0.96,
+      "step": 18684
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.765791731618931e-06,
+      "loss": 0.9352,
+      "step": 18685
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.764771038771968e-06,
+      "loss": 0.8347,
+      "step": 18686
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.763750399689758e-06,
+      "loss": 1.016,
+      "step": 18687
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7627298143852524e-06,
+      "loss": 0.9743,
+      "step": 18688
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.76170928287141e-06,
+      "loss": 1.0786,
+      "step": 18689
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.760688805161192e-06,
+      "loss": 0.8531,
+      "step": 18690
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.759668381267547e-06,
+      "loss": 0.9172,
+      "step": 18691
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.758648011203429e-06,
+      "loss": 0.8709,
+      "step": 18692
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.757627694981789e-06,
+      "loss": 0.9496,
+      "step": 18693
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.756607432615586e-06,
+      "loss": 0.9633,
+      "step": 18694
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.755587224117769e-06,
+      "loss": 0.9934,
+      "step": 18695
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.754567069501285e-06,
+      "loss": 1.029,
+      "step": 18696
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.753546968779087e-06,
+      "loss": 0.9423,
+      "step": 18697
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.75252692196413e-06,
+      "loss": 0.98,
+      "step": 18698
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.751506929069359e-06,
+      "loss": 0.9026,
+      "step": 18699
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.750486990107725e-06,
+      "loss": 1.1022,
+      "step": 18700
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.74946710509217e-06,
+      "loss": 0.8901,
+      "step": 18701
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.748447274035642e-06,
+      "loss": 0.8643,
+      "step": 18702
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.747427496951089e-06,
+      "loss": 0.8488,
+      "step": 18703
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.746407773851461e-06,
+      "loss": 0.8995,
+      "step": 18704
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7453881047497005e-06,
+      "loss": 0.9757,
+      "step": 18705
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.744368489658747e-06,
+      "loss": 0.9037,
+      "step": 18706
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.743348928591552e-06,
+      "loss": 1.0974,
+      "step": 18707
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.742329421561055e-06,
+      "loss": 1.0084,
+      "step": 18708
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.741309968580197e-06,
+      "loss": 0.9204,
+      "step": 18709
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.740290569661917e-06,
+      "loss": 1.0419,
+      "step": 18710
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.739271224819159e-06,
+      "loss": 0.9974,
+      "step": 18711
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.738251934064867e-06,
+      "loss": 0.8535,
+      "step": 18712
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.737232697411976e-06,
+      "loss": 0.9107,
+      "step": 18713
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.736213514873428e-06,
+      "loss": 1.0566,
+      "step": 18714
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7351943864621585e-06,
+      "loss": 0.8917,
+      "step": 18715
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.734175312191101e-06,
+      "loss": 0.8542,
+      "step": 18716
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.733156292073199e-06,
+      "loss": 0.9814,
+      "step": 18717
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.732137326121387e-06,
+      "loss": 0.981,
+      "step": 18718
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.731118414348602e-06,
+      "loss": 1.002,
+      "step": 18719
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.730099556767772e-06,
+      "loss": 1.0265,
+      "step": 18720
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.72908075339184e-06,
+      "loss": 0.9828,
+      "step": 18721
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7280620042337354e-06,
+      "loss": 0.9525,
+      "step": 18722
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.727043309306386e-06,
+      "loss": 0.9526,
+      "step": 18723
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.726024668622732e-06,
+      "loss": 1.0506,
+      "step": 18724
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.725006082195697e-06,
+      "loss": 0.9664,
+      "step": 18725
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7239875500382195e-06,
+      "loss": 0.9531,
+      "step": 18726
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.722969072163225e-06,
+      "loss": 0.9168,
+      "step": 18727
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.721950648583644e-06,
+      "loss": 0.984,
+      "step": 18728
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.720932279312403e-06,
+      "loss": 0.97,
+      "step": 18729
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.719913964362429e-06,
+      "loss": 0.9729,
+      "step": 18730
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71889570374665e-06,
+      "loss": 0.9071,
+      "step": 18731
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7178774974779976e-06,
+      "loss": 1.0366,
+      "step": 18732
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.716859345569395e-06,
+      "loss": 1.0026,
+      "step": 18733
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.71584124803376e-06,
+      "loss": 0.9799,
+      "step": 18734
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.714823204884027e-06,
+      "loss": 1.0327,
+      "step": 18735
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.713805216133116e-06,
+      "loss": 0.953,
+      "step": 18736
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.712787281793946e-06,
+      "loss": 0.9499,
+      "step": 18737
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.711769401879446e-06,
+      "loss": 0.9516,
+      "step": 18738
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.710751576402531e-06,
+      "loss": 1.009,
+      "step": 18739
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7097338053761295e-06,
+      "loss": 0.9123,
+      "step": 18740
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7087160888131576e-06,
+      "loss": 0.9423,
+      "step": 18741
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.707698426726535e-06,
+      "loss": 0.9691,
+      "step": 18742
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.706680819129183e-06,
+      "loss": 1.0421,
+      "step": 18743
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.70566326603401e-06,
+      "loss": 1.1273,
+      "step": 18744
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.704645767453942e-06,
+      "loss": 1.0224,
+      "step": 18745
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.703628323401897e-06,
+      "loss": 0.9074,
+      "step": 18746
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.702610933890791e-06,
+      "loss": 1.0822,
+      "step": 18747
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.701593598933532e-06,
+      "loss": 0.9775,
+      "step": 18748
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.7005763185430455e-06,
+      "loss": 0.9767,
+      "step": 18749
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.699559092732239e-06,
+      "loss": 0.936,
+      "step": 18750
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.69854192151402e-06,
+      "loss": 1.0147,
+      "step": 18751
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.697524804901316e-06,
+      "loss": 0.9961,
+      "step": 18752
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.696507742907024e-06,
+      "loss": 0.9339,
+      "step": 18753
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.695490735544067e-06,
+      "loss": 1.003,
+      "step": 18754
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.694473782825349e-06,
+      "loss": 0.9449,
+      "step": 18755
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.693456884763783e-06,
+      "loss": 0.9009,
+      "step": 18756
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.692440041372276e-06,
+      "loss": 0.985,
+      "step": 18757
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.691423252663733e-06,
+      "loss": 1.0021,
+      "step": 18758
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.690406518651066e-06,
+      "loss": 0.8071,
+      "step": 18759
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6893898393471855e-06,
+      "loss": 0.8984,
+      "step": 18760
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.688373214764995e-06,
+      "loss": 1.0072,
+      "step": 18761
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.687356644917394e-06,
+      "loss": 0.8892,
+      "step": 18762
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.686340129817299e-06,
+      "loss": 0.9257,
+      "step": 18763
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.685323669477607e-06,
+      "loss": 0.9296,
+      "step": 18764
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.68430726391122e-06,
+      "loss": 0.8806,
+      "step": 18765
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.683290913131047e-06,
+      "loss": 0.9472,
+      "step": 18766
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.682274617149983e-06,
+      "loss": 0.91,
+      "step": 18767
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.681258375980937e-06,
+      "loss": 0.9929,
+      "step": 18768
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.680242189636807e-06,
+      "loss": 0.9527,
+      "step": 18769
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.679226058130492e-06,
+      "loss": 0.9702,
+      "step": 18770
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.678209981474889e-06,
+      "loss": 0.9503,
+      "step": 18771
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6771939596828985e-06,
+      "loss": 0.9151,
+      "step": 18772
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.676177992767424e-06,
+      "loss": 0.9187,
+      "step": 18773
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.675162080741359e-06,
+      "loss": 0.916,
+      "step": 18774
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.674146223617599e-06,
+      "loss": 0.9882,
+      "step": 18775
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.673130421409037e-06,
+      "loss": 0.8653,
+      "step": 18776
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.672114674128577e-06,
+      "loss": 1.0639,
+      "step": 18777
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.671098981789107e-06,
+      "loss": 0.9322,
+      "step": 18778
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.670083344403519e-06,
+      "loss": 0.9758,
+      "step": 18779
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.669067761984712e-06,
+      "loss": 0.9743,
+      "step": 18780
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.668052234545573e-06,
+      "loss": 0.9407,
+      "step": 18781
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6670367620990015e-06,
+      "loss": 0.9754,
+      "step": 18782
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.666021344657884e-06,
+      "loss": 0.9388,
+      "step": 18783
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.665005982235109e-06,
+      "loss": 0.95,
+      "step": 18784
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6639906748435645e-06,
+      "loss": 0.9393,
+      "step": 18785
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6629754224961434e-06,
+      "loss": 0.9798,
+      "step": 18786
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.661960225205736e-06,
+      "loss": 1.0236,
+      "step": 18787
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.660945082985228e-06,
+      "loss": 0.9896,
+      "step": 18788
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.659929995847506e-06,
+      "loss": 1.0597,
+      "step": 18789
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.658914963805451e-06,
+      "loss": 0.9769,
+      "step": 18790
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.657899986871957e-06,
+      "loss": 0.9449,
+      "step": 18791
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.656885065059908e-06,
+      "loss": 0.9553,
+      "step": 18792
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.655870198382176e-06,
+      "loss": 0.8852,
+      "step": 18793
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.654855386851661e-06,
+      "loss": 0.9498,
+      "step": 18794
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6538406304812335e-06,
+      "loss": 0.8856,
+      "step": 18795
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.652825929283783e-06,
+      "loss": 0.9304,
+      "step": 18796
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.651811283272188e-06,
+      "loss": 0.9752,
+      "step": 18797
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.650796692459329e-06,
+      "loss": 0.8949,
+      "step": 18798
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.649782156858081e-06,
+      "loss": 1.0388,
+      "step": 18799
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.648767676481331e-06,
+      "loss": 0.8887,
+      "step": 18800
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.647753251341955e-06,
+      "loss": 0.9916,
+      "step": 18801
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.646738881452831e-06,
+      "loss": 0.9119,
+      "step": 18802
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.645724566826837e-06,
+      "loss": 0.7648,
+      "step": 18803
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.644710307476842e-06,
+      "loss": 0.9467,
+      "step": 18804
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.643696103415732e-06,
+      "loss": 0.9675,
+      "step": 18805
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6426819546563774e-06,
+      "loss": 0.9119,
+      "step": 18806
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.641667861211648e-06,
+      "loss": 1.0235,
+      "step": 18807
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.640653823094426e-06,
+      "loss": 0.89,
+      "step": 18808
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6396398403175745e-06,
+      "loss": 1.0049,
+      "step": 18809
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.638625912893977e-06,
+      "loss": 1.0066,
+      "step": 18810
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.637612040836499e-06,
+      "loss": 0.944,
+      "step": 18811
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.63659822415801e-06,
+      "loss": 0.9294,
+      "step": 18812
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.635584462871379e-06,
+      "loss": 0.9131,
+      "step": 18813
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.634570756989477e-06,
+      "loss": 0.9166,
+      "step": 18814
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.633557106525176e-06,
+      "loss": 0.9708,
+      "step": 18815
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.632543511491343e-06,
+      "loss": 1.0247,
+      "step": 18816
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.631529971900843e-06,
+      "loss": 1.1209,
+      "step": 18817
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.630516487766539e-06,
+      "loss": 1.0576,
+      "step": 18818
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6295030591013045e-06,
+      "loss": 0.9406,
+      "step": 18819
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.628489685917998e-06,
+      "loss": 0.8936,
+      "step": 18820
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62747636822949e-06,
+      "loss": 0.9565,
+      "step": 18821
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.626463106048642e-06,
+      "loss": 0.8905,
+      "step": 18822
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6254498993883115e-06,
+      "loss": 0.963,
+      "step": 18823
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.6244367482613704e-06,
+      "loss": 0.9198,
+      "step": 18824
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.623423652680675e-06,
+      "loss": 1.0891,
+      "step": 18825
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.622410612659086e-06,
+      "loss": 0.9673,
+      "step": 18826
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.62139762820946e-06,
+      "loss": 1.0267,
+      "step": 18827
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 5.620384699344661e-06,
+      "loss": 0.9002,
+      "step": 18828
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.619371826077551e-06,
+      "loss": 0.9901,
+      "step": 18829
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.618359008420986e-06,
+      "loss": 0.9981,
+      "step": 18830
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.617346246387821e-06,
+      "loss": 0.8769,
+      "step": 18831
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.616333539990908e-06,
+      "loss": 0.9023,
+      "step": 18832
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.615320889243115e-06,
+      "loss": 1.0396,
+      "step": 18833
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.614308294157284e-06,
+      "loss": 0.9254,
+      "step": 18834
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6132957547462815e-06,
+      "loss": 1.038,
+      "step": 18835
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.612283271022957e-06,
+      "loss": 0.862,
+      "step": 18836
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6112708430001586e-06,
+      "loss": 0.8094,
+      "step": 18837
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.610258470690746e-06,
+      "loss": 1.0212,
+      "step": 18838
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.609246154107567e-06,
+      "loss": 0.8938,
+      "step": 18839
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.608233893263475e-06,
+      "loss": 1.0006,
+      "step": 18840
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.6072216881713134e-06,
+      "loss": 0.959,
+      "step": 18841
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.606209538843937e-06,
+      "loss": 0.9253,
+      "step": 18842
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.605197445294199e-06,
+      "loss": 1.0791,
+      "step": 18843
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.604185407534944e-06,
+      "loss": 1.1444,
+      "step": 18844
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.603173425579019e-06,
+      "loss": 0.9158,
+      "step": 18845
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.602161499439265e-06,
+      "loss": 0.9896,
+      "step": 18846
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.60114962912854e-06,
+      "loss": 1.0418,
+      "step": 18847
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.600137814659676e-06,
+      "loss": 0.9522,
+      "step": 18848
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.59912605604553e-06,
+      "loss": 0.9287,
+      "step": 18849
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5981143532989405e-06,
+      "loss": 1.0084,
+      "step": 18850
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.597102706432746e-06,
+      "loss": 1.0211,
+      "step": 18851
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.596091115459799e-06,
+      "loss": 1.0898,
+      "step": 18852
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.595079580392934e-06,
+      "loss": 1.0115,
+      "step": 18853
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.594068101244991e-06,
+      "loss": 1.1438,
+      "step": 18854
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.593056678028817e-06,
+      "loss": 1.0921,
+      "step": 18855
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.592045310757246e-06,
+      "loss": 0.818,
+      "step": 18856
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.591033999443121e-06,
+      "loss": 0.7975,
+      "step": 18857
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.590022744099279e-06,
+      "loss": 1.0111,
+      "step": 18858
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5890115447385565e-06,
+      "loss": 0.9353,
+      "step": 18859
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.588000401373788e-06,
+      "loss": 0.895,
+      "step": 18860
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.586989314017815e-06,
+      "loss": 0.9578,
+      "step": 18861
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.585978282683465e-06,
+      "loss": 1.0185,
+      "step": 18862
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.584967307383583e-06,
+      "loss": 0.8198,
+      "step": 18863
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.583956388130998e-06,
+      "loss": 0.9466,
+      "step": 18864
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.582945524938539e-06,
+      "loss": 0.8625,
+      "step": 18865
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5819347178190465e-06,
+      "loss": 0.9117,
+      "step": 18866
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.580923966785349e-06,
+      "loss": 1.0038,
+      "step": 18867
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.579913271850272e-06,
+      "loss": 0.9055,
+      "step": 18868
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5789026330266525e-06,
+      "loss": 0.9843,
+      "step": 18869
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5778920503273225e-06,
+      "loss": 0.8489,
+      "step": 18870
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.576881523765106e-06,
+      "loss": 0.9749,
+      "step": 18871
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5758710533528335e-06,
+      "loss": 0.8796,
+      "step": 18872
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.574860639103332e-06,
+      "loss": 0.9903,
+      "step": 18873
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.573850281029422e-06,
+      "loss": 1.0037,
+      "step": 18874
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.572839979143941e-06,
+      "loss": 0.9376,
+      "step": 18875
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.571829733459706e-06,
+      "loss": 0.9902,
+      "step": 18876
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5708195439895475e-06,
+      "loss": 0.9286,
+      "step": 18877
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5698094107462865e-06,
+      "loss": 0.9419,
+      "step": 18878
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.568799333742743e-06,
+      "loss": 0.9424,
+      "step": 18879
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.567789312991747e-06,
+      "loss": 1.0929,
+      "step": 18880
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5667793485061164e-06,
+      "loss": 1.0425,
+      "step": 18881
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5657694402986675e-06,
+      "loss": 0.8637,
+      "step": 18882
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.564759588382227e-06,
+      "loss": 1.0127,
+      "step": 18883
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5637497927696175e-06,
+      "loss": 1.077,
+      "step": 18884
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.562740053473652e-06,
+      "loss": 0.9098,
+      "step": 18885
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.561730370507152e-06,
+      "loss": 0.9153,
+      "step": 18886
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.560720743882932e-06,
+      "loss": 1.0118,
+      "step": 18887
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5597111736138084e-06,
+      "loss": 0.9263,
+      "step": 18888
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.558701659712603e-06,
+      "loss": 0.9456,
+      "step": 18889
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.557692202192123e-06,
+      "loss": 1.0361,
+      "step": 18890
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.556682801065192e-06,
+      "loss": 0.9047,
+      "step": 18891
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55567345634462e-06,
+      "loss": 0.9552,
+      "step": 18892
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.554664168043216e-06,
+      "loss": 0.9861,
+      "step": 18893
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5536549361738e-06,
+      "loss": 0.8826,
+      "step": 18894
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.55264576074918e-06,
+      "loss": 0.9329,
+      "step": 18895
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.551636641782165e-06,
+      "loss": 0.8736,
+      "step": 18896
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.550627579285568e-06,
+      "loss": 0.8055,
+      "step": 18897
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.549618573272202e-06,
+      "loss": 1.0633,
+      "step": 18898
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.548609623754873e-06,
+      "loss": 0.8154,
+      "step": 18899
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.547600730746388e-06,
+      "loss": 0.9981,
+      "step": 18900
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.546591894259556e-06,
+      "loss": 0.9277,
+      "step": 18901
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5455831143071785e-06,
+      "loss": 0.9552,
+      "step": 18902
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.544574390902068e-06,
+      "loss": 1.024,
+      "step": 18903
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54356572405703e-06,
+      "loss": 0.9345,
+      "step": 18904
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.54255711378487e-06,
+      "loss": 1.0685,
+      "step": 18905
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.541548560098389e-06,
+      "loss": 1.0246,
+      "step": 18906
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5405400630103845e-06,
+      "loss": 1.0609,
+      "step": 18907
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.539531622533671e-06,
+      "loss": 0.9372,
+      "step": 18908
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5385232386810435e-06,
+      "loss": 0.9507,
+      "step": 18909
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5375149114653e-06,
+      "loss": 0.9868,
+      "step": 18910
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.536506640899246e-06,
+      "loss": 0.9697,
+      "step": 18911
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5354984269956845e-06,
+      "loss": 1.0821,
+      "step": 18912
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.534490269767407e-06,
+      "loss": 0.9525,
+      "step": 18913
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.533482169227218e-06,
+      "loss": 0.8612,
+      "step": 18914
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.53247412538791e-06,
+      "loss": 0.9018,
+      "step": 18915
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.531466138262278e-06,
+      "loss": 0.9985,
+      "step": 18916
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.530458207863121e-06,
+      "loss": 1.0882,
+      "step": 18917
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.529450334203239e-06,
+      "loss": 1.0108,
+      "step": 18918
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5284425172954225e-06,
+      "loss": 0.887,
+      "step": 18919
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.527434757152465e-06,
+      "loss": 1.0106,
+      "step": 18920
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.526427053787155e-06,
+      "loss": 0.953,
+      "step": 18921
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5254194072122955e-06,
+      "loss": 1.0055,
+      "step": 18922
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.524411817440671e-06,
+      "loss": 1.0487,
+      "step": 18923
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.523404284485072e-06,
+      "loss": 0.9419,
+      "step": 18924
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.52239680835829e-06,
+      "loss": 0.9386,
+      "step": 18925
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.52138938907312e-06,
+      "loss": 0.8804,
+      "step": 18926
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.520382026642344e-06,
+      "loss": 1.041,
+      "step": 18927
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5193747210787556e-06,
+      "loss": 0.9294,
+      "step": 18928
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.518367472395137e-06,
+      "loss": 1.0162,
+      "step": 18929
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.517360280604274e-06,
+      "loss": 1.0213,
+      "step": 18930
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.516353145718953e-06,
+      "loss": 1.0804,
+      "step": 18931
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.515346067751968e-06,
+      "loss": 0.8941,
+      "step": 18932
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.514339046716096e-06,
+      "loss": 0.9393,
+      "step": 18933
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.513332082624123e-06,
+      "loss": 0.9377,
+      "step": 18934
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.512325175488826e-06,
+      "loss": 1.1261,
+      "step": 18935
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.511318325322996e-06,
+      "loss": 0.9144,
+      "step": 18936
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.510311532139412e-06,
+      "loss": 0.9426,
+      "step": 18937
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.509304795950849e-06,
+      "loss": 0.927,
+      "step": 18938
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5082981167700925e-06,
+      "loss": 1.0102,
+      "step": 18939
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.507291494609924e-06,
+      "loss": 0.967,
+      "step": 18940
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.506284929483121e-06,
+      "loss": 0.9205,
+      "step": 18941
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.50527842140246e-06,
+      "loss": 0.9463,
+      "step": 18942
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.504271970380718e-06,
+      "loss": 0.9617,
+      "step": 18943
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.503265576430667e-06,
+      "loss": 0.9304,
+      "step": 18944
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.5022592395650894e-06,
+      "loss": 0.9412,
+      "step": 18945
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.501252959796763e-06,
+      "loss": 1.0085,
+      "step": 18946
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.500246737138457e-06,
+      "loss": 1.0827,
+      "step": 18947
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.499240571602945e-06,
+      "loss": 1.0221,
+      "step": 18948
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4982344632029986e-06,
+      "loss": 0.8814,
+      "step": 18949
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4972284119513945e-06,
+      "loss": 0.9586,
+      "step": 18950
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.496222417860897e-06,
+      "loss": 0.996,
+      "step": 18951
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4952164809442876e-06,
+      "loss": 0.9847,
+      "step": 18952
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.494210601214326e-06,
+      "loss": 1.0018,
+      "step": 18953
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.49320477868379e-06,
+      "loss": 0.8996,
+      "step": 18954
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4921990133654424e-06,
+      "loss": 1.1206,
+      "step": 18955
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.491193305272055e-06,
+      "loss": 0.9754,
+      "step": 18956
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.490187654416391e-06,
+      "loss": 0.9057,
+      "step": 18957
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.489182060811216e-06,
+      "loss": 1.036,
+      "step": 18958
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.488176524469296e-06,
+      "loss": 1.0568,
+      "step": 18959
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.487171045403403e-06,
+      "loss": 0.9137,
+      "step": 18960
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.486165623626296e-06,
+      "loss": 0.9753,
+      "step": 18961
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.485160259150739e-06,
+      "loss": 0.9057,
+      "step": 18962
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.484154951989491e-06,
+      "loss": 1.0722,
+      "step": 18963
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.483149702155319e-06,
+      "loss": 1.1132,
+      "step": 18964
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.482144509660981e-06,
+      "loss": 0.7867,
+      "step": 18965
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.481139374519243e-06,
+      "loss": 0.9733,
+      "step": 18966
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.480134296742857e-06,
+      "loss": 0.9996,
+      "step": 18967
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4791292763445905e-06,
+      "loss": 0.8773,
+      "step": 18968
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.478124313337197e-06,
+      "loss": 0.9173,
+      "step": 18969
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.477119407733434e-06,
+      "loss": 0.9304,
+      "step": 18970
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.476114559546061e-06,
+      "loss": 0.9876,
+      "step": 18971
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.475109768787827e-06,
+      "loss": 1.0188,
+      "step": 18972
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.474105035471493e-06,
+      "loss": 0.972,
+      "step": 18973
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.473100359609818e-06,
+      "loss": 1.0501,
+      "step": 18974
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.472095741215553e-06,
+      "loss": 1.0478,
+      "step": 18975
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.471091180301449e-06,
+      "loss": 0.8911,
+      "step": 18976
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.470086676880254e-06,
+      "loss": 1.0426,
+      "step": 18977
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.469082230964731e-06,
+      "loss": 0.946,
+      "step": 18978
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.468077842567621e-06,
+      "loss": 1.0713,
+      "step": 18979
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.467073511701682e-06,
+      "loss": 0.9603,
+      "step": 18980
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.466069238379657e-06,
+      "loss": 0.9468,
+      "step": 18981
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.465065022614301e-06,
+      "loss": 0.9602,
+      "step": 18982
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.464060864418361e-06,
+      "loss": 0.9616,
+      "step": 18983
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4630567638045816e-06,
+      "loss": 0.9332,
+      "step": 18984
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.462052720785708e-06,
+      "loss": 0.904,
+      "step": 18985
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.461048735374492e-06,
+      "loss": 0.8263,
+      "step": 18986
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4600448075836724e-06,
+      "loss": 0.976,
+      "step": 18987
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.459040937426e-06,
+      "loss": 0.9542,
+      "step": 18988
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4580371249142176e-06,
+      "loss": 1.0356,
+      "step": 18989
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45703337006106e-06,
+      "loss": 0.9469,
+      "step": 18990
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.456029672879282e-06,
+      "loss": 0.992,
+      "step": 18991
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.455026033381617e-06,
+      "loss": 0.9508,
+      "step": 18992
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.454022451580805e-06,
+      "loss": 0.925,
+      "step": 18993
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4530189274895925e-06,
+      "loss": 0.8959,
+      "step": 18994
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.452015461120712e-06,
+      "loss": 0.9534,
+      "step": 18995
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.45101205248691e-06,
+      "loss": 1.0278,
+      "step": 18996
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4500087016009195e-06,
+      "loss": 0.9246,
+      "step": 18997
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.449005408475478e-06,
+      "loss": 0.938,
+      "step": 18998
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4480021731233165e-06,
+      "loss": 0.9479,
+      "step": 18999
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.446998995557178e-06,
+      "loss": 1.028,
+      "step": 19000
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4459958757898e-06,
+      "loss": 0.9056,
+      "step": 19001
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.444992813833913e-06,
+      "loss": 1.0102,
+      "step": 19002
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.443989809702248e-06,
+      "loss": 0.9399,
+      "step": 19003
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.442986863407538e-06,
+      "loss": 0.838,
+      "step": 19004
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.441983974962519e-06,
+      "loss": 1.0181,
+      "step": 19005
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.440981144379922e-06,
+      "loss": 0.8526,
+      "step": 19006
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.439978371672472e-06,
+      "loss": 1.0564,
+      "step": 19007
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.438975656852905e-06,
+      "loss": 0.9228,
+      "step": 19008
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.437972999933945e-06,
+      "loss": 1.0603,
+      "step": 19009
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.436970400928326e-06,
+      "loss": 0.8661,
+      "step": 19010
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.435967859848774e-06,
+      "loss": 0.972,
+      "step": 19011
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.434965376708015e-06,
+      "loss": 0.8858,
+      "step": 19012
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4339629515187695e-06,
+      "loss": 0.9495,
+      "step": 19013
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.432960584293769e-06,
+      "loss": 0.8723,
+      "step": 19014
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.431958275045741e-06,
+      "loss": 0.992,
+      "step": 19015
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.430956023787408e-06,
+      "loss": 1.0731,
+      "step": 19016
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.42995383053149e-06,
+      "loss": 0.9498,
+      "step": 19017
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.428951695290705e-06,
+      "loss": 0.9087,
+      "step": 19018
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.427949618077787e-06,
+      "loss": 0.7897,
+      "step": 19019
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.42694759890545e-06,
+      "loss": 0.8359,
+      "step": 19020
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.425945637786409e-06,
+      "loss": 0.9198,
+      "step": 19021
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.424943734733396e-06,
+      "loss": 0.8211,
+      "step": 19022
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4239418897591165e-06,
+      "loss": 1.0673,
+      "step": 19023
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4229401028763015e-06,
+      "loss": 0.9291,
+      "step": 19024
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.4219383740976606e-06,
+      "loss": 0.952,
+      "step": 19025
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.420936703435913e-06,
+      "loss": 0.9174,
+      "step": 19026
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.419935090903768e-06,
+      "loss": 0.9195,
+      "step": 19027
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.418933536513946e-06,
+      "loss": 0.9147,
+      "step": 19028
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.417932040279167e-06,
+      "loss": 1.0792,
+      "step": 19029
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.416930602212138e-06,
+      "loss": 1.0359,
+      "step": 19030
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.415929222325574e-06,
+      "loss": 1.0766,
+      "step": 19031
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.414927900632182e-06,
+      "loss": 0.9381,
+      "step": 19032
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.413926637144682e-06,
+      "loss": 1.016,
+      "step": 19033
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.412925431875775e-06,
+      "loss": 0.9027,
+      "step": 19034
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.411924284838182e-06,
+      "loss": 0.9589,
+      "step": 19035
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.410923196044604e-06,
+      "loss": 0.8698,
+      "step": 19036
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.409922165507751e-06,
+      "loss": 0.9595,
+      "step": 19037
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.408921193240333e-06,
+      "loss": 1.0204,
+      "step": 19038
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.407920279255057e-06,
+      "loss": 1.0906,
+      "step": 19039
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.406919423564627e-06,
+      "loss": 0.9117,
+      "step": 19040
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.405918626181748e-06,
+      "loss": 0.8552,
+      "step": 19041
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.404917887119123e-06,
+      "loss": 0.8718,
+      "step": 19042
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.403917206389467e-06,
+      "loss": 0.946,
+      "step": 19043
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.402916584005472e-06,
+      "loss": 1.0064,
+      "step": 19044
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.401916019979845e-06,
+      "loss": 0.8414,
+      "step": 19045
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.400915514325285e-06,
+      "loss": 0.9499,
+      "step": 19046
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.399915067054497e-06,
+      "loss": 0.9961,
+      "step": 19047
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.398914678180176e-06,
+      "loss": 0.9536,
+      "step": 19048
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.397914347715027e-06,
+      "loss": 1.0371,
+      "step": 19049
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39691407567175e-06,
+      "loss": 0.8522,
+      "step": 19050
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.395913862063034e-06,
+      "loss": 0.9511,
+      "step": 19051
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.394913706901586e-06,
+      "loss": 0.9553,
+      "step": 19052
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3939136102001e-06,
+      "loss": 0.9927,
+      "step": 19053
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.39291357197127e-06,
+      "loss": 0.9056,
+      "step": 19054
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.391913592227788e-06,
+      "loss": 0.9533,
+      "step": 19055
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.390913670982352e-06,
+      "loss": 1.0046,
+      "step": 19056
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.38991380824766e-06,
+      "loss": 0.8817,
+      "step": 19057
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.388914004036401e-06,
+      "loss": 0.8394,
+      "step": 19058
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.387914258361267e-06,
+      "loss": 0.9868,
+      "step": 19059
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3869145712349446e-06,
+      "loss": 0.969,
+      "step": 19060
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.385914942670134e-06,
+      "loss": 0.9944,
+      "step": 19061
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.384915372679518e-06,
+      "loss": 0.8373,
+      "step": 19062
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.383915861275791e-06,
+      "loss": 0.8745,
+      "step": 19063
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.382916408471636e-06,
+      "loss": 0.9489,
+      "step": 19064
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.381917014279743e-06,
+      "loss": 0.9449,
+      "step": 19065
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3809176787128015e-06,
+      "loss": 0.9119,
+      "step": 19066
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.379918401783496e-06,
+      "loss": 0.8775,
+      "step": 19067
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3789191835045116e-06,
+      "loss": 0.9215,
+      "step": 19068
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.377920023888528e-06,
+      "loss": 0.9186,
+      "step": 19069
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.376920922948235e-06,
+      "loss": 0.9558,
+      "step": 19070
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.37592188069632e-06,
+      "loss": 1.0332,
+      "step": 19071
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.374922897145458e-06,
+      "loss": 1.0162,
+      "step": 19072
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.373923972308335e-06,
+      "loss": 1.0524,
+      "step": 19073
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.372925106197626e-06,
+      "loss": 0.8183,
+      "step": 19074
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.371926298826019e-06,
+      "loss": 0.9321,
+      "step": 19075
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.370927550206186e-06,
+      "loss": 0.9439,
+      "step": 19076
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.369928860350815e-06,
+      "loss": 1.0457,
+      "step": 19077
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.368930229272577e-06,
+      "loss": 0.9419,
+      "step": 19078
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.367931656984147e-06,
+      "loss": 1.1488,
+      "step": 19079
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.366933143498209e-06,
+      "loss": 0.9948,
+      "step": 19080
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.365934688827438e-06,
+      "loss": 0.9772,
+      "step": 19081
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3649362929845005e-06,
+      "loss": 1.0083,
+      "step": 19082
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3639379559820815e-06,
+      "loss": 0.8214,
+      "step": 19083
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.362939677832843e-06,
+      "loss": 1.0216,
+      "step": 19084
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.361941458549471e-06,
+      "loss": 0.8442,
+      "step": 19085
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.360943298144632e-06,
+      "loss": 1.0549,
+      "step": 19086
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.359945196630993e-06,
+      "loss": 0.9666,
+      "step": 19087
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.358947154021227e-06,
+      "loss": 0.9524,
+      "step": 19088
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.357949170328007e-06,
+      "loss": 0.9614,
+      "step": 19089
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3569512455639975e-06,
+      "loss": 1.0108,
+      "step": 19090
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.355953379741871e-06,
+      "loss": 0.8917,
+      "step": 19091
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.354955572874293e-06,
+      "loss": 1.0119,
+      "step": 19092
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.353957824973928e-06,
+      "loss": 1.0162,
+      "step": 19093
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3529601360534466e-06,
+      "loss": 1.0372,
+      "step": 19094
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.351962506125513e-06,
+      "loss": 0.997,
+      "step": 19095
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.350964935202787e-06,
+      "loss": 0.9246,
+      "step": 19096
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.349967423297939e-06,
+      "loss": 1.0432,
+      "step": 19097
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.348969970423627e-06,
+      "loss": 0.9655,
+      "step": 19098
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.347972576592517e-06,
+      "loss": 0.9325,
+      "step": 19099
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.346975241817272e-06,
+      "loss": 0.9919,
+      "step": 19100
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.345977966110548e-06,
+      "loss": 0.8908,
+      "step": 19101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3449807494850035e-06,
+      "loss": 0.8215,
+      "step": 19102
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.343983591953305e-06,
+      "loss": 0.9408,
+      "step": 19103
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.342986493528104e-06,
+      "loss": 1.0322,
+      "step": 19104
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.341989454222065e-06,
+      "loss": 1.0056,
+      "step": 19105
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.340992474047841e-06,
+      "loss": 0.8939,
+      "step": 19106
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.339995553018087e-06,
+      "loss": 1.0659,
+      "step": 19107
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.338998691145464e-06,
+      "loss": 0.993,
+      "step": 19108
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3380018884426235e-06,
+      "loss": 1.0096,
+      "step": 19109
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.337005144922214e-06,
+      "loss": 0.9531,
+      "step": 19110
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3360084605969e-06,
+      "loss": 1.1215,
+      "step": 19111
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.335011835479324e-06,
+      "loss": 1.0149,
+      "step": 19112
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.334015269582146e-06,
+      "loss": 1.0441,
+      "step": 19113
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.333018762918012e-06,
+      "loss": 1.0767,
+      "step": 19114
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.3320223154995755e-06,
+      "loss": 0.9057,
+      "step": 19115
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.331025927339479e-06,
+      "loss": 0.8738,
+      "step": 19116
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.33002959845038e-06,
+      "loss": 0.9535,
+      "step": 19117
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.32903332884492e-06,
+      "loss": 0.8492,
+      "step": 19118
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.328037118535752e-06,
+      "loss": 0.9758,
+      "step": 19119
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.327040967535521e-06,
+      "loss": 0.9776,
+      "step": 19120
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.326044875856868e-06,
+      "loss": 1.0696,
+      "step": 19121
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.325048843512444e-06,
+      "loss": 0.8886,
+      "step": 19122
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.324052870514891e-06,
+      "loss": 0.9974,
+      "step": 19123
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.3230569568768486e-06,
+      "loss": 0.9744,
+      "step": 19124
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.322061102610965e-06,
+      "loss": 0.9082,
+      "step": 19125
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.321065307729883e-06,
+      "loss": 0.8736,
+      "step": 19126
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.320069572246241e-06,
+      "loss": 0.7837,
+      "step": 19127
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.319073896172682e-06,
+      "loss": 0.9316,
+      "step": 19128
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.318078279521841e-06,
+      "loss": 1.0085,
+      "step": 19129
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.317082722306357e-06,
+      "loss": 0.8904,
+      "step": 19130
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.316087224538871e-06,
+      "loss": 0.8873,
+      "step": 19131
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.315091786232025e-06,
+      "loss": 1.0059,
+      "step": 19132
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.314096407398451e-06,
+      "loss": 0.8896,
+      "step": 19133
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.313101088050785e-06,
+      "loss": 1.007,
+      "step": 19134
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.312105828201659e-06,
+      "loss": 0.8876,
+      "step": 19135
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.311110627863714e-06,
+      "loss": 0.8192,
+      "step": 19136
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.31011548704958e-06,
+      "loss": 1.0584,
+      "step": 19137
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.309120405771888e-06,
+      "loss": 0.9606,
+      "step": 19138
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.308125384043273e-06,
+      "loss": 0.9838,
+      "step": 19139
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.30713042187637e-06,
+      "loss": 0.898,
+      "step": 19140
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.306135519283806e-06,
+      "loss": 0.8749,
+      "step": 19141
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.305140676278211e-06,
+      "loss": 0.9695,
+      "step": 19142
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.304145892872213e-06,
+      "loss": 0.9693,
+      "step": 19143
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.303151169078438e-06,
+      "loss": 0.8876,
+      "step": 19144
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.302156504909518e-06,
+      "loss": 1.0748,
+      "step": 19145
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.301161900378082e-06,
+      "loss": 0.9979,
+      "step": 19146
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.300167355496755e-06,
+      "loss": 1.0621,
+      "step": 19147
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.299172870278159e-06,
+      "loss": 0.8616,
+      "step": 19148
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.298178444734916e-06,
+      "loss": 0.8931,
+      "step": 19149
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.297184078879661e-06,
+      "loss": 0.9462,
+      "step": 19150
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.296189772725007e-06,
+      "loss": 1.0198,
+      "step": 19151
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.295195526283576e-06,
+      "loss": 0.9686,
+      "step": 19152
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.294201339567995e-06,
+      "loss": 1.0811,
+      "step": 19153
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.293207212590886e-06,
+      "loss": 0.9849,
+      "step": 19154
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.292213145364866e-06,
+      "loss": 0.9525,
+      "step": 19155
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2912191379025565e-06,
+      "loss": 1.0624,
+      "step": 19156
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.290225190216573e-06,
+      "loss": 0.9485,
+      "step": 19157
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.28923130231953e-06,
+      "loss": 0.939,
+      "step": 19158
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.288237474224049e-06,
+      "loss": 1.0851,
+      "step": 19159
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.287243705942751e-06,
+      "loss": 0.917,
+      "step": 19160
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.286249997488246e-06,
+      "loss": 0.9677,
+      "step": 19161
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.285256348873151e-06,
+      "loss": 1.0237,
+      "step": 19162
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.284262760110073e-06,
+      "loss": 1.0746,
+      "step": 19163
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.283269231211636e-06,
+      "loss": 1.0837,
+      "step": 19164
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.282275762190443e-06,
+      "loss": 0.9828,
+      "step": 19165
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.281282353059115e-06,
+      "loss": 0.9149,
+      "step": 19166
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.280289003830253e-06,
+      "loss": 0.8952,
+      "step": 19167
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2792957145164766e-06,
+      "loss": 0.9455,
+      "step": 19168
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2783024851303935e-06,
+      "loss": 0.921,
+      "step": 19169
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.277309315684607e-06,
+      "loss": 0.9786,
+      "step": 19170
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.27631620619173e-06,
+      "loss": 1.0792,
+      "step": 19171
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.275323156664362e-06,
+      "loss": 0.8826,
+      "step": 19172
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.274330167115115e-06,
+      "loss": 0.9795,
+      "step": 19173
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.273337237556599e-06,
+      "loss": 0.9097,
+      "step": 19174
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.272344368001415e-06,
+      "loss": 0.8772,
+      "step": 19175
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2713515584621665e-06,
+      "loss": 1.0771,
+      "step": 19176
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.270358808951454e-06,
+      "loss": 0.9005,
+      "step": 19177
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.269366119481887e-06,
+      "loss": 1.0306,
+      "step": 19178
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.268373490066059e-06,
+      "loss": 1.0404,
+      "step": 19179
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.26738092071658e-06,
+      "loss": 0.8708,
+      "step": 19180
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.266388411446042e-06,
+      "loss": 1.0353,
+      "step": 19181
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2653959622670525e-06,
+      "loss": 1.0407,
+      "step": 19182
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2644035731922046e-06,
+      "loss": 1.0423,
+      "step": 19183
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2634112442341004e-06,
+      "loss": 0.9298,
+      "step": 19184
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.262418975405335e-06,
+      "loss": 0.9663,
+      "step": 19185
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.261426766718499e-06,
+      "loss": 0.9525,
+      "step": 19186
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2604346181861944e-06,
+      "loss": 0.9726,
+      "step": 19187
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.25944252982102e-06,
+      "loss": 0.9506,
+      "step": 19188
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.258450501635566e-06,
+      "loss": 0.8962,
+      "step": 19189
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.257458533642426e-06,
+      "loss": 0.9771,
+      "step": 19190
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.256466625854188e-06,
+      "loss": 0.9509,
+      "step": 19191
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.255474778283452e-06,
+      "loss": 0.9211,
+      "step": 19192
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2544829909428e-06,
+      "loss": 0.9825,
+      "step": 19193
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.253491263844832e-06,
+      "loss": 0.9074,
+      "step": 19194
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2524995970021305e-06,
+      "loss": 0.9305,
+      "step": 19195
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.251507990427292e-06,
+      "loss": 1.0002,
+      "step": 19196
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.250516444132897e-06,
+      "loss": 0.9934,
+      "step": 19197
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.249524958131538e-06,
+      "loss": 0.9762,
+      "step": 19198
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2485335324357975e-06,
+      "loss": 1.0186,
+      "step": 19199
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.247542167058259e-06,
+      "loss": 1.0081,
+      "step": 19200
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.246550862011513e-06,
+      "loss": 0.9065,
+      "step": 19201
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.245559617308146e-06,
+      "loss": 0.9908,
+      "step": 19202
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.244568432960737e-06,
+      "loss": 1.0025,
+      "step": 19203
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.243577308981871e-06,
+      "loss": 0.9129,
+      "step": 19204
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.242586245384122e-06,
+      "loss": 1.0242,
+      "step": 19205
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2415952421800845e-06,
+      "loss": 0.9492,
+      "step": 19206
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.240604299382326e-06,
+      "loss": 0.9056,
+      "step": 19207
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.239613417003438e-06,
+      "loss": 1.0242,
+      "step": 19208
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.238622595055989e-06,
+      "loss": 0.9547,
+      "step": 19209
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2376318335525656e-06,
+      "loss": 1.0742,
+      "step": 19210
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.236641132505742e-06,
+      "loss": 1.0427,
+      "step": 19211
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.235650491928093e-06,
+      "loss": 1.0331,
+      "step": 19212
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.234659911832194e-06,
+      "loss": 0.8622,
+      "step": 19213
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.233669392230624e-06,
+      "loss": 0.97,
+      "step": 19214
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.232678933135951e-06,
+      "loss": 0.9015,
+      "step": 19215
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.231688534560757e-06,
+      "loss": 0.9165,
+      "step": 19216
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2306981965176095e-06,
+      "loss": 0.9085,
+      "step": 19217
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2297079190190825e-06,
+      "loss": 1.0994,
+      "step": 19218
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.228717702077742e-06,
+      "loss": 1.014,
+      "step": 19219
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.227727545706167e-06,
+      "loss": 0.9613,
+      "step": 19220
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.226737449916919e-06,
+      "loss": 1.0126,
+      "step": 19221
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2257474147225725e-06,
+      "loss": 1.0053,
+      "step": 19222
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.224757440135691e-06,
+      "loss": 0.9041,
+      "step": 19223
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.223767526168849e-06,
+      "loss": 0.9676,
+      "step": 19224
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.222777672834609e-06,
+      "loss": 0.9703,
+      "step": 19225
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.221787880145536e-06,
+      "loss": 0.95,
+      "step": 19226
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2207981481141924e-06,
+      "loss": 0.9002,
+      "step": 19227
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.219808476753145e-06,
+      "loss": 0.8982,
+      "step": 19228
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.218818866074962e-06,
+      "loss": 0.8327,
+      "step": 19229
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.217829316092202e-06,
+      "loss": 0.8389,
+      "step": 19230
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.216839826817428e-06,
+      "loss": 1.07,
+      "step": 19231
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2158503982632e-06,
+      "loss": 0.8825,
+      "step": 19232
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.214861030442075e-06,
+      "loss": 0.9535,
+      "step": 19233
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.2138717233666215e-06,
+      "loss": 1.0011,
+      "step": 19234
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.21288247704939e-06,
+      "loss": 0.8692,
+      "step": 19235
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.211893291502945e-06,
+      "loss": 0.8872,
+      "step": 19236
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.210904166739837e-06,
+      "loss": 0.9994,
+      "step": 19237
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.209915102772631e-06,
+      "loss": 0.7911,
+      "step": 19238
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.208926099613879e-06,
+      "loss": 0.9784,
+      "step": 19239
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.207937157276135e-06,
+      "loss": 1.1182,
+      "step": 19240
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.20694827577195e-06,
+      "loss": 0.9547,
+      "step": 19241
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.20595945511388e-06,
+      "loss": 0.9421,
+      "step": 19242
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.204970695314484e-06,
+      "loss": 1.0714,
+      "step": 19243
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.203981996386309e-06,
+      "loss": 0.9764,
+      "step": 19244
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202993358341906e-06,
+      "loss": 0.9271,
+      "step": 19245
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.202004781193827e-06,
+      "loss": 0.9773,
+      "step": 19246
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.201016264954615e-06,
+      "loss": 1.1115,
+      "step": 19247
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.200027809636827e-06,
+      "loss": 0.8974,
+      "step": 19248
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.199039415253006e-06,
+      "loss": 1.0312,
+      "step": 19249
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.198051081815705e-06,
+      "loss": 0.9787,
+      "step": 19250
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.197062809337463e-06,
+      "loss": 0.9355,
+      "step": 19251
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.196074597830834e-06,
+      "loss": 0.9415,
+      "step": 19252
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.195086447308358e-06,
+      "loss": 0.9027,
+      "step": 19253
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.194098357782581e-06,
+      "loss": 0.9494,
+      "step": 19254
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.193110329266041e-06,
+      "loss": 1.0571,
+      "step": 19255
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.192122361771286e-06,
+      "loss": 1.0161,
+      "step": 19256
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.19113445531086e-06,
+      "loss": 1.0359,
+      "step": 19257
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.190146609897302e-06,
+      "loss": 1.0054,
+      "step": 19258
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.18915882554315e-06,
+      "loss": 0.9998,
+      "step": 19259
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.188171102260946e-06,
+      "loss": 0.9277,
+      "step": 19260
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.187183440063226e-06,
+      "loss": 0.9023,
+      "step": 19261
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.186195838962527e-06,
+      "loss": 0.9352,
+      "step": 19262
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1852082989713945e-06,
+      "loss": 0.9893,
+      "step": 19263
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1842208201023605e-06,
+      "loss": 1.0592,
+      "step": 19264
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.183233402367955e-06,
+      "loss": 1.0259,
+      "step": 19265
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1822460457807215e-06,
+      "loss": 0.9932,
+      "step": 19266
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.181258750353191e-06,
+      "loss": 0.9922,
+      "step": 19267
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1802715160978965e-06,
+      "loss": 0.9335,
+      "step": 19268
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.179284343027366e-06,
+      "loss": 0.9684,
+      "step": 19269
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.178297231154136e-06,
+      "loss": 0.9217,
+      "step": 19270
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.177310180490741e-06,
+      "loss": 0.8549,
+      "step": 19271
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.176323191049708e-06,
+      "loss": 0.9369,
+      "step": 19272
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.175336262843566e-06,
+      "loss": 1.0183,
+      "step": 19273
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.174349395884839e-06,
+      "loss": 0.8389,
+      "step": 19274
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.173362590186064e-06,
+      "loss": 0.8628,
+      "step": 19275
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.172375845759761e-06,
+      "loss": 1.0985,
+      "step": 19276
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.171389162618461e-06,
+      "loss": 0.9964,
+      "step": 19277
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1704025407746895e-06,
+      "loss": 0.8701,
+      "step": 19278
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.169415980240965e-06,
+      "loss": 0.9415,
+      "step": 19279
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.16842948102982e-06,
+      "loss": 0.9735,
+      "step": 19280
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.167443043153775e-06,
+      "loss": 0.8945,
+      "step": 19281
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.166456666625351e-06,
+      "loss": 0.9915,
+      "step": 19282
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1654703514570645e-06,
+      "loss": 1.0291,
+      "step": 19283
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.164484097661444e-06,
+      "loss": 0.9936,
+      "step": 19284
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1634979052510116e-06,
+      "loss": 1.0279,
+      "step": 19285
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.162511774238281e-06,
+      "loss": 0.9655,
+      "step": 19286
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.161525704635775e-06,
+      "loss": 1.0343,
+      "step": 19287
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.160539696456003e-06,
+      "loss": 0.9399,
+      "step": 19288
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.159553749711492e-06,
+      "loss": 0.9139,
+      "step": 19289
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.158567864414752e-06,
+      "loss": 0.9452,
+      "step": 19290
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.157582040578303e-06,
+      "loss": 0.9028,
+      "step": 19291
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.156596278214659e-06,
+      "loss": 1.0717,
+      "step": 19292
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.155610577336326e-06,
+      "loss": 0.8354,
+      "step": 19293
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.154624937955829e-06,
+      "loss": 0.9374,
+      "step": 19294
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.153639360085674e-06,
+      "loss": 1.0154,
+      "step": 19295
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.152653843738371e-06,
+      "loss": 0.9981,
+      "step": 19296
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.151668388926435e-06,
+      "loss": 0.936,
+      "step": 19297
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.150682995662371e-06,
+      "loss": 0.9327,
+      "step": 19298
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.149697663958696e-06,
+      "loss": 0.9044,
+      "step": 19299
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.148712393827913e-06,
+      "loss": 0.9945,
+      "step": 19300
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.147727185282532e-06,
+      "loss": 0.8571,
+      "step": 19301
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.146742038335052e-06,
+      "loss": 1.1172,
+      "step": 19302
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.145756952997991e-06,
+      "loss": 1.0018,
+      "step": 19303
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.144771929283844e-06,
+      "loss": 0.9891,
+      "step": 19304
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.143786967205125e-06,
+      "loss": 0.8582,
+      "step": 19305
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.142802066774332e-06,
+      "loss": 0.9711,
+      "step": 19306
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.141817228003965e-06,
+      "loss": 0.9687,
+      "step": 19307
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.140832450906534e-06,
+      "loss": 0.9603,
+      "step": 19308
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.139847735494539e-06,
+      "loss": 0.9712,
+      "step": 19309
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.138863081780472e-06,
+      "loss": 0.9608,
+      "step": 19310
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.137878489776844e-06,
+      "loss": 0.8957,
+      "step": 19311
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.136893959496144e-06,
+      "loss": 1.0514,
+      "step": 19312
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.13590949095088e-06,
+      "loss": 1.099,
+      "step": 19313
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.134925084153545e-06,
+      "loss": 1.0018,
+      "step": 19314
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.133940739116635e-06,
+      "loss": 1.1154,
+      "step": 19315
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.132956455852642e-06,
+      "loss": 0.9149,
+      "step": 19316
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.131972234374072e-06,
+      "loss": 0.9559,
+      "step": 19317
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.130988074693406e-06,
+      "loss": 0.9158,
+      "step": 19318
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.13000397682315e-06,
+      "loss": 1.0828,
+      "step": 19319
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.129019940775793e-06,
+      "loss": 0.8094,
+      "step": 19320
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.128035966563818e-06,
+      "loss": 0.9215,
+      "step": 19321
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.127052054199728e-06,
+      "loss": 1.0196,
+      "step": 19322
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.126068203696011e-06,
+      "loss": 1.0149,
+      "step": 19323
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.125084415065149e-06,
+      "loss": 0.8645,
+      "step": 19324
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1241006883196395e-06,
+      "loss": 0.9922,
+      "step": 19325
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1231170234719665e-06,
+      "loss": 0.8432,
+      "step": 19326
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.122133420534619e-06,
+      "loss": 0.8499,
+      "step": 19327
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.121149879520084e-06,
+      "loss": 0.9733,
+      "step": 19328
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1201664004408466e-06,
+      "loss": 0.9798,
+      "step": 19329
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.119182983309386e-06,
+      "loss": 1.0039,
+      "step": 19330
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.118199628138196e-06,
+      "loss": 1.0922,
+      "step": 19331
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.117216334939752e-06,
+      "loss": 0.8792,
+      "step": 19332
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.116233103726542e-06,
+      "loss": 1.0602,
+      "step": 19333
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.115249934511047e-06,
+      "loss": 0.9478,
+      "step": 19334
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.114266827305743e-06,
+      "loss": 0.935,
+      "step": 19335
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1132837821231176e-06,
+      "loss": 0.9597,
+      "step": 19336
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.112300798975647e-06,
+      "loss": 0.9953,
+      "step": 19337
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.111317877875803e-06,
+      "loss": 0.911,
+      "step": 19338
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.110335018836076e-06,
+      "loss": 0.8685,
+      "step": 19339
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.109352221868933e-06,
+      "loss": 0.9023,
+      "step": 19340
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1083694869868575e-06,
+      "loss": 0.9498,
+      "step": 19341
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.107386814202321e-06,
+      "loss": 0.9792,
+      "step": 19342
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.1064042035278e-06,
+      "loss": 0.9609,
+      "step": 19343
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.105421654975762e-06,
+      "loss": 0.8161,
+      "step": 19344
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.104439168558685e-06,
+      "loss": 0.9538,
+      "step": 19345
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.103456744289046e-06,
+      "loss": 0.9192,
+      "step": 19346
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.102474382179312e-06,
+      "loss": 1.0385,
+      "step": 19347
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.101492082241954e-06,
+      "loss": 0.9357,
+      "step": 19348
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.100509844489439e-06,
+      "loss": 1.0087,
+      "step": 19349
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.099527668934241e-06,
+      "loss": 0.9999,
+      "step": 19350
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.098545555588827e-06,
+      "loss": 0.9072,
+      "step": 19351
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0975635044656605e-06,
+      "loss": 0.9259,
+      "step": 19352
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0965815155772144e-06,
+      "loss": 0.8937,
+      "step": 19353
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.095599588935949e-06,
+      "loss": 0.9443,
+      "step": 19354
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.094617724554337e-06,
+      "loss": 0.9071,
+      "step": 19355
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0936359224448375e-06,
+      "loss": 0.9405,
+      "step": 19356
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.092654182619916e-06,
+      "loss": 0.8787,
+      "step": 19357
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.091672505092029e-06,
+      "loss": 0.9418,
+      "step": 19358
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.090690889873646e-06,
+      "loss": 0.9731,
+      "step": 19359
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.089709336977228e-06,
+      "loss": 1.0184,
+      "step": 19360
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0887278464152355e-06,
+      "loss": 1.0398,
+      "step": 19361
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.087746418200124e-06,
+      "loss": 1.0388,
+      "step": 19362
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.086765052344354e-06,
+      "loss": 0.9185,
+      "step": 19363
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.085783748860384e-06,
+      "loss": 1.0669,
+      "step": 19364
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.084802507760675e-06,
+      "loss": 0.9747,
+      "step": 19365
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0838213290576754e-06,
+      "loss": 0.7638,
+      "step": 19366
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.082840212763849e-06,
+      "loss": 1.0147,
+      "step": 19367
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.081859158891644e-06,
+      "loss": 0.8687,
+      "step": 19368
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.080878167453521e-06,
+      "loss": 0.9467,
+      "step": 19369
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.079897238461931e-06,
+      "loss": 0.956,
+      "step": 19370
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0789163719293256e-06,
+      "loss": 1.0293,
+      "step": 19371
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.077935567868153e-06,
+      "loss": 1.0474,
+      "step": 19372
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.076954826290868e-06,
+      "loss": 0.9932,
+      "step": 19373
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.075974147209924e-06,
+      "loss": 0.8906,
+      "step": 19374
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074993530637768e-06,
+      "loss": 0.9971,
+      "step": 19375
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.074012976586847e-06,
+      "loss": 1.0013,
+      "step": 19376
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.073032485069607e-06,
+      "loss": 0.9842,
+      "step": 19377
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.072052056098499e-06,
+      "loss": 0.9545,
+      "step": 19378
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.071071689685969e-06,
+      "loss": 0.988,
+      "step": 19379
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.070091385844457e-06,
+      "loss": 0.9673,
+      "step": 19380
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0691111445864145e-06,
+      "loss": 1.017,
+      "step": 19381
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.068130965924281e-06,
+      "loss": 1.0372,
+      "step": 19382
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.067150849870503e-06,
+      "loss": 0.9676,
+      "step": 19383
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.066170796437521e-06,
+      "loss": 0.8671,
+      "step": 19384
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.065190805637777e-06,
+      "loss": 0.879,
+      "step": 19385
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.064210877483705e-06,
+      "loss": 0.9262,
+      "step": 19386
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.063231011987752e-06,
+      "loss": 1.0358,
+      "step": 19387
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.06225120916236e-06,
+      "loss": 0.8549,
+      "step": 19388
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.061271469019962e-06,
+      "loss": 1.0522,
+      "step": 19389
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0602917915729975e-06,
+      "loss": 0.8609,
+      "step": 19390
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.059312176833897e-06,
+      "loss": 1.0521,
+      "step": 19391
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.058332624815107e-06,
+      "loss": 0.9835,
+      "step": 19392
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.057353135529052e-06,
+      "loss": 1.0068,
+      "step": 19393
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.056373708988176e-06,
+      "loss": 1.099,
+      "step": 19394
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.055394345204907e-06,
+      "loss": 1.0021,
+      "step": 19395
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.054415044191674e-06,
+      "loss": 0.9583,
+      "step": 19396
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.053435805960918e-06,
+      "loss": 0.893,
+      "step": 19397
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0524566305250655e-06,
+      "loss": 0.9979,
+      "step": 19398
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.051477517896547e-06,
+      "loss": 0.8614,
+      "step": 19399
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.0504984680877875e-06,
+      "loss": 0.9902,
+      "step": 19400
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.04951948111122e-06,
+      "loss": 1.025,
+      "step": 19401
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.048540556979275e-06,
+      "loss": 0.8561,
+      "step": 19402
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 5.047561695704379e-06,
+      "loss": 0.9365,
+      "step": 19403
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.046582897298955e-06,
+      "loss": 1.0006,
+      "step": 19404
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.045604161775426e-06,
+      "loss": 0.9854,
+      "step": 19405
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0446254891462245e-06,
+      "loss": 0.9748,
+      "step": 19406
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.043646879423767e-06,
+      "loss": 0.86,
+      "step": 19407
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.042668332620483e-06,
+      "loss": 0.8963,
+      "step": 19408
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.041689848748787e-06,
+      "loss": 0.8345,
+      "step": 19409
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.040711427821111e-06,
+      "loss": 0.8952,
+      "step": 19410
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0397330698498694e-06,
+      "loss": 0.9652,
+      "step": 19411
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.038754774847483e-06,
+      "loss": 1.0689,
+      "step": 19412
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.037776542826369e-06,
+      "loss": 1.0184,
+      "step": 19413
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.036798373798946e-06,
+      "loss": 0.8594,
+      "step": 19414
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0358202677776305e-06,
+      "loss": 0.9108,
+      "step": 19415
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.034842224774845e-06,
+      "loss": 0.924,
+      "step": 19416
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.033864244803004e-06,
+      "loss": 0.9778,
+      "step": 19417
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.03288632787452e-06,
+      "loss": 1.0463,
+      "step": 19418
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.031908474001803e-06,
+      "loss": 0.9024,
+      "step": 19419
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.030930683197277e-06,
+      "loss": 0.9801,
+      "step": 19420
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.029952955473344e-06,
+      "loss": 1.0078,
+      "step": 19421
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.028975290842425e-06,
+      "loss": 0.8609,
+      "step": 19422
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027997689316924e-06,
+      "loss": 0.9023,
+      "step": 19423
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.027020150909257e-06,
+      "loss": 0.9666,
+      "step": 19424
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.026042675631831e-06,
+      "loss": 0.9628,
+      "step": 19425
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.025065263497055e-06,
+      "loss": 0.9733,
+      "step": 19426
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.024087914517334e-06,
+      "loss": 0.9451,
+      "step": 19427
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0231106287050815e-06,
+      "loss": 0.8864,
+      "step": 19428
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.022133406072693e-06,
+      "loss": 0.9199,
+      "step": 19429
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.021156246632588e-06,
+      "loss": 0.8896,
+      "step": 19430
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.020179150397163e-06,
+      "loss": 1.1635,
+      "step": 19431
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.019202117378823e-06,
+      "loss": 0.9312,
+      "step": 19432
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.018225147589968e-06,
+      "loss": 1.0065,
+      "step": 19433
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.017248241043006e-06,
+      "loss": 1.0116,
+      "step": 19434
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.016271397750334e-06,
+      "loss": 1.0673,
+      "step": 19435
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.015294617724357e-06,
+      "loss": 1.1096,
+      "step": 19436
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.014317900977468e-06,
+      "loss": 0.9059,
+      "step": 19437
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0133412475220765e-06,
+      "loss": 0.9125,
+      "step": 19438
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.012364657370573e-06,
+      "loss": 0.8786,
+      "step": 19439
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.011388130535357e-06,
+      "loss": 0.9951,
+      "step": 19440
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.010411667028822e-06,
+      "loss": 0.9606,
+      "step": 19441
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.009435266863368e-06,
+      "loss": 0.9179,
+      "step": 19442
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.00845893005139e-06,
+      "loss": 0.8641,
+      "step": 19443
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.007482656605283e-06,
+      "loss": 1.0169,
+      "step": 19444
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.006506446537438e-06,
+      "loss": 0.9777,
+      "step": 19445
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.005530299860248e-06,
+      "loss": 0.9808,
+      "step": 19446
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.004554216586101e-06,
+      "loss": 0.8542,
+      "step": 19447
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.003578196727396e-06,
+      "loss": 0.9918,
+      "step": 19448
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.002602240296517e-06,
+      "loss": 0.9852,
+      "step": 19449
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.001626347305858e-06,
+      "loss": 0.8521,
+      "step": 19450
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.0006505177678025e-06,
+      "loss": 1.0016,
+      "step": 19451
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.999674751694744e-06,
+      "loss": 0.8698,
+      "step": 19452
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9986990490990685e-06,
+      "loss": 1.0158,
+      "step": 19453
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.997723409993159e-06,
+      "loss": 0.9784,
+      "step": 19454
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.996747834389399e-06,
+      "loss": 0.8978,
+      "step": 19455
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9957723223001765e-06,
+      "loss": 0.9628,
+      "step": 19456
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.994796873737878e-06,
+      "loss": 1.0231,
+      "step": 19457
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.993821488714885e-06,
+      "loss": 0.9864,
+      "step": 19458
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.992846167243578e-06,
+      "loss": 0.985,
+      "step": 19459
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.991870909336338e-06,
+      "loss": 1.0455,
+      "step": 19460
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9908957150055436e-06,
+      "loss": 0.8983,
+      "step": 19461
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.989920584263581e-06,
+      "loss": 0.987,
+      "step": 19462
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.988945517122821e-06,
+      "loss": 0.9525,
+      "step": 19463
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.98797051359565e-06,
+      "loss": 0.8779,
+      "step": 19464
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986995573694437e-06,
+      "loss": 0.8931,
+      "step": 19465
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.986020697431569e-06,
+      "loss": 1.0044,
+      "step": 19466
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.985045884819413e-06,
+      "loss": 0.9807,
+      "step": 19467
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.984071135870348e-06,
+      "loss": 1.0011,
+      "step": 19468
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9830964505967426e-06,
+      "loss": 0.924,
+      "step": 19469
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.982121829010974e-06,
+      "loss": 0.9146,
+      "step": 19470
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.981147271125418e-06,
+      "loss": 0.8344,
+      "step": 19471
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.980172776952444e-06,
+      "loss": 0.8955,
+      "step": 19472
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.979198346504423e-06,
+      "loss": 0.996,
+      "step": 19473
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.978223979793722e-06,
+      "loss": 0.9398,
+      "step": 19474
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.97724967683271e-06,
+      "loss": 1.0681,
+      "step": 19475
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9762754376337575e-06,
+      "loss": 0.9547,
+      "step": 19476
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.975301262209236e-06,
+      "loss": 0.9127,
+      "step": 19477
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.974327150571507e-06,
+      "loss": 0.9614,
+      "step": 19478
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.973353102732936e-06,
+      "loss": 1.0161,
+      "step": 19479
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.972379118705894e-06,
+      "loss": 0.9647,
+      "step": 19480
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.971405198502742e-06,
+      "loss": 0.9496,
+      "step": 19481
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.970431342135844e-06,
+      "loss": 1.1082,
+      "step": 19482
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.969457549617557e-06,
+      "loss": 0.957,
+      "step": 19483
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.968483820960248e-06,
+      "loss": 0.9013,
+      "step": 19484
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.967510156176283e-06,
+      "loss": 1.0126,
+      "step": 19485
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.966536555278016e-06,
+      "loss": 0.8823,
+      "step": 19486
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.965563018277809e-06,
+      "loss": 0.9368,
+      "step": 19487
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.96458954518802e-06,
+      "loss": 0.957,
+      "step": 19488
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.963616136021003e-06,
+      "loss": 1.0296,
+      "step": 19489
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.962642790789118e-06,
+      "loss": 0.8868,
+      "step": 19490
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.961669509504726e-06,
+      "loss": 0.8997,
+      "step": 19491
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9606962921801795e-06,
+      "loss": 0.8525,
+      "step": 19492
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.959723138827827e-06,
+      "loss": 1.0034,
+      "step": 19493
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9587500494600315e-06,
+      "loss": 1.0402,
+      "step": 19494
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.957777024089143e-06,
+      "loss": 0.9047,
+      "step": 19495
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.956804062727512e-06,
+      "loss": 0.8921,
+      "step": 19496
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9558311653874866e-06,
+      "loss": 0.9497,
+      "step": 19497
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9548583320814226e-06,
+      "loss": 0.9555,
+      "step": 19498
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.953885562821671e-06,
+      "loss": 0.8371,
+      "step": 19499
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.952912857620579e-06,
+      "loss": 1.0515,
+      "step": 19500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.951940216490494e-06,
+      "loss": 0.9583,
+      "step": 19501
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.950967639443764e-06,
+      "loss": 0.8945,
+      "step": 19502
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9499951264927304e-06,
+      "loss": 0.9277,
+      "step": 19503
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.949022677649744e-06,
+      "loss": 1.0394,
+      "step": 19504
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.948050292927153e-06,
+      "loss": 1.0274,
+      "step": 19505
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.947077972337298e-06,
+      "loss": 0.9289,
+      "step": 19506
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.946105715892517e-06,
+      "loss": 1.0389,
+      "step": 19507
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.945133523605162e-06,
+      "loss": 1.0341,
+      "step": 19508
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94416139548757e-06,
+      "loss": 0.8704,
+      "step": 19509
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9431893315520816e-06,
+      "loss": 0.8646,
+      "step": 19510
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.942217331811034e-06,
+      "loss": 0.9699,
+      "step": 19511
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.941245396276769e-06,
+      "loss": 0.9669,
+      "step": 19512
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.94027352496163e-06,
+      "loss": 0.9854,
+      "step": 19513
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.939301717877949e-06,
+      "loss": 0.8927,
+      "step": 19514
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.938329975038065e-06,
+      "loss": 0.9809,
+      "step": 19515
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.937358296454311e-06,
+      "loss": 0.9487,
+      "step": 19516
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.936386682139021e-06,
+      "loss": 1.0661,
+      "step": 19517
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.935415132104532e-06,
+      "loss": 1.0268,
+      "step": 19518
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.934443646363181e-06,
+      "loss": 0.974,
+      "step": 19519
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.933472224927298e-06,
+      "loss": 0.9943,
+      "step": 19520
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.93250086780921e-06,
+      "loss": 0.9465,
+      "step": 19521
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.931529575021257e-06,
+      "loss": 0.9427,
+      "step": 19522
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.930558346575763e-06,
+      "loss": 0.9745,
+      "step": 19523
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9295871824850565e-06,
+      "loss": 0.9829,
+      "step": 19524
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.928616082761472e-06,
+      "loss": 1.0419,
+      "step": 19525
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9276450474173286e-06,
+      "loss": 0.88,
+      "step": 19526
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.926674076464962e-06,
+      "loss": 0.947,
+      "step": 19527
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.925703169916697e-06,
+      "loss": 0.8957,
+      "step": 19528
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9247323277848545e-06,
+      "loss": 0.8867,
+      "step": 19529
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.923761550081761e-06,
+      "loss": 0.8145,
+      "step": 19530
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.922790836819739e-06,
+      "loss": 0.9763,
+      "step": 19531
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9218201880111105e-06,
+      "loss": 1.0301,
+      "step": 19532
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.920849603668203e-06,
+      "loss": 1.0726,
+      "step": 19533
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.919879083803336e-06,
+      "loss": 0.9745,
+      "step": 19534
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.918908628428823e-06,
+      "loss": 0.9361,
+      "step": 19535
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.917938237556994e-06,
+      "loss": 0.9553,
+      "step": 19536
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.916967911200162e-06,
+      "loss": 0.8618,
+      "step": 19537
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915997649370643e-06,
+      "loss": 0.9616,
+      "step": 19538
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.915027452080759e-06,
+      "loss": 1.0128,
+      "step": 19539
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.914057319342822e-06,
+      "loss": 0.9712,
+      "step": 19540
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.913087251169153e-06,
+      "loss": 1.1656,
+      "step": 19541
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.912117247572063e-06,
+      "loss": 1.0147,
+      "step": 19542
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.911147308563865e-06,
+      "loss": 0.9814,
+      "step": 19543
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.910177434156874e-06,
+      "loss": 0.9247,
+      "step": 19544
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9092076243633975e-06,
+      "loss": 1.0675,
+      "step": 19545
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.908237879195752e-06,
+      "loss": 0.868,
+      "step": 19546
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9072681986662495e-06,
+      "loss": 1.0472,
+      "step": 19547
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.906298582787197e-06,
+      "loss": 0.9369,
+      "step": 19548
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9053290315709004e-06,
+      "loss": 0.9608,
+      "step": 19549
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.904359545029674e-06,
+      "loss": 0.9327,
+      "step": 19550
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.903390123175823e-06,
+      "loss": 0.9448,
+      "step": 19551
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.9024207660216485e-06,
+      "loss": 0.9792,
+      "step": 19552
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.901451473579465e-06,
+      "loss": 0.9253,
+      "step": 19553
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.900482245861569e-06,
+      "loss": 0.9106,
+      "step": 19554
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.899513082880271e-06,
+      "loss": 0.9146,
+      "step": 19555
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.898543984647872e-06,
+      "loss": 0.9677,
+      "step": 19556
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.897574951176675e-06,
+      "loss": 1.0088,
+      "step": 19557
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.896605982478976e-06,
+      "loss": 0.9282,
+      "step": 19558
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.895637078567084e-06,
+      "loss": 0.9173,
+      "step": 19559
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.894668239453292e-06,
+      "loss": 1.0063,
+      "step": 19560
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.893699465149905e-06,
+      "loss": 0.9563,
+      "step": 19561
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.892730755669219e-06,
+      "loss": 1.0813,
+      "step": 19562
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.891762111023527e-06,
+      "loss": 1.0515,
+      "step": 19563
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.890793531225133e-06,
+      "loss": 0.8996,
+      "step": 19564
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8898250162863305e-06,
+      "loss": 0.9639,
+      "step": 19565
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.888856566219408e-06,
+      "loss": 0.9117,
+      "step": 19566
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.887888181036669e-06,
+      "loss": 0.8839,
+      "step": 19567
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.886919860750399e-06,
+      "loss": 1.022,
+      "step": 19568
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.885951605372899e-06,
+      "loss": 0.955,
+      "step": 19569
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884983414916454e-06,
+      "loss": 0.9455,
+      "step": 19570
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.884015289393358e-06,
+      "loss": 0.96,
+      "step": 19571
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.883047228815895e-06,
+      "loss": 0.9709,
+      "step": 19572
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.88207923319636e-06,
+      "loss": 1.0075,
+      "step": 19573
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.881111302547042e-06,
+      "loss": 0.8738,
+      "step": 19574
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8801434368802295e-06,
+      "loss": 0.9729,
+      "step": 19575
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.879175636208204e-06,
+      "loss": 0.9603,
+      "step": 19576
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.878207900543252e-06,
+      "loss": 0.9733,
+      "step": 19577
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.877240229897663e-06,
+      "loss": 0.9828,
+      "step": 19578
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.876272624283719e-06,
+      "loss": 0.9222,
+      "step": 19579
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8753050837136994e-06,
+      "loss": 0.8781,
+      "step": 19580
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.874337608199894e-06,
+      "loss": 1.0307,
+      "step": 19581
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.873370197754578e-06,
+      "loss": 1.0358,
+      "step": 19582
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.872402852390038e-06,
+      "loss": 1.0375,
+      "step": 19583
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.871435572118553e-06,
+      "loss": 1.0385,
+      "step": 19584
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.870468356952399e-06,
+      "loss": 1.0155,
+      "step": 19585
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.869501206903854e-06,
+      "loss": 0.9076,
+      "step": 19586
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.868534121985198e-06,
+      "loss": 0.9405,
+      "step": 19587
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.86756710220871e-06,
+      "loss": 0.8547,
+      "step": 19588
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.866600147586664e-06,
+      "loss": 0.9323,
+      "step": 19589
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.865633258131336e-06,
+      "loss": 0.9844,
+      "step": 19590
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.864666433854994e-06,
+      "loss": 1.0044,
+      "step": 19591
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.863699674769922e-06,
+      "loss": 0.9738,
+      "step": 19592
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.862732980888386e-06,
+      "loss": 0.9638,
+      "step": 19593
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.861766352222656e-06,
+      "loss": 0.9924,
+      "step": 19594
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8607997887850104e-06,
+      "loss": 0.9335,
+      "step": 19595
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85983329058771e-06,
+      "loss": 0.9377,
+      "step": 19596
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.858866857643034e-06,
+      "loss": 1.1149,
+      "step": 19597
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.857900489963245e-06,
+      "loss": 0.9856,
+      "step": 19598
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8569341875606135e-06,
+      "loss": 0.9718,
+      "step": 19599
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8559679504474e-06,
+      "loss": 0.8866,
+      "step": 19600
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.855001778635876e-06,
+      "loss": 1.02,
+      "step": 19601
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.85403567213831e-06,
+      "loss": 0.968,
+      "step": 19602
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.853069630966962e-06,
+      "loss": 1.0638,
+      "step": 19603
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.852103655134096e-06,
+      "loss": 1.0164,
+      "step": 19604
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.851137744651971e-06,
+      "loss": 0.8818,
+      "step": 19605
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.850171899532856e-06,
+      "loss": 0.9878,
+      "step": 19606
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.849206119789005e-06,
+      "loss": 0.8842,
+      "step": 19607
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.848240405432687e-06,
+      "loss": 0.9694,
+      "step": 19608
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.847274756476155e-06,
+      "loss": 0.9819,
+      "step": 19609
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8463091729316646e-06,
+      "loss": 0.9278,
+      "step": 19610
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.845343654811483e-06,
+      "loss": 0.9428,
+      "step": 19611
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.844378202127861e-06,
+      "loss": 0.983,
+      "step": 19612
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.843412814893057e-06,
+      "loss": 0.8906,
+      "step": 19613
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.84244749311932e-06,
+      "loss": 0.9314,
+      "step": 19614
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.841482236818909e-06,
+      "loss": 1.0059,
+      "step": 19615
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8405170460040826e-06,
+      "loss": 0.8882,
+      "step": 19616
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.839551920687088e-06,
+      "loss": 0.8171,
+      "step": 19617
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.838586860880179e-06,
+      "loss": 0.8725,
+      "step": 19618
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.837621866595601e-06,
+      "loss": 1.0219,
+      "step": 19619
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.836656937845613e-06,
+      "loss": 0.8505,
+      "step": 19620
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8356920746424575e-06,
+      "loss": 0.9443,
+      "step": 19621
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.834727276998389e-06,
+      "loss": 0.9676,
+      "step": 19622
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.833762544925652e-06,
+      "loss": 0.9987,
+      "step": 19623
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.83279787843649e-06,
+      "loss": 1.0331,
+      "step": 19624
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.831833277543157e-06,
+      "loss": 0.9459,
+      "step": 19625
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.830868742257894e-06,
+      "loss": 0.9774,
+      "step": 19626
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.829904272592945e-06,
+      "loss": 1.03,
+      "step": 19627
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82893986856055e-06,
+      "loss": 0.8865,
+      "step": 19628
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.827975530172957e-06,
+      "loss": 1.027,
+      "step": 19629
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8270112574424114e-06,
+      "loss": 0.9873,
+      "step": 19630
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.826047050381147e-06,
+      "loss": 0.9906,
+      "step": 19631
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.82508290900141e-06,
+      "loss": 0.8502,
+      "step": 19632
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.824118833315431e-06,
+      "loss": 0.9139,
+      "step": 19633
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.823154823335457e-06,
+      "loss": 1.0045,
+      "step": 19634
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.822190879073721e-06,
+      "loss": 0.9068,
+      "step": 19635
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8212270005424655e-06,
+      "loss": 0.9439,
+      "step": 19636
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.820263187753923e-06,
+      "loss": 0.9665,
+      "step": 19637
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8192994407203255e-06,
+      "loss": 1.0602,
+      "step": 19638
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.818335759453914e-06,
+      "loss": 1.002,
+      "step": 19639
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.817372143966919e-06,
+      "loss": 0.9037,
+      "step": 19640
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.816408594271572e-06,
+      "loss": 1.0397,
+      "step": 19641
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8154451103801035e-06,
+      "loss": 1.133,
+      "step": 19642
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.814481692304747e-06,
+      "loss": 0.9688,
+      "step": 19643
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.813518340057736e-06,
+      "loss": 1.0266,
+      "step": 19644
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.812555053651297e-06,
+      "loss": 1.0934,
+      "step": 19645
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.811591833097659e-06,
+      "loss": 0.9946,
+      "step": 19646
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.810628678409045e-06,
+      "loss": 1.0609,
+      "step": 19647
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80966558959769e-06,
+      "loss": 1.0191,
+      "step": 19648
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.808702566675812e-06,
+      "loss": 1.0794,
+      "step": 19649
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.807739609655644e-06,
+      "loss": 0.8939,
+      "step": 19650
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8067767185494075e-06,
+      "loss": 1.0167,
+      "step": 19651
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.805813893369321e-06,
+      "loss": 1.0104,
+      "step": 19652
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.804851134127615e-06,
+      "loss": 0.9671,
+      "step": 19653
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.80388844083651e-06,
+      "loss": 1.0736,
+      "step": 19654
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.8029258135082195e-06,
+      "loss": 0.9036,
+      "step": 19655
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801963252154972e-06,
+      "loss": 1.0812,
+      "step": 19656
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.801000756788982e-06,
+      "loss": 1.0352,
+      "step": 19657
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.800038327422473e-06,
+      "loss": 0.9796,
+      "step": 19658
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.799075964067661e-06,
+      "loss": 0.9879,
+      "step": 19659
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.798113666736761e-06,
+      "loss": 0.9192,
+      "step": 19660
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.797151435441986e-06,
+      "loss": 0.9656,
+      "step": 19661
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.796189270195558e-06,
+      "loss": 0.9953,
+      "step": 19662
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.795227171009686e-06,
+      "loss": 1.0169,
+      "step": 19663
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.794265137896589e-06,
+      "loss": 1.0216,
+      "step": 19664
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.793303170868476e-06,
+      "loss": 0.9019,
+      "step": 19665
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.792341269937556e-06,
+      "loss": 0.9577,
+      "step": 19666
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.791379435116047e-06,
+      "loss": 0.9281,
+      "step": 19667
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.790417666416155e-06,
+      "loss": 1.0134,
+      "step": 19668
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7894559638500874e-06,
+      "loss": 0.9157,
+      "step": 19669
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.788494327430053e-06,
+      "loss": 0.9424,
+      "step": 19670
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.787532757168267e-06,
+      "loss": 0.9273,
+      "step": 19671
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.786571253076931e-06,
+      "loss": 0.8793,
+      "step": 19672
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.785609815168249e-06,
+      "loss": 1.0214,
+      "step": 19673
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7846484434544285e-06,
+      "loss": 0.8745,
+      "step": 19674
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.783687137947669e-06,
+      "loss": 0.8521,
+      "step": 19675
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.782725898660182e-06,
+      "loss": 0.896,
+      "step": 19676
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.781764725604162e-06,
+      "loss": 1.0048,
+      "step": 19677
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.780803618791819e-06,
+      "loss": 0.9054,
+      "step": 19678
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.779842578235349e-06,
+      "loss": 0.9357,
+      "step": 19679
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.778881603946951e-06,
+      "loss": 0.9583,
+      "step": 19680
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.777920695938829e-06,
+      "loss": 0.8058,
+      "step": 19681
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7769598542231775e-06,
+      "loss": 0.9844,
+      "step": 19682
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775999078812193e-06,
+      "loss": 0.889,
+      "step": 19683
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.775038369718073e-06,
+      "loss": 0.9786,
+      "step": 19684
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.774077726953017e-06,
+      "loss": 0.8674,
+      "step": 19685
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.773117150529219e-06,
+      "loss": 1.0085,
+      "step": 19686
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7721566404588725e-06,
+      "loss": 0.9932,
+      "step": 19687
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.77119619675417e-06,
+      "loss": 1.0594,
+      "step": 19688
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.7702358194272994e-06,
+      "loss": 0.9569,
+      "step": 19689
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.769275508490461e-06,
+      "loss": 0.8909,
+      "step": 19690
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 4.768315263955838e-06,
+      "loss": 0.9765,
+      "step": 19691
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7673550858356275e-06,
+      "loss": 0.9813,
+      "step": 19692
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.766394974142012e-06,
+      "loss": 0.8858,
+      "step": 19693
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.765434928887187e-06,
+      "loss": 0.8442,
+      "step": 19694
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.764474950083335e-06,
+      "loss": 0.8729,
+      "step": 19695
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.763515037742644e-06,
+      "loss": 0.922,
+      "step": 19696
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.762555191877296e-06,
+      "loss": 0.945,
+      "step": 19697
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.761595412499479e-06,
+      "loss": 0.9605,
+      "step": 19698
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.760635699621381e-06,
+      "loss": 1.0085,
+      "step": 19699
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.759676053255181e-06,
+      "loss": 0.9663,
+      "step": 19700
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.758716473413062e-06,
+      "loss": 1.0019,
+      "step": 19701
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.757756960107205e-06,
+      "loss": 0.9806,
+      "step": 19702
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.756797513349788e-06,
+      "loss": 0.9504,
+      "step": 19703
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.755838133152993e-06,
+      "loss": 0.9835,
+      "step": 19704
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.754878819529004e-06,
+      "loss": 0.8497,
+      "step": 19705
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.753919572489996e-06,
+      "loss": 0.9658,
+      "step": 19706
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752960392048141e-06,
+      "loss": 0.967,
+      "step": 19707
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.752001278215623e-06,
+      "loss": 0.9938,
+      "step": 19708
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.751042231004613e-06,
+      "loss": 1.0105,
+      "step": 19709
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.75008325042729e-06,
+      "loss": 1.005,
+      "step": 19710
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.749124336495819e-06,
+      "loss": 1.016,
+      "step": 19711
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.748165489222382e-06,
+      "loss": 0.9117,
+      "step": 19712
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.74720670861915e-06,
+      "loss": 1.0176,
+      "step": 19713
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.746247994698293e-06,
+      "loss": 0.961,
+      "step": 19714
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.745289347471982e-06,
+      "loss": 0.9067,
+      "step": 19715
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.744330766952385e-06,
+      "loss": 0.9304,
+      "step": 19716
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.743372253151669e-06,
+      "loss": 0.972,
+      "step": 19717
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.742413806082005e-06,
+      "loss": 0.9351,
+      "step": 19718
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.741455425755563e-06,
+      "loss": 0.9888,
+      "step": 19719
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.740497112184507e-06,
+      "loss": 0.9958,
+      "step": 19720
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.739538865380997e-06,
+      "loss": 0.9196,
+      "step": 19721
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.738580685357207e-06,
+      "loss": 0.954,
+      "step": 19722
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.737622572125297e-06,
+      "loss": 0.9724,
+      "step": 19723
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.736664525697427e-06,
+      "loss": 0.9675,
+      "step": 19724
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.735706546085759e-06,
+      "loss": 0.9475,
+      "step": 19725
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7347486333024554e-06,
+      "loss": 1.1042,
+      "step": 19726
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.733790787359682e-06,
+      "loss": 0.9769,
+      "step": 19727
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.732833008269593e-06,
+      "loss": 0.9181,
+      "step": 19728
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.731875296044348e-06,
+      "loss": 0.7963,
+      "step": 19729
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.730917650696105e-06,
+      "loss": 0.9976,
+      "step": 19730
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729960072237017e-06,
+      "loss": 0.948,
+      "step": 19731
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.729002560679244e-06,
+      "loss": 0.864,
+      "step": 19732
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.728045116034945e-06,
+      "loss": 1.0241,
+      "step": 19733
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7270877383162705e-06,
+      "loss": 1.0516,
+      "step": 19734
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72613042753537e-06,
+      "loss": 0.9922,
+      "step": 19735
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7251731837044035e-06,
+      "loss": 0.9465,
+      "step": 19736
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.72421600683552e-06,
+      "loss": 0.9777,
+      "step": 19737
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7232588969408665e-06,
+      "loss": 0.9459,
+      "step": 19738
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.722301854032601e-06,
+      "loss": 0.9207,
+      "step": 19739
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.721344878122865e-06,
+      "loss": 0.9744,
+      "step": 19740
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7203879692238145e-06,
+      "loss": 0.9359,
+      "step": 19741
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.719431127347593e-06,
+      "loss": 1.0948,
+      "step": 19742
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.718474352506347e-06,
+      "loss": 1.0754,
+      "step": 19743
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.717517644712224e-06,
+      "loss": 0.8424,
+      "step": 19744
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.716561003977364e-06,
+      "loss": 0.9925,
+      "step": 19745
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.715604430313915e-06,
+      "loss": 0.811,
+      "step": 19746
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.714647923734025e-06,
+      "loss": 1.0302,
+      "step": 19747
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.713691484249832e-06,
+      "loss": 0.8846,
+      "step": 19748
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7127351118734746e-06,
+      "loss": 0.947,
+      "step": 19749
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.711778806617099e-06,
+      "loss": 0.8912,
+      "step": 19750
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.710822568492844e-06,
+      "loss": 1.0123,
+      "step": 19751
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.709866397512844e-06,
+      "loss": 1.1133,
+      "step": 19752
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.708910293689245e-06,
+      "loss": 1.006,
+      "step": 19753
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.707954257034176e-06,
+      "loss": 1.0219,
+      "step": 19754
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.706998287559782e-06,
+      "loss": 0.8782,
+      "step": 19755
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7060423852781955e-06,
+      "loss": 0.9637,
+      "step": 19756
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7050865502015496e-06,
+      "loss": 0.9519,
+      "step": 19757
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.704130782341979e-06,
+      "loss": 0.9367,
+      "step": 19758
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.703175081711614e-06,
+      "loss": 0.8897,
+      "step": 19759
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.702219448322589e-06,
+      "loss": 0.9402,
+      "step": 19760
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.70126388218704e-06,
+      "loss": 0.9762,
+      "step": 19761
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.7003083833170946e-06,
+      "loss": 0.9193,
+      "step": 19762
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6993529517248785e-06,
+      "loss": 0.9165,
+      "step": 19763
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.698397587422528e-06,
+      "loss": 0.917,
+      "step": 19764
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6974422904221655e-06,
+      "loss": 0.8638,
+      "step": 19765
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.696487060735916e-06,
+      "loss": 1.0334,
+      "step": 19766
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.695531898375914e-06,
+      "loss": 0.9743,
+      "step": 19767
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.694576803354276e-06,
+      "loss": 0.8899,
+      "step": 19768
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.693621775683136e-06,
+      "loss": 0.8845,
+      "step": 19769
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.692666815374612e-06,
+      "loss": 0.857,
+      "step": 19770
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.691711922440828e-06,
+      "loss": 1.0283,
+      "step": 19771
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6907570968939055e-06,
+      "loss": 0.9941,
+      "step": 19772
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.689802338745961e-06,
+      "loss": 0.9663,
+      "step": 19773
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.688847648009122e-06,
+      "loss": 0.8183,
+      "step": 19774
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.687893024695508e-06,
+      "loss": 1.0271,
+      "step": 19775
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6869384688172345e-06,
+      "loss": 0.8862,
+      "step": 19776
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6859839803864186e-06,
+      "loss": 0.9134,
+      "step": 19777
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.685029559415181e-06,
+      "loss": 0.9667,
+      "step": 19778
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6840752059156345e-06,
+      "loss": 0.8701,
+      "step": 19779
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.683120919899893e-06,
+      "loss": 0.9218,
+      "step": 19780
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.682166701380076e-06,
+      "loss": 0.8855,
+      "step": 19781
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.681212550368289e-06,
+      "loss": 0.9212,
+      "step": 19782
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.680258466876655e-06,
+      "loss": 0.9753,
+      "step": 19783
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.67930445091728e-06,
+      "loss": 1.0384,
+      "step": 19784
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6783505025022755e-06,
+      "loss": 0.9299,
+      "step": 19785
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.677396621643747e-06,
+      "loss": 0.9706,
+      "step": 19786
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.676442808353812e-06,
+      "loss": 0.9289,
+      "step": 19787
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.675489062644571e-06,
+      "loss": 0.9993,
+      "step": 19788
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.674535384528138e-06,
+      "loss": 0.8507,
+      "step": 19789
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.673581774016618e-06,
+      "loss": 0.8689,
+      "step": 19790
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.672628231122111e-06,
+      "loss": 0.9182,
+      "step": 19791
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.671674755856731e-06,
+      "loss": 0.9858,
+      "step": 19792
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.670721348232577e-06,
+      "loss": 1.0415,
+      "step": 19793
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.669768008261749e-06,
+      "loss": 0.9579,
+      "step": 19794
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.668814735956358e-06,
+      "loss": 1.0104,
+      "step": 19795
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.667861531328496e-06,
+      "loss": 0.9138,
+      "step": 19796
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.66690839439027e-06,
+      "loss": 0.9934,
+      "step": 19797
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.66595532515378e-06,
+      "loss": 0.8783,
+      "step": 19798
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.665002323631123e-06,
+      "loss": 0.9226,
+      "step": 19799
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.664049389834392e-06,
+      "loss": 0.8843,
+      "step": 19800
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.663096523775689e-06,
+      "loss": 0.9151,
+      "step": 19801
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.662143725467113e-06,
+      "loss": 1.1043,
+      "step": 19802
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6611909949207575e-06,
+      "loss": 0.78,
+      "step": 19803
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6602383321487165e-06,
+      "loss": 0.9674,
+      "step": 19804
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.65928573716308e-06,
+      "loss": 0.9075,
+      "step": 19805
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.658333209975947e-06,
+      "loss": 0.8761,
+      "step": 19806
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6573807505994074e-06,
+      "loss": 0.9397,
+      "step": 19807
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6564283590455476e-06,
+      "loss": 0.9772,
+      "step": 19808
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.655476035326466e-06,
+      "loss": 0.8202,
+      "step": 19809
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.654523779454244e-06,
+      "loss": 1.0994,
+      "step": 19810
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.653571591440979e-06,
+      "loss": 0.8838,
+      "step": 19811
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.652619471298753e-06,
+      "loss": 0.985,
+      "step": 19812
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.651667419039654e-06,
+      "loss": 0.9653,
+      "step": 19813
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.650715434675763e-06,
+      "loss": 0.9582,
+      "step": 19814
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6497635182191706e-06,
+      "loss": 0.8353,
+      "step": 19815
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.648811669681964e-06,
+      "loss": 0.9419,
+      "step": 19816
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6478598890762225e-06,
+      "loss": 0.9887,
+      "step": 19817
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.646908176414029e-06,
+      "loss": 1.1495,
+      "step": 19818
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6459565317074614e-06,
+      "loss": 0.9775,
+      "step": 19819
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.645004954968608e-06,
+      "loss": 0.985,
+      "step": 19820
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.644053446209546e-06,
+      "loss": 0.9319,
+      "step": 19821
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.643102005442348e-06,
+      "loss": 1.0927,
+      "step": 19822
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.642150632679102e-06,
+      "loss": 0.9772,
+      "step": 19823
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.641199327931877e-06,
+      "loss": 0.9352,
+      "step": 19824
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.640248091212758e-06,
+      "loss": 0.9506,
+      "step": 19825
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.639296922533815e-06,
+      "loss": 0.9454,
+      "step": 19826
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.638345821907124e-06,
+      "loss": 0.9813,
+      "step": 19827
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6373947893447554e-06,
+      "loss": 1.012,
+      "step": 19828
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6364438248587855e-06,
+      "loss": 0.8334,
+      "step": 19829
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.63549292846129e-06,
+      "loss": 1.0016,
+      "step": 19830
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.634542100164336e-06,
+      "loss": 0.8713,
+      "step": 19831
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.633591339979996e-06,
+      "loss": 0.9574,
+      "step": 19832
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6326406479203336e-06,
+      "loss": 1.0178,
+      "step": 19833
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.631690023997426e-06,
+      "loss": 1.0758,
+      "step": 19834
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.630739468223333e-06,
+      "loss": 0.9425,
+      "step": 19835
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62978898061013e-06,
+      "loss": 0.8462,
+      "step": 19836
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.628838561169879e-06,
+      "loss": 1.018,
+      "step": 19837
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.627888209914641e-06,
+      "loss": 0.9376,
+      "step": 19838
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.626937926856487e-06,
+      "loss": 0.9216,
+      "step": 19839
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.625987712007481e-06,
+      "loss": 0.9961,
+      "step": 19840
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6250375653796795e-06,
+      "loss": 0.9329,
+      "step": 19841
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.624087486985145e-06,
+      "loss": 0.9186,
+      "step": 19842
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.623137476835941e-06,
+      "loss": 0.8065,
+      "step": 19843
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.62218753494413e-06,
+      "loss": 0.9892,
+      "step": 19844
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.621237661321769e-06,
+      "loss": 0.9843,
+      "step": 19845
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.620287855980916e-06,
+      "loss": 0.9371,
+      "step": 19846
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.619338118933624e-06,
+      "loss": 1.0287,
+      "step": 19847
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.618388450191959e-06,
+      "loss": 1.0148,
+      "step": 19848
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.617438849767966e-06,
+      "loss": 0.9129,
+      "step": 19849
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61648931767371e-06,
+      "loss": 0.8029,
+      "step": 19850
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.61553985392124e-06,
+      "loss": 0.8607,
+      "step": 19851
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.614590458522606e-06,
+      "loss": 0.8707,
+      "step": 19852
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6136411314898675e-06,
+      "loss": 0.9372,
+      "step": 19853
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.612691872835074e-06,
+      "loss": 0.9784,
+      "step": 19854
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.6117426825702716e-06,
+      "loss": 1.0192,
+      "step": 19855
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.610793560707509e-06,
+      "loss": 0.8845,
+      "step": 19856
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.609844507258838e-06,
+      "loss": 0.923,
+      "step": 19857
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.608895522236312e-06,
+      "loss": 0.8945,
+      "step": 19858
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.607946605651973e-06,
+      "loss": 0.8589,
+      "step": 19859
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606997757517866e-06,
+      "loss": 1.0372,
+      "step": 19860
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.606048977846035e-06,
+      "loss": 1.0102,
+      "step": 19861
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.605100266648529e-06,
+      "loss": 1.0588,
+      "step": 19862
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.604151623937386e-06,
+      "loss": 0.875,
+      "step": 19863
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.603203049724657e-06,
+      "loss": 1.005,
+      "step": 19864
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.602254544022379e-06,
+      "loss": 1.001,
+      "step": 19865
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.601306106842588e-06,
+      "loss": 0.9121,
+      "step": 19866
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.600357738197333e-06,
+      "loss": 1.0333,
+      "step": 19867
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.599409438098649e-06,
+      "loss": 0.8479,
+      "step": 19868
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.598461206558571e-06,
+      "loss": 0.9798,
+      "step": 19869
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.597513043589143e-06,
+      "loss": 0.9993,
+      "step": 19870
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.596564949202396e-06,
+      "loss": 0.9699,
+      "step": 19871
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.595616923410371e-06,
+      "loss": 0.9415,
+      "step": 19872
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.594668966225101e-06,
+      "loss": 0.9539,
+      "step": 19873
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.593721077658618e-06,
+      "loss": 0.8964,
+      "step": 19874
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.592773257722953e-06,
+      "loss": 0.9773,
+      "step": 19875
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.591825506430145e-06,
+      "loss": 1.0191,
+      "step": 19876
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.590877823792219e-06,
+      "loss": 0.8939,
+      "step": 19877
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.589930209821212e-06,
+      "loss": 1.0274,
+      "step": 19878
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.588982664529149e-06,
+      "loss": 1.0033,
+      "step": 19879
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.588035187928057e-06,
+      "loss": 0.9195,
+      "step": 19880
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5870877800299695e-06,
+      "loss": 0.8912,
+      "step": 19881
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.586140440846911e-06,
+      "loss": 0.95,
+      "step": 19882
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.585193170390903e-06,
+      "loss": 1.0063,
+      "step": 19883
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.584245968673976e-06,
+      "loss": 0.8227,
+      "step": 19884
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.583298835708157e-06,
+      "loss": 0.9368,
+      "step": 19885
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.582351771505465e-06,
+      "loss": 0.8707,
+      "step": 19886
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.581404776077925e-06,
+      "loss": 1.0249,
+      "step": 19887
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.580457849437556e-06,
+      "loss": 0.9344,
+      "step": 19888
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.579510991596378e-06,
+      "loss": 0.8235,
+      "step": 19889
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.578564202566416e-06,
+      "loss": 0.9332,
+      "step": 19890
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.577617482359682e-06,
+      "loss": 0.9793,
+      "step": 19891
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.576670830988204e-06,
+      "loss": 0.8714,
+      "step": 19892
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.575724248463993e-06,
+      "loss": 1.016,
+      "step": 19893
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.574777734799063e-06,
+      "loss": 0.8308,
+      "step": 19894
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.573831290005438e-06,
+      "loss": 1.0565,
+      "step": 19895
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.572884914095128e-06,
+      "loss": 0.9193,
+      "step": 19896
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.571938607080143e-06,
+      "loss": 0.957,
+      "step": 19897
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5709923689725e-06,
+      "loss": 0.9311,
+      "step": 19898
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.570046199784215e-06,
+      "loss": 1.0216,
+      "step": 19899
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.569100099527295e-06,
+      "loss": 0.964,
+      "step": 19900
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.568154068213752e-06,
+      "loss": 0.9389,
+      "step": 19901
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.567208105855594e-06,
+      "loss": 1.0072,
+      "step": 19902
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.566262212464826e-06,
+      "loss": 0.8911,
+      "step": 19903
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.565316388053464e-06,
+      "loss": 0.8806,
+      "step": 19904
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.564370632633507e-06,
+      "loss": 1.001,
+      "step": 19905
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.563424946216969e-06,
+      "loss": 0.9249,
+      "step": 19906
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.562479328815851e-06,
+      "loss": 0.8821,
+      "step": 19907
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.561533780442153e-06,
+      "loss": 0.9652,
+      "step": 19908
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.560588301107886e-06,
+      "loss": 0.8813,
+      "step": 19909
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5596428908250515e-06,
+      "loss": 0.9442,
+      "step": 19910
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5586975496056425e-06,
+      "loss": 0.95,
+      "step": 19911
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.557752277461668e-06,
+      "loss": 1.0067,
+      "step": 19912
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.556807074405128e-06,
+      "loss": 1.0816,
+      "step": 19913
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.555861940448022e-06,
+      "loss": 0.9196,
+      "step": 19914
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.554916875602344e-06,
+      "loss": 1.0282,
+      "step": 19915
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553971879880094e-06,
+      "loss": 1.0311,
+      "step": 19916
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.553026953293262e-06,
+      "loss": 0.8943,
+      "step": 19917
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.552082095853855e-06,
+      "loss": 1.0326,
+      "step": 19918
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.551137307573857e-06,
+      "loss": 0.9185,
+      "step": 19919
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.550192588465268e-06,
+      "loss": 0.957,
+      "step": 19920
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.549247938540082e-06,
+      "loss": 0.924,
+      "step": 19921
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.548303357810283e-06,
+      "loss": 0.9496,
+      "step": 19922
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.547358846287872e-06,
+      "loss": 0.9614,
+      "step": 19923
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.546414403984833e-06,
+      "loss": 1.0709,
+      "step": 19924
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.545470030913154e-06,
+      "loss": 1.0512,
+      "step": 19925
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.544525727084826e-06,
+      "loss": 0.9683,
+      "step": 19926
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.543581492511842e-06,
+      "loss": 1.0112,
+      "step": 19927
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.542637327206183e-06,
+      "loss": 0.9044,
+      "step": 19928
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.541693231179836e-06,
+      "loss": 0.9399,
+      "step": 19929
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.540749204444784e-06,
+      "loss": 0.9901,
+      "step": 19930
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.539805247013011e-06,
+      "loss": 0.8704,
+      "step": 19931
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5388613588965e-06,
+      "loss": 0.8517,
+      "step": 19932
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.537917540107241e-06,
+      "loss": 0.8436,
+      "step": 19933
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536973790657208e-06,
+      "loss": 0.9449,
+      "step": 19934
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.536030110558384e-06,
+      "loss": 0.9757,
+      "step": 19935
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.535086499822743e-06,
+      "loss": 0.8945,
+      "step": 19936
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.534142958462273e-06,
+      "loss": 1.0143,
+      "step": 19937
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.533199486488948e-06,
+      "loss": 0.9217,
+      "step": 19938
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.532256083914741e-06,
+      "loss": 1.0048,
+      "step": 19939
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.531312750751632e-06,
+      "loss": 0.9749,
+      "step": 19940
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.530369487011599e-06,
+      "loss": 0.9384,
+      "step": 19941
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.529426292706614e-06,
+      "loss": 1.0423,
+      "step": 19942
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5284831678486495e-06,
+      "loss": 0.8902,
+      "step": 19943
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.52754011244968e-06,
+      "loss": 0.9259,
+      "step": 19944
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5265971265216714e-06,
+      "loss": 1.005,
+      "step": 19945
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.525654210076599e-06,
+      "loss": 0.9404,
+      "step": 19946
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5247113631264365e-06,
+      "loss": 1.0504,
+      "step": 19947
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5237685856831495e-06,
+      "loss": 0.989,
+      "step": 19948
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.522825877758707e-06,
+      "loss": 0.9123,
+      "step": 19949
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.521883239365071e-06,
+      "loss": 0.9567,
+      "step": 19950
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5209406705142165e-06,
+      "loss": 0.9816,
+      "step": 19951
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519998171218105e-06,
+      "loss": 0.8719,
+      "step": 19952
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.519055741488697e-06,
+      "loss": 0.9695,
+      "step": 19953
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.51811338133796e-06,
+      "loss": 0.8589,
+      "step": 19954
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.517171090777863e-06,
+      "loss": 0.9472,
+      "step": 19955
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.516228869820363e-06,
+      "loss": 0.9729,
+      "step": 19956
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.515286718477418e-06,
+      "loss": 0.9201,
+      "step": 19957
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.514344636760994e-06,
+      "loss": 1.0416,
+      "step": 19958
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.513402624683042e-06,
+      "loss": 0.8562,
+      "step": 19959
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.512460682255526e-06,
+      "loss": 0.8633,
+      "step": 19960
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.511518809490407e-06,
+      "loss": 0.872,
+      "step": 19961
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.510577006399639e-06,
+      "loss": 0.8633,
+      "step": 19962
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5096352729951765e-06,
+      "loss": 1.0819,
+      "step": 19963
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.508693609288972e-06,
+      "loss": 0.8877,
+      "step": 19964
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5077520152929865e-06,
+      "loss": 0.9043,
+      "step": 19965
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.506810491019165e-06,
+      "loss": 1.0033,
+      "step": 19966
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.505869036479469e-06,
+      "loss": 0.9645,
+      "step": 19967
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5049276516858395e-06,
+      "loss": 1.0203,
+      "step": 19968
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5039863366502375e-06,
+      "loss": 0.981,
+      "step": 19969
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.503045091384609e-06,
+      "loss": 0.8674,
+      "step": 19970
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5021039159009e-06,
+      "loss": 0.9956,
+      "step": 19971
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.501162810211062e-06,
+      "loss": 0.9384,
+      "step": 19972
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.5002217743270345e-06,
+      "loss": 1.0201,
+      "step": 19973
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.499280808260771e-06,
+      "loss": 1.0803,
+      "step": 19974
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.498339912024218e-06,
+      "loss": 1.1322,
+      "step": 19975
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.497399085629316e-06,
+      "loss": 0.9346,
+      "step": 19976
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.496458329088006e-06,
+      "loss": 0.8847,
+      "step": 19977
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 4.495517642412237e-06,
+      "loss": 0.9151,
+      "step": 19978
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.494577025613947e-06,
+      "loss": 1.0571,
+      "step": 19979
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.493636478705075e-06,
+      "loss": 0.9986,
+      "step": 19980
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.492696001697568e-06,
+      "loss": 0.9559,
+      "step": 19981
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.491755594603354e-06,
+      "loss": 0.9854,
+      "step": 19982
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.490815257434382e-06,
+      "loss": 1.0549,
+      "step": 19983
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.489874990202585e-06,
+      "loss": 1.0084,
+      "step": 19984
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.488934792919899e-06,
+      "loss": 0.8897,
+      "step": 19985
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.48799466559826e-06,
+      "loss": 0.9963,
+      "step": 19986
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.487054608249597e-06,
+      "loss": 0.9245,
+      "step": 19987
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.486114620885849e-06,
+      "loss": 0.8457,
+      "step": 19988
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.485174703518954e-06,
+      "loss": 1.0036,
+      "step": 19989
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.484234856160838e-06,
+      "loss": 0.9691,
+      "step": 19990
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.483295078823428e-06,
+      "loss": 0.8639,
+      "step": 19991
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.482355371518662e-06,
+      "loss": 0.9533,
+      "step": 19992
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.481415734258467e-06,
+      "loss": 1.042,
+      "step": 19993
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.480476167054765e-06,
+      "loss": 0.9795,
+      "step": 19994
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.479536669919493e-06,
+      "loss": 0.9016,
+      "step": 19995
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.4785972428645705e-06,
+      "loss": 0.8414,
+      "step": 19996
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.477657885901929e-06,
+      "loss": 1.0458,
+      "step": 19997
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.476718599043491e-06,
+      "loss": 0.9738,
+      "step": 19998
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.47577938230118e-06,
+      "loss": 0.91,
+      "step": 19999
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 4.474840235686914e-06,
+      "loss": 1.0667,
+      "step": 20000
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 28746,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 10000,
+  "total_flos": 2270571145052160.0,
+  "train_batch_size": 8,
+  "trial_name": null,
+  "trial_params": null
+}